CAMRADIIvolume II

CAMRAD II
COMPREHENSIVE ANALYTICAL MODEL OF

ROTORCRAFT AERODYNAMICS AND DYNAMICS
Volume II: Components Theory
Wayne Johnson Distributed by
Johnson Aeronautics AMI Aero, LLC

Palo Alto, California Bellevue, Washington
CAMRAD II
COMPREHENSIVE ANALYTICAL MODEL OF
ROTORCRAFT AERODYNAMICS AND DYNAMICS
Volume II: Components Theory
Release 4.11
Wayne Johnson
Johnson Aeronautics
Copyright
c 1992–2019 Wayne Johnson
Published by Johnson Aeronautics, Palo Alto, California
This volume is part of the Documentation of the proprietary computer software CAMRAD II, provided to Licensee
under a software license agreement. Documentation and all information and trade secrets contained therein are
Confidential Information. Confidential Information constitutes valuable trade secrets and proprietary property of
Johnson Aeronautics.
Notice of Proprietary Rights: This document discloses subject matter in which Johnson Aeronautics has proprietary
rights. Neither receipt nor possession of this document confers or transfers any right to reproduce or disclose the
document, any part thereof, or any information contained therein, except by written permission from or written
agreement with Johnson Aeronautics.
LIMITED RIGHTS LEGEND: The restrictions governing the use and disclosure of technical data marked with this
legend are set forth in the definition of “Limited Rights” in paragraph (a)(15) of the clause at DFARS 252.227-7013.
The limited rights are not subject to an expiration date. Manufacturer: Johnson Aeronautics, PO Box 1253, Palo
Alto, California 94302.
CONTENTS
1. INTRODUCTION 1
1–1 Background 1
1–2 Approach 2
1–3 Nomenclature 3
1–4 General Warnings and Limitations 5
2. COMPONENTS 7
3. STRUCTURAL DYNAMIC COMPONENTS 11

3–1 Description 11
3–2 Component Variables 11
3–3 Implementation 13
3-3.1 Frames and Rigid Motion 13
3-3.2 Geometry and Kinematics 13
3-3.3 Joints 17
3-3.4 Structural Dynamic Interface 18
3-3.5 Aerodynamic Interface 19
3-3.6 Applied Load Interface 19
3-3.7 Controls 21
3-3.8 Sensors 21
3-3.9 Spring/Damper/Actuator 21
3–4 Theory 21
3-4.2 Joints 23
3-4.3 Aerodynamic Interface 27
3-4.4 Sensors 28
3-4.5 Reaction for Equations of Motion 32
3-4.6 Spring/Damper/Actuator 32
4. RIGID BODY COMPONENT 39

4–2 Theory 39
4-2.1 Rigid Body Equations of Motion 39
5. LINEAR NORMAL MODES COMPONENT 41

iv CONTENTS
5-2.1 Linear Normal Modes 41

5-2.2 Aerodynamics 42
5-2.3 Modal Sensors 42
5–3 Theory 42
5-3.2 Locations and Elastic Motion 43
5-3.3 Elastic Equations of Motion 43
5-3.4 Sensors 44
6. FINITE ELEMENT BEAM COMPONENT 45

6-2.1 Beam Model 47
6-2.2 Rigid Body Motion 48
6-2.3 Beam Properties 48
6-2.4 Locations, Joints, and Connections 50
6-2.5 Sensors 50
6–3 Theory 51
6-3.2 Extension and Torsion Produced by Bending 56
6-3.3 Elastic Variables and Shape Functions 59
6-3.4 Integration Over Beam Length 62
6-3.5 Beam Cross-Section Motion 63
6-3.6 Hamilton’s Principle 63
6-3.7 Strain Energy 65
6-3.8 Extension/Torsion Coupling 72
6-3.9 Kinetic Energy 72
6-3.12 Sensors 79
6-3.13 Cross-Section Structural Properties 82
6–4 References 82
7. ROD/CABLE COMPONENT 85
7–2 Theory 85
7-2.4 Sensors 88
8. TRANSMISSION COMPONENT 89
CONTENTS v

8-3.2 Interfaces 92
8-3.3 Controls 93
8-3.4 Sensors 93
8–4 Theory 93
8-4.1 Rigid Motion 93
8-4.2 Rigid Equation of Motion 94
8-4.4 Sensors 94
9. REFERENCE FRAME COMPONENT 97

9-3.1 Functions 97
9-3.2 Sensors 97
9–4 Theory 99
9-4.1 Definition of Axes T 99
9-4.2 Sensors 99
10. FILTER COMPONENT 105

10–4 Theory 107
11. REFERENCE PLANE COMPONENT 111

11-3.1 Geometry 113
11-3.2 Identification Method 113
11-3.3 Filter Model 114
11–4 Theory 114
11-4.1 Component Equations 114
11-4.2 Geometry 114
11-4.3 Plane Through Position Identification Method 115
11-4.4 Harmonic Identification Methods 115
12. DIFFERENTIAL EQUATION COMPONENT 119

vi CONTENTS

12-3.2 Differential Equation Input 121
12-3.3 Static Equation 121
12–4 Theory 121
12-4.1 Alternate Input Form 121
12-4.2 Output Transformation 122
13. PROGRAMMABLE COMPONENT 125

13-3.2 Differential Equation Input 127
13-3.3 User-Defined Calculations 127
13–4 Theory 128
13-4.1 Transformation from Input Form 128
14. TRANSFER FUNCTION COMPONENT 131

14-3.1 Transfer Function 131
14-3.2 Transfer Function Input and Output 131
14–4 Theory 133
14-4.1 Transfer Function 133
14-4.2 Differential Equation 133
15. FOURIER SERIES COMPONENT 137

15–4 Theory 139
16. PRESCRIBED CONTROL COMPONENT 141

16–4 Theory 143
16-4.1 Derivatives and Integrals 143
16-4.2 General Time History 146
CONTENTS vii
16-4.3 Random Time History 147
17. GUST COMPONENT 149

17-3.1 Geometry 149
17-3.2 Gust Models 151
17–4 Theory 151
17-4.1 Geometry 151
17-4.2 Elementary Gust Models 153
17-4.3 Prescribed Gust Model 153
17-4.4 Tabular Gust Model 155
18. AERODYNAMIC COMPONENTS 157

18–2 Wings and Bodies 157
18–3 Inflow and Wake 157
18–4 Wake Geometry 160
18–5 Solution Procedure 160
18-5.1 Aerodynamic Solution Partition 160
18-5.2 Wake Geometry Loop 160
18-5.3 Circulation Loop 162
19. RIGID AIRFRAME AERODYNAMICS COMPONENT 163

19-3.1 Geometry and Axes 165
19-3.2 Aerodynamic Models 165
19-3.3 Nonlinear Model 167
19-3.4 Stability Derivatives Model 167
19-3.5 Airframe Lift/q 168
19-3.6 Controls 168
19-3.7 Sensors 168
19–4 Theory 168
19-4.1 Nonlinear Model 168
19-4.2 Stability Derivatives Model 172
19–5 References 174
20. AIRFRAME FLOW FIELD COMPONENT 177

viii CONTENTS

20-3.2 Velocity From Table File 179
20-3.3 Calculated Velocity 179
20-3.4 Sensors 182
20–4 Theory 182
20-4.1 Calculated Velocity, Wings 182
20-4.2 Calculated Velocity, Bodies 186
21. LIFTING LINE WING COMPONENT 193

21-3.1 Lifting-Line Theory 197
21-3.2 Practical Implementation of Lifting-Line Theory 201
21-3.4 Trailing-Edge Flaps 205
21-3.5 Airfoil Table 205
21-3.6 Induced and Interference Division 206
21-3.7 Controls 206
21-3.8 Sensors 207
21-3.9 External Aeroacoustic Analysis 208
21–4 Theory 210
21-4.1 Wing Geometry 210
21-4.2 Section Motion and Loads 216
21-4.3 Section Aerodynamic Model 225
21-4.4 Unsteady Airfoil Motion 234
21-4.5 Dynamic Stall 242
21-4.6 Solving the State Equations 254
22. WING INFLOW COMPONENT 259

22-3.1 Geometry and Frames 261
22-3.2 Component Input 261
22-3.3 Uniform Inflow Model 261
22-3.4 Dynamic Inflow Model 262
22–4 Theory 262
22-4.1 Force and Velocity 262
CONTENTS ix

22-4.4 Lagged Inflow Model 264
23. ROTOR INFLOW COMPONENT 267

23–4 Theory 271
23-4.1 Force and Velocity 271
23-4.3 Vortex Ring State 277
23-4.5 Lagged Inflow Model 283
23-4.6 Ducted Fan 283
24. ROTOR DYNAMIC WAKE COMPONENT 289

24-3.3 Dynamic Wake Model 291
24–4 Theory 292
24-4.1 Geometry, Velocity, and Force 292
24-4.2 Dynamic Wake Model 294
24-4.3 Rotor Loading and Generalized Forces 302
24-4.4 Wake-Induced Velocity 303
24-4.5 Dynamic Inflow 304
25. WING WAKE COMPONENT 307

25-3.1 Vortex Wake Analysis 308
x CONTENTS
25-3.5 Induced Velocity and Influence Coefficients 341

25–4 Theory 345
25-4.1 Induced Velocity Calculation 345
25-4.2 Influence Coefficient Calculation 346
25-4.3 Vortex Line Segment 361
25-4.4 Vortex Sheet Element 364
25-4.5 Circular-Arc Vortex Segment 369
25-4.6 Ground Effect 374
26. WAKE GEOMETRY COMPONENTS 379

26-2.1 Character of Wake Geometry 379
26-2.2 Lifting Line Wing 380
26-2.3 Description of Geometry 380
26–3 Theory 383
26-3.1 Description of Geometry 383
26-3.2 Betz Rollup 385
26-3.3 Interpolation and Extrapolation of Distortion 385
26-3.4 Lifting Line Wing 386
26-3.5 Display Geometry 387
27. WING WAKE GEOMETRY COMPONENT 389

27-3.1 Character of Wing Wake Geometry 391
27-3.4 Wake Geometry Models 393
27–4 Theory 394
27-4.1 Rigid Distortion 394
27-4.2 Prescribed Distortion 394
27-4.3 Free Distortion 395
28. ROTOR WAKE GEOMETRY COMPONENT 413

28-3.1 Character of Rotor Wake Geometry 415
CONTENTS xi

28-3.4 Wake Geometry Models 418
28–4 Theory 418
28-4.1 Rigid Distortion 418
28-4.2 Prescribed Distortion 419
28-4.3 Scully Free Distortion 421
28-4.4 Johnson Free Distortion 423
28-4.5 Comparison of Algorithms 424
29. WING PERFORMANCE COMPONENT 427

29-3.4 Sensors 429
29–4 Theory 430
29-4.1 Ideal Induced Velocity 430
29-4.2 Sensors 430
30. ROTOR PERFORMANCE COMPONENT 435

30-3.4 Controls 438
30-3.5 Sensors 438
30–4 Theory 439
30-4.1 Definition of Axes T and C 439
30-4.2 Ideal Induced Velocity 439
30-4.3 Sensors 440
30-4.4 Ducted Fan 446
31. ROTORCRAFT PERFORMANCE COMPONENT 449

31-3.1 Operating Condition 449
xii CONTENTS

31-3.4 Sensors 451
31–4 Theory 451
31-4.1 Sensors 451
32. RIGID WING COMPONENT 455

32–4 Theory 458
33. HELICOPTER TAIL BOOM COMPONENT 461

33-3.1 Geometry 463
33-3.2 Airfoil Table 463
33-3.3 Controls 464
33-3.4 Sensors 464
33–4 Theory 465
33-4.1 Boom Geometry 465
33-4.2 Section Motion and Loads 467
33-4.3 Induced Angle of Attack 469
33-4.4 Reaction Jet Load 470
33-4.5 Total Loads 470
33-4.6 Simple Aerodynamic Model 470
34. COMPUTATIONAL FLUID DYNAMICS COMPONENT 473

34-3.1 General CFD Interface 475
34–4 Theory 477
35. PLUGIN COMPONENT 481

35-2.1 Plugin Calculations 481
CONTENTS xiii
35-2.2 Shell Plugin 482

35-2.3 Core Component Input 483
35-2.4 Initialization 483
35-2.5 Solution Process 483
35-2.6 Utilities 484
xiv CONTENTS
Chapter 1
INTRODUCTION
CAMRAD II is an analysis of aeromechanical systems that incorporates a combination of advanced tech-

nology, including multibody dynamics, nonlinear finite elements, structural dynamics, and aerodynamics. It is
a comprehensive analysis: for the design, testing, and evaluation of systems — at all stages, including research,
conceptual design, detailed design, and development — CAMRAD II calculates performance, loads, vibration,
response, and stability — with a consistent, balanced, yet high level of technology in a single computer program
— applicable to a wide range of problems, and a wide class of configurations. Such an approach is essential for
many modern problems, which are inherently complex and multidisciplinary.
CAMRAD II uses a building block approach to achieve flexibility in the model of the dynamic and aerodynamic
configuration. Hence it can model the true geometry of a system, including multiple load paths. CAMRAD II
provides a powerful analysis capability, including advanced aerodynamics; rigorous kinematics and dynamics
(with consistent structural loads and dynamic response, and general interfaces between aerodynamic and structural
dynamic components); and general transient solutions. For ease of use a shell is provided to build typical system
models, while the core input capability always gives complete flexibility to define and revise the model. A range
of components and modelling options makes it a practical engineering tool, allowing the best balance of efficiency
and accuracy to be found for a particular problem. CAMRAD II offers a design for growth that makes it an
appropriate platform for future developments, for continuing access to new technology.
This volume of the CAMRAD II documentation describes the theoretical basis for the components of the
analysis.
1–1 Background
CAMRAD II incorporates a combination of advanced technology. The existing technology was not found
sufficient however, individually or in combination, to solve the required problems, as designers take modern systems
rapidly beyond the capability of present analyses. Yet there is much to build on in this background technology.
Multibody dynamics technology provides a background on rigid body components, frames, and joints; non-
linear kinematics; and joint operations. In addition however, it is necessary to deal with elastic motion.
Structural dynamics technology provides a background on substructure coupling and static residuals; modal
analysis and truncation; and elimination of constraints. Most of this work is for linear systems, however; here it
is necessary to deal with nonlinear systems.
Nonlinear finite element technology provides a background on nonlinear elements, numerical integration,
and beam components. In addition however, it is necessary to deal with large motion. Hence the present approach
2 INTRODUCTION
diverges from finite elements by not using nodal motion as degrees of freedom.
In general the problems of interest here involve more than just structural dynamics. In particular, there is no
system Lagrangian. For numerous reasons, CAMRAD II also uses explicit constraint forces (for example, in order
to handle partitioned solutions, static residuals, input/output constraints, and evaluation of reactions).
Rotorcraft aeromechanics provides a background on aircraft dynamics, rotating systems, and aerodynamics.
This field also helps define the required tasks and results of the analysis.
The configuration and operation of modern systems requires an analysis and design tool capable of han-
dling complex configurations with unusual load paths and interactions, and with many subsystems; structural,
aerodynamic, and kinematic nonlinearities; arbitrary large motion, including rigid body motions and rotation of
components relative to each other; and components that are not defined by the equations and interfaces of structural
dynamics. The analysis must be flexible, which demands that the specification of the configuration, the aerome-
chanical models, and the solution procedures be separated. The analysis must be designed for growth, so it can
incorporate future technology developments. It is also important that the analysis be easy to use, especially for
normal configurations. A final requirement is that the analysis be practical: efficient, accurate, and reliable.
1–2 Approach
CAMRAD II performs a nonlinear dynamic/static analysis of an aeromechanical system. Flexibility and

generality of the system configuration are obtained by assembling standard components with standard interfaces,
and solving the system by standard procedures. The basic approach of the analysis is to make no approximations
at the highest levels. A finite-dimension description of the system is required. The components and interfaces
introduce spatial discretization; the solution procedures introduce time discretization. No further approximations
are made at the top level. The coupling and solution procedures can handle arbitrary and exact models, including
nonlinear and time-varying equations. Approximations are possible (indeed required for most practical solutions),
but enter at the component level. With this approach, new technology and more accurate models can be developed
and implemented in the form of new components, without changing the framework of the analysis.
For configuration generality, the analysis splits the system into pieces, with connections between. The
following pieces are available to construct the system:
environmental physical logical

case component loop
wind frame part
operating condition interface transform
period output modes
input response
weights
These system pieces constitute the core analysis, providing a flexible, building-block oriented modelling capability.
There are physical pieces (components and interfaces) and logical pieces (solution procedure). The components
available include structural dynamic, aerodynamic, differential equation, and performance models. Standard
descriptions of components, interfaces, and solution procedures are required. Structural dynamic systems are an
important subset of the problems of interest, but other types of components are equally important, particularly
INTRODUCTION 3
aerodynamic models. The conventions and methodology must encompass all types of components, interfaces, and
solution procedures; and yet also accommodate the conventional approaches for specific types.
The analysis solves differential, integral, static, and implicit equations for the motion of the system, and
evaluates required output quantities from the response. Figure 1 illustrates the analysis tasks: trim, transient, and
flutter. The trim task finds the equilibrium solution (constant or periodic) for a steady state operating condition.
Often it is necessary for the trim task to identify parameter values required to achieve a specified operating condition.
The transient task integrates the equations in time, from trim, for a prescribed excitation. The flutter task obtains
and analyzes differential equations, linearized about trim. With a quasistatic reduction of the system, the flutter
task can produce a stability derivative model.
It is important that the analysis be easy to use, especially for normal configurations. Modelling a large system
requires definition of many core pieces (particularly interface and response pieces). Hence the analysis has a shell
to facilitate application to specific problems. The shell constructs the core input for typical configurations and
typical problems. The objective is to isolate the user from details of the system definition at the core level. The
shell creates the components and other system pieces required, from parametric input and from assumptions about
the system configuration and model. By using such a shell, it is typically possible to accomplish 2–3 orders of
magnitude compression of the amount of input data supplied by the user.
The shell does not have the flexibility of the core input, and may not be able to model exactly every config-
uration. The shell constructs most of the system, minimizing the need to deal directly with the core input; and
what the shell constructs will provide guidance for the use of the analysis. The user can still use the core input to
change the model constructed by the shell, as required for specific problems. In addition, there is no need for the
shell input to implement features that can be obtained by simple changes to the core input (as long as the shell at
least creates the system piece).
1–3 Nomenclature
The nomenclature for kinematics of rigid motion employs the following conventions. A vector x is a column
matrix of three elements, measuring the vector relative to a particular basis (or axes, or frame). The basis is
indicated as follows:
a) xA is a vector measured in axes A;

b) xEF/A is a vector from point F to point E, measured in axes A.
A rotation matrix C is a three-by-three matrix that transforms vectors from one basis to another:
c) C BA transforms vectors from basis A to basis B, so xB = C BA xA .
The matrix C BA defines the orientation of basis B relative basis A, so it also may be viewed as rotating the axes
is defined as follows:
from A to B. For a vector u, a cross-product matrix u
⎡ ⎤
0 −u3 u2
= ⎣ u3
u 0 −u1 ⎦
−u2 u1 0
v is equivalent to the vector cross-product u × v. Note that outside the context of rigid motion
such that u
kinematics, the tilde symbol does not imply this matrix. The cross-product matrix enters the relation between
4 INTRODUCTION
solve equations (differential, integral, static, implicit) for motion of system

evaluate required output quantities from response
equilibrium solution
steady or periodic RESPONSE
steady state operating condition TRIM and
identify parameters to achieve OUTPUT
specified operating condition
integrate in time RESPONSE

from trim TRANSIENT and
for prescribed excitation OUTPUT
differential equations
linearized about trim RESPONSE
full dynamics or FLUTTER
quasistatic reduction
and
including stability derivative OUTPUT
representation
Figure 1-1 CAMRAD II tasks.

INTRODUCTION 5
angular velocity and the time derivative of a rotation matrix:
Ċ AB = − BA/B
ω AB/A C AB = C AB ω
(the Poisson equations). For rotation by an angle α about the x, y, or z axis (1, 2, or 3 axis), the following notation
is used: ⎡ ⎤
1 0 0
Xα = ⎣ 0 cos α sin α ⎦
0 − sin α cos α
⎡ ⎤
cos α 0 − sin α
Yα = ⎣ 0 1 0 ⎦
sin α 0 cos α
⎡ ⎤
cos α sin α 0
Zα = ⎣ − sin α cos α 0 ⎦
0 0 1
Thus for example, C BA = Xφ Yθ Zψ means that the axes B are located relative the axes A by first rotating by angle
ψ about the z-axis, then by angle θ about the y-axis, and finally by angle φ about the x-axis. Euler-Rodrigues
parameters (or Rodrigues parameters) are defined by the relation p = 2u tan ψ/2, for rotation by angle ψ about
an axis in the direction of the unit vector u. This differs by a factor of 2 from the classical definition, but has
the advantage that p =∼ uψ for small rotations. Consider the relative motion of axes A and B. Let xBA/A be the
displacement from the origin of A to the origin of B, measured in A axes; and C BA the rotation from axes A to
axes B. Time derivatives of the motion of B relative A can be described by body axis velocity and angular velocity.
Thus the following standard convention is used for a description of the rigid motion of axes B relative axes A:
a) linear motion: xBA/A and v BA/B = C BA ẋBA/A and v̇ BA/B

BA/B = C BA Ċ AB and ω̇ BA/B
b) angular motion: C BA and ω
The axes A and B might be frames, or rigid body degrees of freedom, or the position of some point on a structure.
1–4 General Warnings and Limitations
The state-of-the-art of engineering is such that it is not yet possible to accurately calculate or predict aerome-
chanical system behavior in all circumstances. Any computer program must be used in that context. It is always
possible to get a bad answer from bad input or misuse of an analysis. It is also possible that the best results of
an analysis may not be accurate. A computer program is a tool, not substituting for but rather to be used with
judgement, experience, and much testing of the actual system.
The analysis models a system by dividing it into components. Each component is a model of the actual device.
Rarely are the complete equations being solved for the exact response. Rather empiricism and approximations
are required in order to make the analysis practical. Hence engineering judgement is required to use and interpret
the analysis, based on experience with the code, particularly projects using the code to correlate with measured
results.
6 INTRODUCTION
Chapter 2
COMPONENTS
For configuration generality, the analysis splits the system into pieces — physical pieces (components and
interfaces) and logical pieces (solution procedure) — with connections between. The components perform all
computations associated with the physics of the model of a system. The following components are available to
construct the system:
rigid body Fourier series wing wake geometry

linear normal modes prescribed control rotor wake geometry
finite element beam gust
rod/cable wing performance
rigid airframe aerodynamics rotor performance
transmission airframe flow field rotorcraft performance
reference frame
lifting line wing helicopter tail boom
filter rigid wing computational fluid dynamics
reference plane
wing inflow
differential equation rotor inflow
programmable rotor dynamic wake
transfer function wing wake plugin
This volume describes the theoretical model for each of the components.
A component can have degrees of freedom ξi , where i enumerates the degree of freedom vectors. Dependence
on degrees of freedom means that equations of motion exist, and usually a differential equation formulation is
possible. The component can depend on frame motion β. The component has input fi and output xj , where i
and j enumerate the input and output vectors. The component input is of structural dynamic kind or input/output
kind; it can be connected to an interface or to a system input piece. The component output is of structural dynamic
kind or input/output kind; it can be connected to an interface or to a system output piece. For a structural dynamic
interface, the input and output occur in pairs; the input is the vector of force and moment at the connection, and
the output is the axes motion at the connection.
Hence the component produces motion and output equations, which depend on the degrees of freedom, frame
motion, and input:
0 = Aj (ξi , β, fi )
xj = Bj (ξi , β, fi )
All the variables are vectors; each vector has one or more elements. In general these equations are nonlinear and
time varying. If the component does not have degrees of freedom, the motion equations do not exist.
8 COMPONENTS
A component can be considered an operator that evaluates a vector or matrix. Figure 1 illustrates the component
functionality. Specifically, at time t a component evaluates one of the following vectors:
a) motion equation, Aj ;
b) component output for input/output interface, xj = Bj ;
c) component output for structural dynamic interface, axes motion xj ;
from the degrees of freedom ξi (including frame motion) and from the component input fi . The component can
also perturb this vector (analytically or numerically), to construct a column of a matrix.
A component must provide the analysis information about its functionality: how the equations of motion and
output equations depend on the frame motion, degrees of freedom, and component input. The documentation for
each component type concludes with figures that describe the functionality. Figure 2 illustrates the format. The
rows correspond to the equations, and the columns correspond to the variables. The dependence is identified by
vector. This information is used by the analysis in several ways, including generation of matrices and tests for loop
convergence. The identification should be as complete as possible. If some true functionality is not identified,
in particular the corresponding submatrix is omitted when the equations are linearized. It is simplest to specify
complete functionality: every equation depending on all variables, and second-order for degrees of freedom. Such
an approach can be too inefficient however, in both computational effort and data structure size. Moreover the
analysis tasks and solution procedures may make assumptions about the functionality, particularly for structural
dynamic equations, that will not be satisfied if complete functionality is assumed.
A frame must be specified for structural dynamic components. For other component types a frame may not be
required; in such cases the frame dependence is omitted from the functionality description. The analysis allows the
component to identify its frame dependence up to some reference frame. Unless otherwise stated, this capability
is not used: this reference is assumed to be the inertial frame. Then dependence on a frame means dependence on
the degrees of freedom of the component frame, and on the degrees of freedom of all its parent frames.
COMPONENTS 9
component
type = TTTT
component input name = NNNN component output
IO f IO x = B
component input component output
SD f SD x
(constraint force) (axes motion)
degree of freedom ξ motion equation

frame β
A
Figure 2-1 Component functionality.

10 COMPONENTS
frame β degree of freedom ξ component input f
equation of motion A yes (0,1,2) or no yes or no
output equation x = B yes (0,1,2) or no yes or no
Dependence on frame motion and degrees of freedom: displacement, or velocity

and displacement, or acceleration and velocity and displacement. Designated static
(0), or first order (1), or second order (2) respectively.
The dependence on component input can be specified separately for the trim, tran-
sient, and flutter tasks. “X” indicates a dependence that is impossible.
The analysis asks for information about an equation/variable combination only if
both the equation and the variable exist for the specific use of the component.
Figure 2-2 Component functionality.

Chapter 3
STRUCTURAL DYNAMIC COMPONENTS
3–1 Description
All structural dynamic components of the analysis share common characteristics, including:
a) rigid body motion and frame;

b) mass, hence inertial and gravitational forces;
c) joints;
d) structural dynamic interfaces;
e) aerodynamic interfaces;
f) applied load interfaces;
g) controls and sensors.
Hence all structural dynamic components have a standard form that implements these characteristics. The compo-
nent motion is described by rigid body motion, which can always be large; plus elastic motion, which is measured
relative the rigid motion. The structural dynamic components differ in regard to their modelling of the elastic
motion.
Exact kinematics are used in the standard relations, for all structural dynamic components. In particular,
each component has rigid body motion that can be large without restriction, and exact kinematics are maintained
for all structural dynamic interfaces. A structural dynamic component typically makes some approximations
in its representation of the elastic motion. Thus for a given motion of the rigid body axes and the connection
axes, there will be approximations to the kinetic and potential energy associated with the elastic motion. But no
approximations are made in the rigid body or connection kinematics, so the interfaces between components remain
correct.
3–2 Component Variables
Figure 1 illustrates the standard functionality of structural dynamic components.
Degrees of Freedom
The component can have up to three degree-of-freedom vectors: rigid motion, elastic motion, and joint motion.
The rigid motion vector is absent for a constrained component. The number of joint degrees of freedom (length
of the joint motion vector) need not equal the number of joint variables. The elastic motion vector depends on the
component type.
12 STRUCTURAL DYNAMIC COMPONENTS
controls degrees of sensors

(for joint, actuator, freedom
aux force) rigid
elastic
joint
aerodynamic interface aerodynamic interface
interference, gust velocity frame velocity relative air
force, moment position relative origin axes
force, torque
moment
rigid rotation
motion
structural dynamic structural dynamic

interface (complete) interface (torque)
Figure 3-1 Functionality of structural dynamic components.

STRUCTURAL DYNAMIC COMPONENTS 13
Component Input
a) Structural dynamic interface: the input is force and moment for a complete interface; or torque for a torque
interface.
b) Aerodynamic interface: the input can be force, moment, interference velocity, or gust velocity.
c) Controls: for use by joints, actuators, or applied loads.
Component Output
a) Structural dynamic interface: the output is axes motion for a complete interface; or rotation for a torque interface.
b) Aerodynamic interface: the output can be the velocity relative to the air, or the position relative the origin of
the axes.
c) Sensors.
3–3 Implementation
3-3.1 Frames and Rigid Motion
A frame F (perhaps inertial) must be identified for a structural dynamic component. All motion of the
component is measured relative to that frame. The component has rigid body motion, described by the motion
of specified body axes B. For a “constrained component,” this motion is connected to the frame, and there are
no rigid body degrees of freedom. For a “frame component,” the rigid body degrees of freedom are the frame
motion. In other cases, the rigid body degrees of freedom exist, and represent motion relative to the frame. Figure
2 summarizes the options for the rigid body motion. The rigid body motion can consist of just the linear degrees of
freedom, or both linear and angular degrees of freedom. It is constant for a constrained component (no degrees of
freedom). Ignoring the angular motion is appropriate for a point mass (no moments of inertia or center-of-gravity
offset) with only pinned interfaces.
3-3.2 Geometry and Kinematics
A convention is required to describe the geometry and kinematics of a structural dynamic component. In-
terfaces occur at connections (C) on the structure, hence every load acts at a connection. Geometry is specified
in terms of locations (E) on the structure. Figure 3 summarizes the construction of motion at a connection. The
position of a connection must be described, ultimately relative the inertial frame (I). That position is calculated as
a series of steps, symbolically:
I → P → F → B → E → J → C
For the kinematics of rigid motion, the position of each step consists of the orientation and displacement (and
derivatives if required) of an axis system, measured relative the axes of the preceding step. The addition of steps
involves the addition of axes motion, not scalar summation. Each step may itself consist of several motions. The
definition of the steps is as follows.
a) Inertial frame I: general reference for all motion.
b) Parent frame P: some parent frame of the component. PI is the sum of all frame motions between P and the
inertial frame. Each parent frame might be frame degrees of freedom (rigid body degrees of freedom of some
other component).
CONSTRAINED COMPONENT
body
axes B
component constant
frame F position BF
(displacement
and rotation)
FRAME COMPONENT
component frame F
= body axes B
motion BP = FP =
nominal + degree of freedom
OTHER CASES
body
axes B
component
motion BF =
frame F
nominal + degree of freedom
Figure 3-2 Rigid body motion of structural dynamic component.

CI = CJ + JE + EB + BF + FP + PI

PI = Pi Pi−1 = (nominal + dof)
FP = (nominal + dof); or (dof + nominal)
BF = (nominal + dof); or (dof + nominal);
or 0 (frame degrees of freedom);
or constant BF (constrained component)
EB = EL + LB; or constant EB (no elastic)
JE = JD + DE; or 0 (no joint)
CJ = constant CJ; or constant CE (no joint)
Figure 3-3a Summary of motion at a connection.
c) Frame F: designated frame of the component. FP is the motion of the frame relative its parent. FP itself is the
sum of the nominal and degree-of-freedom motions (in either order, standard rigid-type response).
d) Body motion B: rigid body motion of the component. BF is the motion of the component body axes relative to
its frame. BF itself is the sum of the nominal and degree-of-freedom motions (in either order, standard rigid-type
response).
Frame component: the body axes of this component are the frame (B = F). Hence the frame
motion FP provides the rigid body degrees of freedom of the component, while the motion BF
is zero.
Constrained component: the body axes of this component are fixed relative to the frame. The
constant position of the body relative to the frame is specified by z BF/F and C BF (B relative
F, in F axes). Note that if the component is not constrained, these quantities are contained
instead in the nominal response of the rigid body degrees of freedom.
e) Elastic motion E: geometric location and elastic motion. EB is the position of the location relative the body axes,
z EB/B and C EB (E relative B, in B axes). EB is the sum of elastic degrees of freedom and a constant position
(EB = EL + LB). The analysis has the option to suppress the effects of the elastic motion or the elastic force at a
location (an approximation).
f) Joint motion J: joint motion of the component (only present if a joint exists at the location). JE is the position
of the joint relative the location axes, z JE/E and C JE (J relative E, in E axes). JE is the sum of joint degrees of
freedom and a constant position (JE = JD + DE).
g) Connection C: position of a connection on the component. The constant position of the connection relative the
joint axes is specified by z CJ/J and C CJ (C relative J, in J axes). If the connection is not at a joint, then CJ is
actually the position relative the location axes (C relative E, in E axes), although the notation is not changed.
Any subset of this motion can be calculated, as required by the component analysis. For each component, it is
possible to define any number of locations; any number of joints at a location; any number of connections at a
interfaces
and sensors
C = connection axes
constant
position
CJ
J = joint axes
JE
joint
E = location axes
degrees of
D freedom
constant
position
elastic
degrees of EB
freedom
L
constant position
degree of
B = rigid freedom
body axes nominal
BF
F = component
frame
FP
degree of
freedom
degree of nominal
freedom
nominal
P = parent frame
PI
I = inertial frame
Figure 3-3b Summary of motion at a connection.

location, or at a joint; any number of uses of a connection, including interfaces and sensors. There are some
restrictions however. For example, aerodynamic interfaces can not be defined at a joint.
3-3.3 Joints
A general convention for defining joints is used for all structural dynamic components. The joint motion
occurs after the rigid body and elastic motion; more complex motions can be accommodated by breaking the
system up into more components. It is assumed that the component has no mass on the connection side of the
joint. Then the joint equations of motion can be obtained by simply considering balance of forces acting on the
joints. It also follows that there must be at least one interface producing forces on the joint. The joint kinematics
describe the position of the joint axes relative to a location (JE). JE is the sum of joint degrees of freedom and a
constant position (JE = JD + DE). The constant position of the joint (the position for zero joint variables) relative
the location axes is specified by z DE/E and C DE (D relative E, in E axes). Then the joint motion occurs, described
by xJD/D and C JD (and their derivatives).
A joint can consist of just linear motion; just angular motion; linear followed by angular motion; or angular
followed by linear motion. The linear motion can have one of the following configurations:
a) 1-linear (slide, prism, linear hinge);

b) 2-linear (plane);
c) 3-linear (space);
described by one to three variables (with gain factors). The axes of the linear motions can be arbitrarily specified.
The angular motion can have one of the following configurations:
a) hinge (hinge, revolute);

b) 2-hinge (universal);
c) 3-hinge (hinge plus universal);
d) hinge (Rodrigues parameters);
e) gimbal (2-Rodrigues);
f) rod end (3-Rodrigues, ball, spherical, pinned);
g) hinge (third variable) then gimbal;
h) gimbal then hinge (first variable);
described by one to three variables (with gain factors; the first variable is the last rotation). The axes of the angular
motions can be arbitrarily specified. These options are sufficient to define joints with combined motions, such
as a screw or rack-and-pinion. Here Rodrigues parameters are used as a convenient way to generate a joint in
which all rotations are treated the same. Note that for the Rodrigues joints, the spring/damper/actuator uses the
Rodrigues parameters rather than the rotation angles. The 3-hinge joint has a singularity at 90 deg motion of one
of the variables; often this singularity can be avoided by defining the last rotation as the one that has large motion.
The Rodrigues joint has a singularity at 180 deg total motion. A linear or angular displacement in the joint equals
a joint variable θ times a gain factor. Each linear and angular joint variable θ is one of the following:
a) Degree of freedom: a joint degree of freedom can be shared by more than one variable, of
one or more than one joint.
b) Prescribed: the joint motion can be constant velocity, rotating variable, or constant position.
c) Controlled: the joint motion can be obtained from a control vector, implementing a dis-
placement actuator. The control vector can be just the joint displacement (an approximation
if the motion is time varying), or include the joint velocity and acceleration.
Note that a force or offset actuator can be implemented using the spring/damper/actuator convention for a joint
degree of freedom. With a controlled joint, a structural dynamic no-residual constraint depends on the component
input. Hence a controlled joint can not be used in the flutter task. A joint degree of freedom with an offset actuator
can be used instead (and the degree of freedom can be quasistatic). A controlled joint can be used in the trim and
transient tasks, as long as the control variable is solved in a separate part. Each joint degree of freedom (involving
one or more joint variables) has an equation of motion obtained by equilibrium of loads on the joints:
fspring = finterfaces
where the reaction fspring is obtained from a spring/damper/actuator model, and finterfaces is obtained by summing
the effects of all interface loads acting on all the joints involved in this degree of freedom.
A structural dynamic torque interface is only allowed on joints where the last motion (first variable) is angular
rotation about an axis u: only for 1-hinge, 2-hinge, 3-hinge, and gimbal-then-hinge joint kinds. The torque
interface is intended for use with a joint that has a single rotational motion (1-hinge). A structural dynamic pinned
interface is inconsistent with a joint that has angular degrees of freedom. Usually a joint with angular degrees of
freedom must have at least one cantilever or torque interface.
3-3.4 Structural Dynamic Interface
Only a structural dynamic component can have structural dynamic input and output, for connection to another
component through a structural dynamic interface. The interface occurs at a connection (C). The interface kind
can be complete or torque.
3-3.4.1 Complete Interface
A complete interface is a true physical connection. It connects two components relative their common parent
frame (P). The component input consists of the force and moment acting on the connection: F x and M C . The
force can be in connection (F x = F C ) or common frame (F x = F P ) axes; the moment is always in connection
axes. All six elements of the force and moment always exist in the component input, regardless of the interface
kind. For example, with a pinned interface only the force is used, but M C = 0 still exists in the component. The
component output consists of the axes motion at the connection, relative the common frame axes: xCP/P and
C CP (and the derivatives, for a no-residual interface).
3-3.4.2 Torque Interface
A torque interface is an approximation, dealing with only rotational motion and torques. A torque interface
produces no net moment on a true structural dynamic component; otherwise equilibrium of moments on the
component would be incorrect, since this interface does not properly account for the direction or load path of the
torque. The torque interface can be viewed as a device attached to the structural dynamic component that produces
the required torque on a joint (as required for the interface solution), which is reacted by an equal and opposite
torque on the body of the component, so there is no net load to influence the rigid or elastic equation of motion. The
torque interface is only involved in the joint equations of the component, and can be used only on an appropriate
angular joint. The component input is the torque Q. The moment acting on the joint is then
M C = uQ
where u is unit vector along the rotational axis of the joint. The component output is rotation φ about the u-axis
(and the derivatives, for a no-residual interface).
3-3.5 Aerodynamic Interface
A standard input/output interface (component input and/or component output) is defined for interfaces with
aerodynamic components. The interface occurs at a connection on the structure (which can not be at a joint). The
component input can be:
a) force (F );
b) force and moment (F and M );
c) interference velocity (vA );
d) gust velocity (vG );
e) gust angular velocity (ωG );
in body, frame, or parent frame axes. The component output can be
a) velocity relative the air (v, q, v̇, ω);

b) position relative origin of axes (r and ṙ);
in connection, body, frame, or parent frame axes. The component output can be a subset of these quantities. The
velocity relative the air is calculated using all interference and gust velocities (input) at this connection; figure 4
illustrates the calculation. An aerodynamic component typically involves a set of collocation points. The task
of calculating the geometry of the collocation points belongs to the structural dynamic component, which knows
about the physical configuration. Calculating the velocity of a collocation point relative to the air also belongs
in the structural dynamic component, since the task requires the kinematics of the connection. Typically at a
connection that is a collocation point, there will be defined aerodynamic interfaces for the interference and gust
velocities (input), the velocity relative the air (output), the position (output), and the resulting aerodynamic load
(input).
3-3.6 Applied Load Interface
An applied force can act on the component. The interface occurs at a connection on the structure, the force
acting on the component at C. The axis u of the force in the C frame must be specified: u is a unit vector in the
x, y, or z-axis direction. The magnitude of the force is a gain a times the value f of some element of a control
vector (component input). Thus the applied force is
F C = uaf
Similarly
M C = uaf
for an applied moment acting on the structure.

AT A CONNECTION
interference
velocity vA
interference
velocity vA
gust
velocity v and wind
G
velocity of air velocity

relative inertial frame relative
the air
velocity of structure
relative inertial frame
body rigid and

elastic motion,
and frame motion
Figure 3-4 Aerodynamic interface to calculate velocity relative the air.

3-3.7 Controls
A set of control vectors is available to the component. These vectors are component input, for connection
to a system input piece or to an input/output interface. The standard features of a structural dynamic component
can use the controls for joints, actuators, or applied load interfaces. Component-specific uses are permitted. A
particular control vector and element can be used once, more than once, or not at all.
3-3.8 Sensors
A set of sensor vectors can be defined for the component. These vectors are component output, for connection
to a system output piece or to an input/output interface. The sensors can measure the following standard quantities:
a) motion (motion of axes at a point, velocity relative air, degrees of freedom, motion of torque interface);
b) structural dynamic interface constraint (force, moment, torque);
c) reaction (total load at a point, joint reaction, weight);
d) power (joint, joint degree of freedom, reaction).
There may also be component-specific sensors available. The value of any sensor can be multiplied by a scale
factor. Automatic scaling produces as appropriate a value in g’s, degrees, Mach number, or horsepower.
3-3.9 Spring/Damper/Actuator
A standard spring/damper/actuator model is used to provide a reaction force as required by the component.
For each degree of freedom θ (each element of the degree-of-freedom vector), a scalar reaction force f is generated:
fspring = f (θ, θ̇, c)
where c is an optional control (an element of a specified control vector) for an actuator model. The force is
evaluated as the sum of spring, damper, table, and bias terms. The spring is modelled by linear and elastomeric
terms. The damper is modelled by linear, elastomeric, and hydraulic terms. The table can implement more general
behavior than provided in the elastomeric and hydraulic models, including dependence on a switch parameter
(perhaps time). Optionally the device can support only compression, or only tension. Specifically, this convention
is used for joints, rod/cables, or transmission branches. A component can use the reaction force in only one context,
depending on the component type.
3–4 Theory
The nomenclature for the geometry and kinematics employs the following conventions. A vector x is a
column matrix of three elements, measuring the vector relative to a particular basis (or axes, or frame). A rotation
matrix C is a three-by-three matrix that transforms vectors from one basis to another. The basis is indicated as
follows:
a) xA is a vector measured in axes A.

b) xEF/A is a vector from point F to point E, measured in axes A.
c) C BA transforms vectors from basis A to basis B, so xB = C BA xA .
The matrix C BA defines the orientation of basis B relative basis A, so it also may be viewed as rotating the axes
is defined as follows:
from A to B. For a vector u, a cross-product matrix u
⎡ ⎤
0 −u3 u2
= ⎣ u3
u 0 −u1 ⎦
−u2 u1 0
v is equivalent to the vector cross-product u × v. Note that outside the context of rigid motion
such that u
kinematics, the tilde symbol does not imply this matrix. The cross-product matrix enters the relation between
angular velocity and the time derivative of a rotation matrix:
Ċ AB = − BA/B
ω AB/A C AB = C AB ω
(the Poisson equations). The rotation matrix Xα , Yα , or Zα indicates a rotation by an angle α about the x, y, or z
axis (1, 2, or 3 axis) respectively: ⎡ ⎤
1 0 0
Xα = ⎣ 0 cos α sin α ⎦
0 − sin α cos α
⎡ ⎤
cos α 0 − sin α
Yα = ⎣ 0 1 0 ⎦
sin α 0 cos α
⎡ ⎤
cos α sin α 0
Zα = ⎣ − sin α cos α 0 ⎦
0 0 1
Thus for example, C BA = Xφ Yθ Zψ means that the axes B are located relative the axes A by first rotating by angle
ψ about the z-axis, then by angle θ about the y-axis, and finally by angle φ about the x-axis. Euler-Rodrigues
parameters (or Rodrigues parameters) are defined by the relation p = 2u tan ψ/2, for rotation by angle ψ about
an axis in the direction of the unit vector u. This differs by a factor of 2 from the classical definition, but has the
advantage that p ∼= uψ for small rotations.
Consider the relative motion of axes A and B. Let xBA/A be the displacement from the origin of A to the
origin of B, measured in A axes; and C BA the rotation from axes A to axes B. Time derivatives of the motion of
B relative A can be described by body axis velocity and angular velocity. Thus the following standard convention
is used for a description of the rigid motion of axes B relative axes A:
a) linear motion: xBA/A and v BA/B = C BA ẋBA/A and v̇ BA/B

BA/B = C BA Ċ AB and ω̇ BA/B
b) angular motion: C BA and ω
The axes A and B might be frames, or rigid body degrees of freedom, or the position of some point on a structure.
Evaluating the geometry and kinematics requires the addition of axes motion. The equations
xCA/A = C AB xCB/B + xBA/A
C CA = C CB C BA
BA/B xCB/B )
v CA/C = v CB/C + C CB (v BA/B + ω
ω CA/C = ω CB/C + C CB ω BA/B
BA/B xCB/B )
v̇ CA/C = v̇ CB/C + C CB (v̇ BA/B + ω̇
+ (C CB
ω BA/B )v CB/C − ω CB/C C CB (v BA/B + ω
BA/B xCB/B )
ω̇ CA/C = ω̇ CB/C + C CB ω̇ BA/B − ω
CB/C C CB ω BA/B
implement the sum CA = CB + BA.
3-4.2 Joints
The joint kinematics describe the position of the joint axes relative to a location. JE is the sum of joint degrees
of freedom and a constant position (JE = JD + DE). The constant position of the joint (the position for zero joint
variables) relative the location axes is specified by z DE/E and C DE (D relative E, in E axes). Then the joint
motion occurs, described by xJD/D and C JD (and their derivatives). The linear and angular motion of the joint
are defined separately.
The linear motion is a combination of one to three displacements. Each displacement has value aθ, where a is
a gain and θ is a joint variable. The direction of the displacement is the vector u (or v or w), which is a unit vector
in the x, y, or z-axis direction of the D axes (a negative gain produces motion in the negative-axis direction). The
linear motion can have one of the following configurations:
a) 1-linear: x = ua1 θ1
b) 2-linear: x = ua1 θ1 + va2 θ2
c) 3-linear: x = ua1 θ1 + va2 θ2 + wa3 θ3
using one, two, or three joint variables. The velocity and acceleration for the 3-linear joint are thus:
x = ua1 θ1 + va2 θ2 + wa3 θ3

v = ua1 θ̇1 + va2 θ̇2 + wa3 θ̇3
v̇ = ua1 θ̈1 + va2 θ̈2 + wa3 θ̈3
and similarly for the 1-linear and 2-linear configurations. The partial velocity (required for the equation of motion)
is vθ = ua1 for a single displacement (and similarly for the other configurations).
The angular motion is a combination of one to three rotations. Each rotation has value aθ, where a is a gain
and θ is a joint variable. The direction of the rotation is the vector u (or v or w), which is a unit vector in the x, y,
or z-axis direction of the D axes (a negative gain produces motion about the negative-axis direction). The rotation
is performed by a rotation matrix Uaθ (or V or W ); or using Rodrigues parameters p = uaθ. The angular motion
can have one of the following configurations:
a) hinge: C = Ua1 θ1
b) 2-hinge: C = Ua1 θ1 Va2 θ2
c) 3-hinge: C = Ua1 θ1 Va2 θ2 Wa3 θ3
d) Rodrigues hinge: C = C(p), p = ua1 θ1
e) gimbal: C = C(p), p = ua1 θ1 + va2 θ2
f) rod end: C = C(p), p = ua1 θ1 + va2 θ2 + wa3 θ3
g) hinge then gimbal: C = C(p)Wa3 θ3 , p = ua1 θ1 + va2 θ2
h) gimbal then hinge: C = Ua1 θ1 C(p), p = va2 θ2 + wa3 θ3
using one, two, or three joint variables. For the hinge joints, the rotation matrix, angular velocity, angular
acceleration, and partial velocity (required for the equation of motion) are as follows.
C = Ua1 θ1 Va2 θ2 Wa3 θ3
ω = ua1 θ̇1 + U va2 θ̇2 + U V wa3 θ̇3

⎛ ⎞
a1 θ̇1
= [ u U v U V w ] ⎝ a2 θ̇2 ⎠
a3 θ̇3
= R(aθ̇)
ω̇ = R(aθ̈) + Ṙ(aθ̇)
= ua1 θ̈1 + U va2 θ̈2 + U V wa3 θ̈3 + (ua1 θ̇1 + U va2 θ̇2 )(ua1 θ̇1 − U V wa3 θ̇3 )
⎡ ⎤
a1 0 0
ωθ = R ⎣ 0 a2 0 ⎦
0 0 a3
= [ ua1 U va2 U V wa3 ]
C = Ua1 θ1 Va2 θ2
ω = ua1 θ̇1 + U va2 θ̇2
ω̇ = ua1 θ̈1 + U va2 θ̈2 + (U va2 θ̇2 )(ua1 θ̇1 )
ωθ = [ ua1 U va2 ]
C = Ua1 θ1
ω = ua1 θ̇1
ω̇ = ua1 θ̈1
ωθ = ua1
for three hinges, two hinges, and one hinge respectively. For the Rodrigues joints, the velocity and acceleration of
the Rodrigues parameters are:
p = ua1 θ1 + va2 θ2 + wa3 θ3
ṗ = ua1 θ̇1 + va2 θ̇2 + wa3 θ̇3
p̈ = ua1 θ̈1 + va2 θ̈2 + wa3 θ̈3
and similarly for the 1-Rodrigues and 2-Rodrigues configurations; C, ω, and ω̇ are obtained from p, ṗ, and p̈.
Since ω = Rṗ, it follows that the partial velocity is ωθ = Rua1 for a single variable (and similarly for the other
configurations). The last two angular joints are a combination of 2-Rodrigues and 1-hinge motions, in either order:
CA = CB + BA. Hence:
C CA = C CB C BA
ω CA/C = ω CB/C + C CB ω BA/B
ω̇ CA/C = ω̇ CB/C + C CB ω̇ BA/B − ω
CB/C C CB ω̇ BA/B
∂ω CA/C ∂ω CB/C ∂ω BA/B

= + C CB
∂ θ̇ ∂ θ̇ ∂ θ̇
In particular:
ω = Rg (ua1 θ̇1 + va2 θ̇2 ) + Cg wa3 θ̇3
ω = ua1 θ̇1 + Ch Rg (va2 θ̇2 + wa3 θ̇3 )

for the hinge-then-gimbal and gimbal-then-hinge configurations respectively. A structural dynamic torque interface
is only allowed on joints where the last motion (first variable) is angular rotation about an axis u: only for 1-hinge,
2-hinge, 3-hinge, and gimbal-then-hinge joint kinds. The component output for the interface is the rotation about
the u-axis:
φ = a1 θ1
(and the derivatives). The component input is the torque Q, which produces the moment acting on the joint,
M C = uQ.
A joint can consist of just linear motion; just angular motion; linear followed by angular motion (x in D axes):
xJD/D = x
v JD/J = Cv
v̇ JD/J = C v̇ − ω
Cv
C JD = C
ω JD/J = ω
ω̇ JD/J = ω̇
or angular followed by linear motion (x in J axes):
xJD/D = C T x
x
v JD/J = v + ω
+ω
v̇ JD/J = v̇ + ω̇x v
C JD = C
ω JD/J = ω
ω̇ JD/J = ω̇
These expressions follow from addition of the linear motion (x, v = ẋ, v̇) and angular motion (C, ω, ω̇) of the
joint. The corresponding partial velocities (required for the equation of motion) are
∂v JD/J
= Cvθ
∂ θ̇
∂ω JD/J
= ωθ
∂ θ̇
for linear followed by angular motion; and
∂v JD/J
= vθ − x
ωθ
∂ θ̇
∂ω JD/J
= ωθ
∂ θ̇
for angular followed by linear motion.
Optionally each linear and angular variable θ may be a degree of freedom, prescribed, or controlled. A joint
degree of freedom can be shared by more than one variable, of one or more than one joint; and gains relate the joint
degree of freedom to the magnitude of the linear displacement or angular rotation. Thus complex configurations
such as a screw or a transmission can be defined with this convention. For a prescribed variable, the joint motion
can be specified as constant velocity:
θ = θ̇0 t + θ0
θ̇ = θ̇0
or rotating variable:
θ = Ωt + θ0
θ̇ = Ω
or constant position:
θ = θ0
Here the velocity θ̇0 and offset θ0 are constant parameters; and the rotational speed Ω is obtained from a specified
period. For a controlled variable, the joint motion is obtained from a control vector (component input):
⎛ ⎞ ⎛ ⎞
θ c + θ0
⎝ θ̇ ⎠ = ⎝ ċ ⎠
θ̈ c̈
including a constant offset θ0 . This implements a displacement actuator. The control vector must consist of the
displacement, velocity, and acceleration (length 3), in order to fully evaluate the joint motion. If the control vector
provides only the displacement, then ⎛ ⎞ ⎛ ⎞
θ c + θ0
⎝ θ̇ ⎠ = ⎝ 0 ⎠
θ̈ 0
is used, which is an approximation if c is time-varying. A system input piece does not provide the derivatives of
the input variables. A component can be constructed to provide the derivatives for a time-varying control. Note
that a force or offset actuator can be implemented using the spring/damper/actuator convention for a joint degree
of freedom. The offset actuator with a large spring stiffness effectively provides a displacement actuator, and does
not require the derivatives of the control.
Each joint degree of freedom θ (involving one or more joint variables) has an equation of motion obtained by
equilibrium of loads on the joints:
fspring = vθT F J + ωθT M J
where the reaction fspring is obtained from a spring/damper/actuator model. The partial velocities
∂v JE/J ∂v JD/J
vθ = =
∂ θ̇ ∂ θ̇
∂ω JE/J ∂ω JD/J
ωθ = =
∂ θ̇ ∂ θ̇
are evaluated as described above (depending on the joint kind). The reactions for the equations of motion, F J
and M J , are evaluated from all interface loads acting on all connections, of all joints involved in this degree of
freedom (as described below). The interpretation of the degree of freedom θ depends on the definition of the joint.
The spring and damper parameters used in fspring must be consistent with that interpretation. In particular, for
large rotations a Rodrigues parameter is not equal to the rotation angle.
3-4.3 Aerodynamic Interface
The component input for an aerodynamic interface can be one of the following quantities, acting at a connection
(C):
a) force (F R );
b) force and moment (F R and M R );
R
c) interference velocity (vA );
R
d) gust velocity (vG );
R
e) gust angular velocity (ωG ).
The axes R of the quantity must be identified: body, frame, or parent frame axes (R=BFP). A frame that is not
a parent of the component frame can not be handled (because of the frame functionality). The interference and
gust velocities are needed to calculate the total velocity of the air at the connection. The component output for an
aerodynamic interface can be one of the following quantities, at a connection (C):
a) velocity relative air (v R , q, v̇ R , ω R );

b) position relative origin of axes (rR and ṙR ).
A subset of these quantities can be selected. The axes R of the quantity must be identified: connection, body,
frame, or parent frame axes (R=CBFP). A frame that is not a parent of the component frame can not be handled
(because of the frame functionality). The velocity relative to the air is calculated using all the interference and
gust velocities (component input) defined at this connection.
The position of the connection C, relative the origin of the axes R, is calculated as follows:
rR = xCR/R
ṙR = C RC v CR/C
from the motion CR. The velocity of the connection C relative to the air, in axes R, is calculated as follows:
v R = vB
R
− vW
R
− R
vG − R
vA
v̇ R = v̇B
R
− v̇W
R
ω R = ωB
R
− R
ωG
q = 1/2ρ|v R |2
where the first term in each equation is the motion of the connection relative to the inertial frame, and the remaining
terms are the motion of the air relative to the inertial frame (wind, gust, and aerodynamic interference). Note that
time derivatives of the interference and gust velocities are not considered. The required terms are calculated as
follows. The motion of the connection C relative to the inertial frame I, in R axes is:
R
vB = C RC v CI/C
R
v̇B CR/C v CI/C )
= C RC (v̇ CI/C + ω
R
ωB = C RC ω CI/C
from the motion CI and CR. The wind at connection C relative to the inertial frame I, in R axes is:
R
vW = C RI vW
I
R
v̇W = −
ω RI/R C RI vW
I
I I
from the motion RI, and the wind velocity vW . With a ground boundary layer in the wind model, the velocity vW
is obtained at the position xCI/I (and the resulting time variation of vW I R
is neglected in calculating v̇W ). A gust
at connection C (perturbation of the air velocity relative to the inertial frame) in R axes is:
R Q
vG = C RQ vG
R Q
ωG = C RQ ωG
from the motion RQ, and the gust velocity in Q axes (component input). An aerodynamic interference at connection
C (perturbation of the air velocity relative to the inertial frame) in R axes is:
R Q
vA = C RQ vA
from the motion RQ, and the interference velocity in Q axes (component input).
3-4.4 Sensors
The following standard sensor quantities can be defined as component output: motion of axes at a point;
velocity relative air; degrees of freedom; motion of torque interface; structural dynamic constraint; reaction;
power.
3-4.4.1 Motion of Axes at a Point
The sensor quantity is the axes motion at Q, relative R, in Q axes (except in R axes for displacement). The
point Q can be a connection, a joint, a location, or the body origin (Q=CJEB). The point R can be the origin of
the body, frame, or parent frame axes (R=BFP). A frame that is not a parent of the component frame can not be
handled (because of the frame functionality). The following motion can be obtained:
a) Displacement: xQR/R
b) Velocity: C QR ẋQR/R = v QR/Q
c) Acceleration: QR/Q v QR/Q
C QR ẍQR/R = v̇ QR/Q + ω
d) Load factor: QR/Q v QR/Q − C QI g I
n = v̇ QR/Q + ω
e) Angular velocity: ω QR/Q
e) Angular acceleration: ω̇ QR/Q
f) Rodrigues displacement: p from C QR
g) Rodrigues velocity: ṗ from ω QR/Q
h) Rodrigues acceleration: p̈ from ω̇ QR/Q
i) Rotational displacement: uψ from C QR
j) Euler angles: φ, θ, ψ from C QR
The Rodrigues parameters are defined by the relation p = 2u tan ψ/2, for rotation by angle ψ about an axis in the
direction of the unit vector u. Hence u = p/|p| and ψ = 2 tan−1 |p|/2 give the rotational displacement sensor:
p |p| tan−1 |p|/2

uψ = 2 tan−1 =p
|p| 2 |p|/2
The Euler angles are obtained from one of the following definitions:
⎡ ⎤
cos θ cos ψ cos θ sin ψ − sin θ
C QR = Xφ Yθ Zψ = ⎣ − cos φ sin ψ + sin φ sin θ cos ψ cos φ cos ψ + sin φ sin θ sin ψ sin φ cos θ ⎦
sin φ sin ψ + cos φ sin θ cos ψ − sin φ cos ψ + cos φ sin θ sin ψ cos φ cos θ
⎡ ⎤
cos ψ cos θ sin ψ − cos ψ sin θ
C QR = Xφ Zψ Yθ = ⎣ − cos φ sin ψ cos θ + sin φ sin θ cos φ cos ψ cos φ sin ψ sin θ + sin φ cos θ ⎦
sin φ sin ψ cos θ + cos φ sin θ − sin φ cos ψ − sin φ sin ψ sin θ + cos φ cos θ
⎡ ⎤
cos θ cos ψ − sin θ sin φ sin ψ cos θ sin ψ + sin θ sin φ cos ψ − sin θ cos φ
C QR = Yθ Xφ Zψ = ⎣ − cos φ sin ψ cos φ cos ψ sin φ ⎦
sin θ cos ψ + cos θ sin φ sin ψ sin θ sin ψ − cos θ sin φ cos ψ cos θ cos φ
⎡ ⎤
cos θ cos ψ cos θ sin ψ cos φ + sin θ sin φ cos θ sin ψ sin φ − sin θ cos φ
C QR = Yθ Zψ Xφ = ⎣ − sin ψ cos ψ cos φ cos ψ sin φ ⎦
sin θ cos ψ sin θ sin ψ cos φ − cos θ sin φ sin θ sin ψ sin φ + cos θ cos φ
⎡ ⎤
cos ψ cos θ + sin ψ sin φ sin θ sin ψ cos φ − cos ψ sin θ + sin ψ sin φ cos θ
C QR = Zψ Xφ Yθ = ⎣ − sin ψ cos θ + cos ψ sin φ sin θ cos ψ cos φ sin ψ sin θ + cos ψ sin φ cos θ ⎦
cos φ sin θ − sin φ cos φ cos θ
⎡ ⎤
cos ψ cos θ cos ψ sin θ sin φ + sin ψ cos φ − cos ψ sin θ cos φ + sin ψ sin φ
C QR = Zψ Yθ Xφ = ⎣ − sin ψ cos θ − sin ψ sin θ sin φ + cos ψ cos φ sin ψ sin θ cos φ + cos ψ sin φ ⎦
sin θ − cos θ sin φ cos θ cos φ
Hence the angles are:
C QR φ θ ψ
−1
Xφ Yθ Zψ tan−1 (C23 /C33 ) sin (−C13 ) tan−1 (C12 /C11 )
Xφ Zψ Yθ −1
tan (−C32 /C22 ) −1
tan (−C13 /C11 ) sin−1 (C12 )
Yθ Xφ Zψ sin−1 (C23 ) tan−1 (−C13 /C33 ) tan−1 (−C21 /C22 )
Yθ Zψ Xφ tan−1 (C23 /C22 ) tan−1 (C31 /C11 ) sin−1 (−C21 )
−1 −1
Zψ Xφ Yθ sin (−C32 ) tan (C31 /C33 ) tan−1 (C12 /C22 )
Zψ Yθ Xφ tan−1 (−C32 /C33 ) sin−1 (C31 ) tan−1 (−C21 /C11 )
The range of the middle rotation (sin−1 ) is −90 to 90 deg; the range of the other variables (tan−1 ) is −180 to 180
deg. The singularity at the middle variable equal ±90 deg (|C| = 1, cos = 0) is resolved by assuming that the
first variable is zero, and then
C QR φ θ ψ
−C − C13 C32
sin−1 (−C13 ) = ±90
21
Xφ Yθ Zψ 0 tan−1
C + C C C22 − C13 C31
sin−1 (C12 ) = ±90
31 12 23
Xφ Zψ Yθ 0 tan−1
C33 − C12 C21 C + C C
sin−1 (C23 ) = ±90
12 23 31
Yθ Xφ Zψ 0 tan−1
−C − C C C11 − C23 C32
sin−1 (−C21 ) = ±90
32 21 13
Yθ Zψ Xφ tan−1 0
C33 − C21 C12 −C − C32 C21
sin−1 (−C32 ) = ±90
13
Zψ Xφ Yθ tan−1 0
C + C C C11 − C32 C23
sin−1 (C31 ) = ±90
23 31 12
Zψ Yθ Xφ tan−1 0
C22 − C31 C13
In order to get the motion at a structural dynamic interface, identify Q = C (the connection of the interface). In
order to get the rigid body degrees of freedom of the component, identify Q = B. Optionally the motion can be
obtained in output axes O, identified as either the component frame or a parent frame, by multiplying the vector
by C OQ (C OR for the displacement).
3-4.4.2 Velocity Relative Air
The sensor quantity is the velocity relative the air at R, in Q axes. The point R is a connection (R=C). The axes
Q can be connection, body, frame, or parent frame axes (Q=CBFP). A frame that is not a parent of the component
frame can not be handled (because of the frame functionality). The following motion can be obtained: air velocity,
air angles. The air velocity is the velocity of the connection relative to the air, in Q axes:
Q Q Q Q
v Q = vB − vW − vG − vA
calculated as for the aerodynamic interface. The air angles are the longitudinal velocity fraction, sideslip angle,
and angle of attack: ⎛ ⎞ ⎛ ⎞
u/V v1 /|v|
⎝ β ⎠ = ⎝ sin v2 /|v| ⎠
−1
α tan−1 v3 /v1
calculated from the components of the air velocity v Q .
3-4.4.3 Aerodynamic Interface
The sensor quantity consists of the terms in the velocity relative to the air for an aerodynamic interface R,
in Q axes. The aerodynamic interface R identifies a connection. The axes Q can be connection, body, frame, or
parent frame axes (Q=CBFP). The sensor vector contains the following terms: velocity (v, and terms vB , vW , all
vG , all vA ), acceleration (v̇, and terms v̇B , v̇W ), angular velocity (ω, and terms ωB , sum of ωG ); calculated as for
the aerodynamic interface.
3-4.4.4 Degrees of Freedom
The sensor quantity consists of the degrees of freedom of the vector Q. The degrees of freedom Q can be
joint, elastic, or body (Q=JEB). The body degrees of freedom are not available if they are rigid body motion. The
displacement, velocity, or acceleration can be obtained.
3-4.4.5 Motion of Torque Interface
The sensor quantity is the rotation angle of a structural dynamic interface Q (torque kind only). The displace-
ment, velocity or acceleration can be obtained.
3-4.4.6 Structural Dynamic Interface Constraint
The sensor quantity is the constraint force or moment of a structural dynamic interface Q (complete kind
only). The quantity can be obtained in connection axes (C), or in common parent frame axes (P). It is also possible
to define a system output piece for the interface variable fl , which for a structural dynamic interface consists of
the constraint forces.
3-4.4.7 Structural Dynamic Interface Torque
The sensor quantity is the constraint torque of a structural dynamic interface Q (torque kind only). It is also
possible to define a system output piece for the interface variable fl = Q.
3-4.4.8 Reaction
The sensor quantity is the total load (force or moment) acting on Q, about a reference R, in R axes. Q can be a
connection, a joint, a location, or the body origin (Q=CJEB). The reference R is a connection (R=C), or the origin
of Q, or the origin of a frame. Loads act on the component at connections. All the loads on all the connections of
Q=CJEB are summed, in Q axes, about the origin of Q:
FQ = C QS F S
MQ = CQ/Q C QS F S
C QS M S + x
where S is the axes of the load at a connection (S=CBFP). If the reference R is the origin of Q, this is the result.
If the reference R is a connection or frame, then
F R = C RQ F Q
M R = C RQ M Q + Δ
xQR/R C RQ F Q
transfers the reaction to R axes, about the point R. If the reference R is a connection, then
ΔxQR/R = C RB (xQB/B − xRB/B )
since B is always a common parent of R=C and Q=CJEB. Note that the connection R might be at a joint, a location,
or the origin of the body axes. If the reference R is a frame (R=FP), then ΔxQR/R is the position of Q relative the
origin of frame R. Optionally the loads can be obtained in output axes O, identified as either the component frame
or a parent frame, by multiplying the vector by C OQ or C OR .
3-4.4.9 Joint Degree of Freedom Reaction
The sensor quantity is total force acting on the equation for joint degree of freedom Q. Since the joint equation
of motion is fspring = finterfaces , the total force can be evaluated from the joint motion, using the spring/damper/
actuator reaction fspring .
3-4.4.10 Weight and Load Factor
The sensor quantity is the weight, acceleration force, or inertial force. The weight is the total gravity force
B
acting on the component: Fgrav = M C BI g I , in body axes. The acceleration force is
B
Facc = −M (v̇ BI/B + ω
BI/B v BI/B )
B
in body axes. The inertial force is the sum (Fgrav − Facc
B
), which is the weight times the load factor. Here M is
the total mass of the component. Optionally the loads can be obtained in output axes O, identified as either the
component frame or a parent frame, by multiplying the vector by C OB .
3-4.4.11 Joint Power
The sensor quantity is power of a joint Q=J. The total load (reaction force F J and moment M J ) on the joint
is calculated, and then
P = − (v JE/J )T F J + (ω JE/J )T M J
is the definition of the power (rate of work performed by the forces and moments). Note that this definition typically
gives
d 1
P = θ̇f = θ̇(Kθ + C θ̇) = ( Kθ2 ) + C θ̇2
dt 2
so P includes the rate of change of potential energy (which may average to zero) as well as energy losses.
3-4.4.12 Joint Degree of Freedom Power
The sensor quantity is power of a joint degree of freedom Q. The definition of the power (rate of work
performed by the forces and moments) is P = finterfaces θ̇ = fspring θ̇, where θ is the joint degree of freedom. The
last expression is used, since the spring/damper/actuator reaction fspring can be evaluated from the joint motion.
3-4.4.13 Reaction Power
The sensor quantity is power of the total load (force or moment) acting on Q, about a reference R. Q can be
a connection, a joint, a location, or the body origin (Q=CJEB). The reference R is a parent of Q (R=JEBFP). All
the loads on all the connections of Q=CJEB are summed, in Q axes, about the origin of Q (reaction force F Q and
moment M Q ). Then
P = − (v QR/Q )T F Q + (ω QR/Q )T M Q
is the definition of the power (rate of work performed by the forces and moments). The result is the same as the
joint power if Q=J and R=E.
3-4.5 Reaction for Equations of Motion
The equations of motion depend on the component type, but in all cases require that a total load be evaluated.
The reaction (force or moment) acting on Q, about the origin of Q, in Q axes is required. Q can be a joint, a
location, or the body origin (Q=JEB), for the joint, elastic, or rigid body equations respectively. Loads act on
the component at connections. A particular load at a particular connection can be in connection, body, frame, or
parent frame axes. All the loads on all the connections of Q=JEB are summed, in Q axes, about the origin of Q:
FQ = C QS F S
MQ = CQ/Q C QS F S
C QS M S + x
where S is the axes the load at a connection (S=CBFP). The summation is over all connections of Q.
3-4.6 Spring/Damper/Actuator
A standard spring/damper/actuator model is used to provide a reaction force as required by the component.
For each degree of freedom θ (an element of the degree-of-freedom vector), a scalar reaction force f is generated:

where c is an optional control (an element of a specified control vector). Note that the power of the device can be
calculated from P = fspring θ̇. The force is evaluated as the sum of spring, damper, table, and bias terms:
fspring = K(θ − θ0 ) + C(θ̇, θ − θ0 ) + T (θ̇, θ − θ0 , s) − c0
with offset θ0 and bias c0 . The offset θ0 is usually the value of θ for which the spring gives zero load. The definition
of the degree of freedom θ can include an offset as well. The bias c0 gives the load for zero displacement. The
parameter s in the table term is an optional switch (an element of a specified control vector, or time). Optionally
the device can support only compression, or only tension. An actuator is added to this model by introducing the
control c. For a force actuator, c0 is replaced by (c0 + c):
fspring = K(θ − θ0 ) + C(θ̇, θ − θ0 ) + T (θ̇, θ − θ0 , s) − (c0 + c)
For an offset actuator, θ0 is replaced by (θ0 + c):
fspring = K(θ − θ0 − c) + C(θ̇, θ − θ0 − c) + T (θ̇, θ − θ0 − c, s) − c0
For very large stiffness, the offset actuator should thus produce θ = θ0 + c, which is a displacement actuator. Note
that a joint has a separate implementation for a displacement actuator. To compare the role of offset and bias, and
of offset actuator and force actuator, consider the case of linear spring and linear damper:
fspring = Kθ + C θ̇ − (Kθ0 + c0 ) for no actuator

fspring = Kθ + C θ̇ − (Kθ0 + c0 + Kc) for an offset actuator
fspring = Kθ + C θ̇ − (Kθ0 + c0 + c) for a force actuator
Figure 5 summarizes the actuator models.
The spring is modelled by linear and elastomeric terms. The damper is modelled by linear, elastomeric, and
hydraulic terms. Thus the contributions to the reaction force are:
K(θ) = Klin θ + Kelast

C(θ̇, θ) = Clin θ̇ + Celast + Chyd
(used with an offset θ0 ). The linear terms are defined by constant spring and damping coefficients, Klin and Clin .
The elastomeric spring/damper is described by polynomials in the displacement θ and rate θ̇:

Kelast = sign(θ) Pekd (|θ|) + Pekr (|θ̇|)

Celast = sign(θ̇) Pecd (|θ|) + Pecr (|θ̇|)
where each polynomial is implemented as fifth-order (six terms). The Pekd term is a nonlinear spring; the Pecr
term is nonlinear viscous damping. The Pecd term is friction damping. If a harmonic motion θ = δ cos ωt
produces a reaction f = K δ cos ωt − K δ sin ωt, and K is independent of frequency, then friction damping is
the appropriate time domain model. The hydraulic damper is described by polynomials in the rate θ̇:

Chyd = sign(θ̇) min Pha (|θ̇|), Phb (|θ̇|)
DISPLACEMENT ACTUATOR (controlled joint)
θ = c (need c, c also)
OFFSET ACTUATOR
c R
FORCE ACTUATOR
F=c
R
θ = variable to be controlled
c = control (component input)
F = force
R = reaction
= spring (no load length θ 0 )
Figure 3-5 Summary of actuator models.

where each polynomial is implemented as fifth-order (six terms). A hydraulic damper would typically be of the
form
Pha = ca θ̇2 = fmax (θ̇/θ̇max )2
Phb = cb = fmax
The hydraulic model can also represent more general behavior of a viscous damper.
The term T (θ̇, θ, s) is obtained from a two-dimensional table of the force T as a function of displacement
and rate; or from a three-dimensional table of the force T as a function of displacement and rate and a switch s
(an element of a specified control vector, or time). A spring is modelled if the table data are only a function of
θ; a viscous damper is modelled if the table data are only a function of θ̇. The table can implement more general
behavior than provided by the elastomeric and hydraulic polynomials.
The behavior a nonlinear spring/damper can be described in terms of an equivalent linear spring and linear
damper, the magnitude of which will depend on the amplitude and frequency of the motion. Such behavior depends
on a finite amplitude of the motion, hence can not be obtained by a local linearization of the force. Thus when the
analysis perturbs the spring/damper/actuator, it can optionally use
K(θ) = Kequiv θ
C(θ̇, θ) = Cequiv θ̇
T (θ̇, θ) = 0
for the trim, transient, or flutter task.

β, ξ SD torque aero control
motion (a) yes (e) yes (g) yes
SD 0 (b) no (f) no (i) (f)
torque no (c) no no (i)
aero 0–2 (d) (j) no no (h) no
sensor:
motion or spring 2 (j) no no yes
reaction 2 (k) yes no yes
motion air 2 no yes yes
degree of freedom 2 (l) no no no
modal motion no no no no
beam reaction 2 yes yes yes
“SD” means a complete structural dynamic interface; “torque” means a torque struc-
tural dynamic interface; “aero” means an aerodynamic interface.
sensor = motion or spring: QUANT = 1–11, 24–26, 43, 52
rod/cable component, QUANT = 67–68
sensor = reaction: QUANT = 31–35, 41–42, 51, 53
sensor = motion air: QUANT = 12–14
sensor = degree of freedom: QUANT = 21–23
finite element beam component, QUANT = 61–62
rod/cable component, QUANT = 61–66
sensor = modal motion: linear normal modes component, QUANT = 61
sensor = beam reaction: finite element beam component, QUANT = 63–64
Notes:
a) see next tables
b) no if joint degree of freedom and interface not at joint
c) yes if joint degree of freedom; no frame dependence
d) depending on interface kind
e) no if joint equation and interface not at joint
f) yes if full residual
g) no if joint equation, or vA or vG
h) yes if vA or vG for v, and at same connection
i) yes if at controlled joint, or joint degree of freedom with actuator
j) frame dependence no for QUANT = 1–11 (maybe), 24–26, 43, 52
k) frame dependence no for QUANT = 35, 51
l) frame dependence no for QUANT = 21–23, rod/cable 61–66
Figure 3-6a Functionality of structural dynamic components.

rigid body component
frame β rigid ξR joint ξJ elastic ξE
rigid motion 2 2 0 X
joint motion 0 0 1 X
linear normal modes component
rigid motion 2 2 0 0
joint motion 0 0 1 0
elastic motion 0 0 0 2
finite element beam component
rigid motion 2 2 0 2
joint motion 0 0 1 0
elastic motion 2 2 0 2
rod/cable component
rigid motion 2 2 X 0
elastic motion 1 1 X 1
Figure 3-6b Functionality of structural dynamic components.

Chapter 4
RIGID BODY COMPONENT
4–1 Description
A rigid body component is a structural dynamic component. It has no elastic motion, hence consists of just
the standard features.
A point mass is a compact rigid body. It can be modelled using the rigid body component with no angular
motion, no moments, and only pinned structural dynamic interfaces. The center of gravity and all locations and
connections will be at the origin of the body axes.
A rod or link can be modelled as a rigid body component with two pinned structural dynamic interfaces,
separated by a distance equal to the length of the rod. A linear joint at one end can be used to model axial flexibility
of the rod.
4–2 Theory
4-2.1 Rigid Body Equations of Motion
The equations for the rigid body motion are obtained from equilibrium of forces and moments on the com-
ponent, in the body axes:
maB B B
CM = F + Fgrav
BI/B I B ω BI/B + m
I B ω̇ BI/B + ω B
zCM aB B
CM
0 =M +z
B B
Fgrav
F B and M B are the total force and moment acting on the body, about the origin of B, in B axes. The gravitational
force
B
Fgrav = mC BI g I
is calculated from the acceleration produced by gravity (constant in inertial axes). The body mass is m, and the
symmetric moment of inertia matrix is
⎡ ⎤
Ixx −Ixy −Ixz
IB = ⎣ −Iyx Iyy −Iyz ⎦
−Izx −Izy Izz
B
in B axes (hence constant). The quantity zCM is the position of the body center-of-mass, in B axes (hence constant).
The acceleration is:
d2 BI/I
aB
CM = C
BI
(x + C IB zCM
B
)
dt2
40 RIGID BODY COMPONENT
So the acceleration of the origin of the B axes and of the center-of-mass:
aB
0 = v̇
BI/B
BI/B v BI/B
+ω
aB
CM = v̇
BI/B
+ω BI/B zCM
BI/B v BI/B + ω̇ B
BI/B ω
+ω BI/B zCM
B
are obtained from the rigid motion BI. These equations of motion are written in terms of the velocities v BI/B and
ω BI/B , which are then represented in terms of linear degrees of freedom q̇ and angular degrees of freedom ṗ. In
order to maintain symmetry in terms of the degrees of freedom, the equations must be multiplied by the transpose
of the appropriate partial velocity matrices:
∂v BI/B
vq =
∂ q̇
for the linear equations, and
∂ω BI/B
ωp =
∂ ṗ
for the angular equations. These matrices are identified in the description of the conventions for representing rigid
motion, for the system response piece.
When modelling a rod, care must be taken with the rotation about the rod axis, which may be undetermined.
Consider a rigid body with two pinned interfaces, one at the origin and one at a position relative the origin (so
the length of the rod is | |). If the body is massless, equilibrium of force and moment give simply
0 = F1 + F 2
0 = F2
where F1 and F2 are the interface forces. However, the three moment equations are not independent (the solution
always must give zero axial moment). Problems can be avoided by using one of the following approaches.
a) Include a non-zero axial moment of inertia; the axial motion is still free however, which can lead to difficulties
with the trim and flutter solutions.
b) Align an axis of the B coordinates with the rod axis, and use an Euler angle representation of the angular motion,
with rotation about the rod axis as the last rotation; and then omit this axial rotation degree of freedom from the
system solution.
c) Replace one of the pinned interfaces with a cantilever interface, with a two-variable gimbal on one side of the
interface; so the axial rotation of the rod is constrained.
Chapter 5
LINEAR NORMAL MODES COMPONENT
5–1 Description
A linear normal modes component is a structural dynamic component. It has the standard features, and the
elastic motion is described by free-vibration normal modes. Typically these modes are obtained from a large
finite-element analysis. If the component is constrained, then constrained modes can be used. For this component,
the rigid motion describes the center-of-mass, mean-axes motion of the body. Thus the rigid body degrees of
freedom are orthogonal to the free-vibration modes. The elastic motion is small (consistent with using linear
modes), measured relative to the rigid motion. Aerodynamic spring, damping, and control terms are included in
the model, as an approximation to the effects of aerodynamics on the elastic motion.
5-2.1 Linear Normal Modes
The modes are calculated for a free body, including the rigid motion. Small motion is assumed, consistent
with the linear modes. Consequently there is no effect of the frame motion on the elastic modes, and no effect of
the elastic motion on the inertial properties of the rigid modes. Orthogonality of the rigid and elastic modes means
that the rigid modes represent motion of the center-of-mass, mean-axes of the body.
A linear structural dynamic model can be formulated using constraint modes rather than free-vibration modes.
For example, consider normal modes calculated with the node at the origin of the body axes constrained. These are
fixed interface modes, with the constraint modes equal to the rigid body motion. The normal modes have diagonal
(uncoupled) mass and spring matrices, and the spring matrix is fully decoupled since the constraint modes are here
just rigid motion (only the normal modes have nonzero spring). However the normal modes and rigid motions are
still coupled in the mass matrix. To avoid input data for this inertial coupling, the option of using fixed interface
modes is not implemented for this component. Note that this inertial coupling would also require that inertial
relief be considered in the quasistatic reduction of the component modes. A constrained component does not have
rigid body motion however. So fixed interface modes can be used for a constrained component, since the resulting
equations are identical to those obtained using free-vibration modes and omitting the rigid degrees of freedom.
The model is extended here by replacing the rigid modes by the large rigid body motion of the body axes B.
The origin of the body axes must be at the center-of-mass, but it is not necessary to use principal axes. The elastic
modes then represent motion relative to the body axes. The elastic and rigid equations are fully decoupled still
(from orthogonality). Let q be the vector of the generalized coordinates of the elastic modes. The linear equations
42 LINEAR NORMAL MODES COMPONENT
of motion take the form:

M q̈ + Kq = QA + ξeT F + γeT M
where the mass and spring matrices are diagonal (normal modes); and ξ and γ are the linear and angular mode
shapes at the connection where the force and moment are applied. The elastic motion at an arbitrary location on
the structure can be obtained from
Δu = ξd q
Δφ = γd q
where ξ and γ are the linear and angular mode shapes at the location.
5-2.2 Aerodynamics
The generalized force QA is produced by aerodynamic loads acting on the elastic component. The following
form is used here:
QA = 1/2ρV 2 [Fqδ δ − Fqq̇ (q̇/V ) − Fqq q]
As aerodynamic loads, they scale with the air density ρ and an appropriate component velocity through the air V .
The coefficients F are assumed to be constant. The vector δ consists of the modal controls, each element identified
as an element of a control vector. This is a quasistatic model, since no additional states are introduced. Further
approximations are made, as follows. Only the effects of aerodynamics on the elastic modes are considered here.
It is assumed that the aerodynamic forces involving the rigid body motions are handled by a separate component.
Only the diagonal terms in the aerodynamic spring and damping matrices are used. Using a full matrix would
be straightforward, but not really consistent unless the aerodynamic coupling of the elastic and rigid modes was
also included. This model of the aerodynamics is quite simple, serving only to represent the basic effects such as
aerodynamic damping of modes, and the effect of aerodynamic control surfaces on the modal deflection.
An aerodynamic component that calculates the forces and moments for the rigid body motion of the component
may not be able to handle the modal elastic motion. For example, a small number of aerodynamic collocation
points are sufficient to describe the rigid body aerodynamics, but a large number are required to define the loads
acting on an elastic mode. For this circumstance, the analysis has the option of suppressing the effects of the elastic
motion or force at a location (a standard feature for all structural dynamic components). Thus the elastic motion
can be suppressed in evaluating the collocation point motion for the aerodynamic component; and the resulting
aerodynamic forces can be omitted from the elastic equations of motion.
5-2.3 Modal Sensors
Quantities in the elastic structure such as load, stress, and strain can be expressed as a linear combination of
the modal deflections:
x = (∂z/∂q)T q
using the appropriate modal reaction coefficients (∂z/∂q).
5–3 Theory
The rigid body equations of motion are the same as for the rigid body component, with the restriction that
LINEAR NORMAL MODES COMPONENT 43
B
the origin of the body axes is at the center-of-mass: zCM = 0. The elastic motion is included however when
evaluating the point of application of forces and moments on the body.
5-3.2 Locations and Elastic Motion
The elastic motion affects the position of a location relative to the body origin (EB). EB is the sum of elastic
degrees of freedom and a constant position (EB = EL + LB). The constant position of the location (the position
for zero elastic motion) relative the body axes is specified by z LB/B and C LB (L relative B, in B axes). Then
the elastic motion occurs, described by xEL/L and C EL (and their derivatives). The elastic motion is described
in terms of the generalized coordinates θ and the mode shapes. The mode shapes define the linear and angular
motion at a location E. They can be input in either body axes B, or undeflected location axes L; and used in either
B or L axes, transforming with the constant rotation matrix C LB as required. Thus
xEL/L = ξdL θ
v EL/E = C EL ξdL θ̇ ∼
= ξdL θ̇
∼ ξ L θ̈
v̇ EL/E = d
= I − (γ
C EL ∼ L
d θ)
ω EL/E = C EL γ L θ̇ ∼
= γ L θ̇
d d
ω̇ EL/E ∼
= γdL θ̈
from the linear modes. It is necessary to retain exact kinematics for all interfaces with other components. Hence
it is not acceptable to use an approximate rotation matrix as above. Consistent rigid motion can be obtained by
taking the small rotation at the location to be Rodrigues parameters, and retaining all terms. Thus:
xEL/L = ξdL θ
v EL/E = C EL ξdL θ̇
v̇ EL/E = C EL ξdL θ̈ − ω
EL/E C EL ξdL θ̇
p = γdL θ
C EL = I − R
p
ω EL/E = Rṗ
ω̇ EL/E = Rp̈ + Ṙṗ
is used here (symmetry considerations imply similar changes to the elastic equations of motion). The model is
still only valid for small elastic motion. The purpose of the above expressions is to provide exact kinematics of
the axes motion, which is conveniently accomplished using Rodrigues parameters.
5-3.3 Elastic Equations of Motion
The elastic equations of motion are decoupled from each other (normal modes). The modal properties consist
of the generalized mass Mk ; frequency ωk ; structural damping g (twice the critical damping ratio), implemented
as viscous damping; aerodynamic coefficients; and the mode shapes at locations on the component. The equation
of motion for the k-th mode, with degree of freedom θk (generalized coordinate), is:

Mk θ̈k + Ck θ̇k + Kk θk = Dk δ + vθT F E + ωθT M E
44 LINEAR NORMAL MODES COMPONENT
where the modal damping, spring, and control terms are:
Ck = Mk gk ωk + CAk = Mk gk ωk + 1/2ρV Fqq̇

Kk = Mk ωk2 + KAk = Mk ωk2 + 1/2ρV 2 Fqq
Dk = DAk = 1/2ρV 2 Fqδ
including the aerodynamic loads (subscript A). The reactions F E and M E are evaluated from all interface loads
acting on all connections of a location (calculated as described for standard structural dynamic components). The
summation is over all locations of the component. The partial velocities
∂v EB/E ∂v EL/E
vθ = = = C EL ξdk
L
∂ θ̇k ∂ θ̇k
∂ω EB/E ∂ω EL/E L
ωθ = = = Rγdk
∂ θ̇k ∂ θ̇k
are calculated from the mode shapes at the location. The matrices C EL and R are nearly unity, but are required
to maintain symmetry of the structural dynamic equations. For reference in preparing the input parameters, the
modal equation of motion has the following form:

M q̈ + gω q̇ + ω 2 q = ξ T F + γ T M + 1/2ρV 2 [Fqδ δ − Fqq̇ (q̇/V ) − Fqq q]
(dimensional).
5-3.4 Sensors
The following component-specific sensors are implemented, in addition to the standard sensors of a structural
dynamic component.
5-3.4.1 Modal Sensors
The sensor quantity is the modal reaction (such as load, stress, or strain), calculated as

∂z
x= θk
∂qk
k
from the input modal reaction coefficients.

Chapter 6
FINITE ELEMENT BEAM COMPONENT
6–1 Description
A finite element beam component is a structural dynamic component. It has the standard features, and
the elastic motion is described by the axial, bending, and torsion deflection of a beam. The designation of the
component type refers to the technological basis of the theory, but the analysis does not use a finite element
approach. In particular, the equations of motion are obtained from a component Lagrangian rather than from a
system Lagrangian. A consequence of this approach is that all constraint forces appear explicitly in the component
equations. The following beam models are implemented:
a) Euler-Bernoulli beam theory for isotropic materials with an elastic axis, the undistorted elastic axis straight
within the component;
b) beam theory for anisotropic or composite materials including transverse shear deformation, the undistorted
beam axis straight within the component.
The effects of cross-section warping are included in the section structural properties. The component rigid body
motion can be large, and the kinematics of the interfaces and rigid body motion are always exact. The kinematics
of the elastic motion can be exact. For an almost-exact model, a second-order approximation is used for the
extension and torsion produced by bending. Alternatively, the equations of motion can retain only second-order
effects of elastic motion in the strain energy and kinetic energy, restricting the elastic motion to moderate deflection.
The structural model assumes that the strain is small. In addition, the geometric model requires that the cross-
section rotation produced by bending (relative the body axes at one end) be less than 90 degrees within each beam
component; so very large elastic motion must be modelled using several beam components.
For this component, the rigid motion describes the motion of one end of the beam. The elastic motion is
measured relative to the large rigid motion. Thus the model does not use nodal coordinates as degrees of freedom.
The beam axis is straight within the component. The component can have any number of locations (on the beam
axis); any number of joints at a location; any number of connections at a location or at a joint; and any number
of interfaces and sensors at a connection. Typically joints and connections are defined in a beam section (a plane
normal to the bent beam axis). Figure 1 illustrates the use of a beam component.
It is assumed that the component has a straight beam axis (undeflected). Thus a structure described by a
beam axis with curvature, kinks, or jumps must be modelled by defining nodes to break it up into several straight
beam segments. Here “nodes” are points where segments are joined (in CAMRAD II terminology, these are
connections at a joint or location on each structural dynamic component, with a structural dynamic interface
between). The equations of motion require integration of the beam properties (such as mass and stiffness) along
46 FINITE ELEMENT BEAM COMPONENT
component rigid body

degrees of freedom
component elastic
degrees of freedom
structural dynamic interface,

joint eliminates rigid body degrees of
freedom
connection and structural

dynamic interface
J
J beam axis locations (on beam axis)
J joint
Figure 6-1 Use of beam component.

FINITE ELEMENT BEAM COMPONENT 47
its length. Gaussian integration is used, implying a polynomial approximation to the variation of the properties,
which is accurate only if the variation is sufficiently smooth. Also, the shape functions for the elastic motion
are continuous, and can not accurately represent large changes in the curvature or slope. Thus a beam having
properties that vary rapidly along its length must be also modelled by defining nodes to break it into segments,
with the major jumps in properties at the nodes. If very short beam segments are required to accommodate the
properties, then beam theory is probably not applicable.
6-2.1 Beam Model
The elastic motion of the component is described by the axial, bending, and torsion deflection of a beam.
It is assumed that the beam axis is straight within the component. The geometry and the structural and inertial
properties can be defined in a general manner relative the beam axis. Yet identification of the beam axis is not
entirely arbitrary, since it has the following consequences.
a) The structure must be slender relative the beam axis, allowing application of the beam theory
assumptions.
b) Structural and inertial properties, including the centroid offset and twist of the principal
axes, are defined in planes perpendicular to the beam axis.
c) The elastic motion is described by extension, bending, and torsion of the beam axis. The
axes of the engineering strain and section elastic loads are defined by the orientation of the
beam axis.
d) The beam axis defines the origin and orientation of the component body axes, hence the
component rigid body motion.
As for all structural dynamic components, the component rigid body motion can be large, and the kinematics of
the interfaces and rigid body motion are always exact. With the assumption of small strain, beam theory produces
a linear relationship between the section structural loads and the strain measures (such as curvature). Some
interesting (if not important) nonlinear terms are retained as well, coupling extension and torsion. The kinematics
of the elastic motion, including these strain measures, can be exact. For an almost-exact model, a second-order
approximation is used for the extension and torsion produced by bending. Alternatively, the equations of motion
can retain only second-order effects of elastic motion in the strain energy and kinetic energy. These second-order
models do not however compromise the correctness of the kinematics at interfaces. The derivation of the model
with exact kinematics is influenced by references 1 and 2.
A beam theory for anisotropic or composite materials is developed, including transverse shear deformation.
This development is based on references 3 to 6. The effects of cross-section warping and transverse shear are
included in the section structural properties; their effects on the inertial forces and interface geometry are neglected.
Any variables describing the warp amplitude are eliminated by expressing them in terms of the strain measures.
This treatment of warp might be a concern with open sections, or restrained warping at end conditions. Transverse
shear is introduced by variables that describe the rotation of the cross-section relative the plane perpendicular to
the bent beam axis. These variables could be retained as separate degrees of freedom of the component. It would
also be necessary then to implement a kind of structural dynamic interface that enforces equality of strain between
two components, in addition to equality of position and orientation (enforced by the cantilever interface). There
are several reasons for not using such an approach. A beam model only deals with two of the three shear strains,
so the interface would have to be constructed to allow use of a subset of the strains (in appropriate axes). A proper
method has not been developed to enforce continuity of shear strains at a node where there is a change in position
or orientation of the beam axis, or at a node where more than two beams are connected. The transverse shear
variables would be static or very high frequency degrees of freedom, so a quasistatic reduction would be used in
the solution procedure (although for a subsystem or component, rather than a beam section). Finally, good results
have been obtained using reduced section properties (refs. 7 to 9). Hence in the present approach, a quasistatic
reduction is used to eliminate the transverse shear variables from the section structural relations.
As a simplification, Euler-Bernoulli beam theory for isotropic materials can be used, with the assumption
that an elastic axis exists and is coincident with the beam axis. The Euler-Bernoulli beam theory model with
second-order approximations for the elastic motion is based on reference 10.
6-2.2 Rigid Body Motion
A component frame F is identified, and all motion measured relative to that frame. The component has rigid
body motion, described by the motion of body axes B. The undeflected structure has a straight beam axis, of
length . The rigid body motion is the motion of one end of the beam. The beam axis is assumed to be on the
positive x-axis of the B frame. Thus the origin of the B axes is located at one end of the beam, with the beam
extending from x = 0 to x = . Figure 2 illustrates the configuration. The motion BF describes the orientation and
displacement of the beam end, relative to the frame. In order to interpret the rigid body degrees of freedom as the
beam motion in B axes, the response should be defined as degrees-of-freedom relative nominal. The nominal part
of the response BF typically is the rest position of the beam relative to the frame. For a constrained component,
BF is a constant position.
6-2.3 Beam Properties
Beam properties are defined relative to the beam axes B. Displacements y and z are measured from the
beam axis (the x-axis). Pitch angles are measured from the x–y plane, positive for rotation about the x-axis. The
structural and inertial properties of the undeflected beam are defined as follows.
a) θC : pitch of the structural principal axes, relative the x–y plane. The pitch angle can be large.
b) yC and zC : offset of the tension center (modulus-weighted centroid) from the beam axis, relative the principal
axes (at θC ).
c) kP : modulus-weighted radius of gyration, about the beam axis.
d) θI : pitch of the inertial principal axes, relative the x–y plane. The pitch angle can be large.
e) yI and zI : offset of the center of gravity (mass-weighted centroid) from the beam axis, relative the principal
axes (at θI ).
Additional structural and inertial properties of the beam section are required by the model, and are defined below.
The beam properties are specified in terms of a piecewise-linear distribution in x (a step in the input properties is
approximated by a steep ramp, using two close values of x). Then the properties can be evaluated at an arbitrary
point on the beam by linear interpolation of the input values.
beam (with elastic)

y E
B x
E beam (no elastic)
rigid motion BF
degrees of freedom
F nominal (typically rest position)
Figure 6-2 Beam configuration.

6-2.4 Locations, Joints, and Connections
Locations are defined to be on the beam axis at axial station x, with the undeflected location axes parallel the
body axes. Hence without elastic motion, the position of a location relative the body axes (EB) is xEB/B = (x 0 0)T
and C EB = I. Elastic motion produces an additional displacement and rotation of the location axes. Thus the
E axes are the bent and twisted section axes, with origin on the beam axis. Joint and connection geometry are
specified using the standard conventions for structural dynamic components. A connection has a constant position
relative the location axes (relative the joint axes if the connection is at a joint), specified by z CJ/J and C CJ
(C relative E, in E axes; or C relative J, in J axes). The position of the joint relative the location axes (without
joint motion) is specified by z DE/E and C DE (D relative E, in E axes). Joints and connections usually are on a
plane normal to the beam axis at the location axial station x, but in general the specification can include an axial
displacement as well.
A typical application requires that the connection be a distance Δz F from the beam axis at x, and the
connection axes be parallel to the frame axes (for no rigid or bending motion). This is accomplished using
z CJ/J = C BF Δz F
C CJ = C F B
with the location axes parallel to the body axes (C EB = I) and parallel to the joint axes (z JE/E = 0 and C JE = I,
or no joint). Another possibility is for the connection to be on the y–z plane of the location (at axial station x).
The connection axes are rotated relative the location axes by angle θX about the x-axis. The displacement in the
plane (y,z) is measured relative the connection axes at θX . Thus
⎛ ⎞
0
z CJ/J = C JC ⎝ y ⎠
z
C CJ = XθX
For example, yC , zC , and θC give a connection at the tension center; or yI , zI , and θI give a connection at the
center of gravity.
6-2.5 Sensors
The component-specific sensors can measure the section torsion and bending moments, the axial tension, and
the section shear forces. These sensors provide the beam section load at a location (axial station xL ), acting at the
tension center, in the structural principal axes. The section load can be calculated from the deflection, or by force
balance. The section load can also be calculated at a node using the standard sensor for the structural dynamic
load.
The deflection method obtains the section load from the elastic motion and structural coefficients. The
shear forces are not available with this method. Essentially the load is evaluated from the stiffness and elastic
displacement at xL : moment = EI × curvature. The accuracy of this calculation depends on the accuracy of the
representation of the curvature or slope (the product of the degrees of freedom and shape functions). At a step in
stiffness there should be a corresponding step in curvature or slope, such that the load remains continuous. With a
small number of shape functions it is not possible to simulate such a step well, so the results for the reaction will
not be accurate near a step in stiffness. Also, the theory does not imply continuity of curvature on the two sides of
a node.
The force balance method obtains the section load from the difference between the applied forces and inertial
forces acting on the beam segment to one side of xL . The section loads calculated using the forces on either side
of xL are combined, so that this sensor gives at the beam ends the same result as the nodal reaction. All forces and
moments acting on a structural dynamic component are discretized (although the shape functions are not consistent
with discrete loads except at the beam ends). The force balance method can capture the steps in the section load
produced by discrete loads on the beam. If appropriate, such steps can be eliminated for the aerodynamic forces
by treating them as distributed loads in the force balance method.
Alternatively, the beam reaction can be calculated at a node using a standard sensor for the structural dynamic
load (about a point at the tension center, in the structural principal axes). The accuracy of this nodal reaction only
depends on the tolerance in the solution for equilibrium of the beam. However, it is necessary to define a node
(structural dynamic interface) at the sensor point.
6–3 Theory
The elastic motion affects the position of a location relative to the body origin (EB). EB is the combination of
elastic degrees of freedom and a constant position. The B axes (rigid body motion) have origin at one end of the
beam, and the x-axis along the beam axis. Locations are defined to be on the beam axis at axial station x, with the
undeflected location axes parallel the body axes. Elastic motion produces an additional displacement and rotation
of the location axes. Thus the E axes are the bent and twisted section axes, with origin on the beam axis. The
position of the location relative the body axes (E relative B, in B axes), is constructed as follows.
a) Constant axial position x;

b) then elastic axial deflection u along the x-axis;
c) then elastic bending deflections v then w, along the y and z-axes respectively; this bending produces a rotation
of the location axes;
d) then elastic torsion, and a constant rotation θX about the x-axis.
At a location, the angle θX = 0. For the structural and inertial analyses of the beam, appropriate principal axes
are used, so θX = θC or θX = θI . The order of the bending (v then w) follows from the use of Euler angles to
describe the rotation of the section. If Rodrigues parameters were used instead, the bending deflections would be
treated identically. Thus the position is
⎛ ⎞
x+u
xEB/B = ⎝ v ⎠
w
C EB = Xθ Y−β Zζ
Transverse shear and warping deflections are not considered here (they are only needed for the beam structural
analysis). Hence the cross-section is still perpendicular to the bent beam axis, and β and ζ are the rotations of the
cross-section produced by bending deflection. The rotation angles are obtained from the kinematics of the elastic
deflection. Let r be the arc length along the deflected beam axis. The notation (. . .)+ is used for the derivative
with respect to r, while (. . .) is the derivative with respect to x. The tangent to the beam axis is
dxEB/B 1 dxEB/B
tB = =
dr r dx
⎛ + ⎞ ⎛ ⎞
x + u+ 1 + u
1
= ⎝ v+ ⎠ = ⎝ v ⎠
r
w +
w
where EB/B
dr dx
r = = = (1 + u )2 + v 2 + w 2
dx dx
Since the first row of Y−β Zζ equals tB :
⎛ ⎞ ⎛ ⎞
cos β cos ζ ...
⎝ cos β sin ζ ⎠ = ⎝ v + ⎠
sin β w+
the rotation angles can be obtained from the bending:
sin β = w+

sin ζ = v + / 1 − w+2
So β is a rotation about the negative y-axis, produced by bending w ; and ζ is a rotation about the z-axis, produced
√
by bending v . The magnitudes of w+ and v + / 1 − w+2 are less than one for values of u, v, and w describing
a realizable deflection of the beam. Note however that the polynomial shape functions used can violate this
requirement, giving an inconsistent geometric model. Also, it is assumed that the magnitudes of β and ζ are less
than 90 degrees. So very large elastic motion must be modelled using several beam components. The linear and
angular motion of the location E are then evaluated as follows.
⎛ ⎞
x+u
⎜ ⎟
xEB/B = ⎝ v ⎠
w
⎛ ⎞
u̇
⎜ ⎟
ẋEB/B = ⎝ v̇ ⎠
ẇ
⎛ ⎞
ü
⎜ ⎟
ẍEB/B = ⎝ v̈ ⎠
ẅ
v EB/E = C EB ẋEB/B
v̇ EB/E = C EB ẍEB/B − ω
EB/E C EB ẋEB/B
⎡ ⎤
Cβ Cζ Cβ Sζ Sβ
⎢ ⎥
C EB = Xθ Y−β Zζ = Xθ ⎣ −Sζ Cζ 0 ⎦
−Sβ Cζ −Sβ Sζ Cβ
⎡ ⎤
1 0 0
⎢ ⎥
Xθ = ⎣ 0 Cθ Sθ ⎦
0 −Sθ Cθ
⎡ ⎤ ⎡ ⎤
1 0 Sβ 1 0 Sβ
⎢ ⎥ ⎢ ⎥
R = Xθ ⎣ 0 −1 0 ⎦ = ⎣0 −Cθ Sθ Cβ ⎦
0 0 Cβ 0 Sθ Cθ Cβ
⎛ ⎞
θ̇
⎜ ⎟
ω EB/E = R ⎝ β̇ ⎠
ζ̇
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
θ̈ θ̇ θ̈ θ̇ θ̇ − Sβ ζ̇
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟
ω̇ EB/E = R ⎝ β̈ ⎠ + Ṙ ⎝ β̇ ⎠ = R ⎝ β̈ ⎠ + ⎝ −Cθ β̇ ⎠ ⎝ −Sθ Cβ ζ̇ ⎠
ζ̈ ζ̇ ζ̈ Sθ β̇ −Cθ Cβ ζ̇
⎛ ⎞ ⎛ ⎞
θ̈ Cβ β̇ ζ̇
⎜ ⎟ ⎜ ⎟
= R ⎝ β̈ ⎠ + ⎝ Cθ Cβ θ̇ζ̇ + (θ̇ − Sβ ζ̇)Sθ β̇ ⎠
ζ̈ −Sθ Cβ θ̇ζ̇ + (θ̇ − Sβ ζ̇)Cθ β̇
using the notation Cθ = cos θ, Sθ = sin θ; Cβ = cos β, Sβ = sin β; and Cζ = cos ζ, Sζ = sin ζ. Similarly, the
virtual rotation and the torsion/curvature are:
⎛⎞
δθ
δψ = R ⎝ δβ ⎠
δζ
⎛⎞
θ+
κ = R ⎝ β+ ⎠
ζ+
EB/E = C EB Ċ BE , and κ
Note that ω = C EB (C BE )+ . An exact evaluation of the geometry requires the
following quantities.

Sβ = w + Cβ = 1 − Sβ2

Sζ = v + / 1 − w+2 Cζ = 1 − Sζ2
! !
Sβ w+
=
Sζ Cβ v+
! " # ! !
β̇ Cβ 0 β̇ ẇ+
A = =
ζ̇ −Sβ Sζ Cβ Cζ ζ̇ v̇ +
" #
1 Cβ Cζ 0
A−1 = 2
Cβ Cζ Sβ Sζ Cβ
! " # !
β̇ 1 Cβ Cζ 0 ẇ+
= 2
ζ̇ Cβ Cζ Sβ Sζ Cβ v̇ +
! " # !
β̈ 1 Cβ Cζ 0 ẅ+
= 2
ζ̈ Cβ Cζ Sβ Sζ Cβ v̈ +
" # !
Sβ β̇/Cβ2 0 ẇ+
+
(1 + Sβ2 )Sζ β̇/Cβ3 Cζ + Sβ ζ̇/Cβ2 Cζ2 Sβ β̇/Cβ2 Cζ + Sζ ζ̇/Cβ Cζ2 v̇ +
w+ = w /r

ẇ+ = ẇ /r − w ṙ /r 2 = ẇ − w+ ṙ /r

ẅ+ = ẅ − 2ẇ+ ṙ − w+ r̈ /r
The structural analysis also requires the following quantities.
! " # !
δβ 1 Cβ Cζ 0 δ(w+ )
= 2
δζ Cβ Cζ Sβ Sζ Cβ δ(v + )
! " # !
β 1 Cβ Cζ 0 (w+ )
= 2
ζ Cβ Cζ Sβ Sζ Cβ (v + )

δ(w+ ) = δ(w /r ) = δw − w+ δr /r

(w+ ) = (w /r ) = w − w+ r /r
Then from
⎛ ⎞ ⎛ ⎞
! " # ṙ " # ṙ
ẇ +
−w +
0 1 1 ⎜ ⎟ −Sβ 0 1 1 ⎜ ⎟
= ⎝ v̇ ⎠ = ⎝ v̇ ⎠
v̇ +
−v +
1 0 r −Cβ Sζ 1 0 r
ẇ ẇ
the derivatives can be written as follows:
⎛ ⎞
! " # ṙ
β̇ 1 −Sβ Cβ Cζ 0 Cβ Cζ 1 ⎜ ⎟
= ⎝ v̇ ⎠
ζ̇ Cβ2 Cζ −Sζ Cβ Sβ Sζ r
ẇ
⎛ ⎞
! " # δr
δβ 1 −Sβ Cβ Cζ 0 Cβ Cζ 1 ⎜ ⎟
= ⎝ δv ⎠
δζ Cβ2 Cζ −Sζ Cβ Sβ Sζ r
δw
⎛ ⎞
! " # r
β 1 −Sβ Cβ Cζ 0 Cβ Cζ 1 ⎜ ⎟
= ⎝v ⎠
ζ Cβ2 Cζ −Sζ Cβ Sβ Sζ r
w
and
⎛ ⎞
! " # r

−β 1 Sβ Cζ 0 −Cζ 1 ⎜ ⎟
= ⎝v ⎠
Cβ ζ Cβ Cζ −Sζ Cβ Sβ Sζ r
w
⎛ ⎞
! " # δr − (r /r )δr

−β 1 Sβ Cζ 0 −Cζ
1 ⎜ ⎟
δ = ⎝
δv − (v /r )δr ⎠

Cβ ζ Cβ Cζ −Sζ Cβ Sβ Sζ r
δw − (w /r )δr
⎛ ⎞
" # r
1/Cβ2 0 −Sβ /Cβ2 1 ⎜ ⎟
+ δβ ⎝ v ⎠
−Sβ Sζ /Cβ2 Cζ 0 Sζ /Cβ2 Cζ r
w
⎛ ⎞
" # r
0 0 0 1 ⎜ ⎟
+ δζ ⎝ v ⎠
−1/Cβ Cζ2 Sζ /Cζ2 Sβ /Cβ Cζ2 r
w
⎛ ⎞
" # δr
1 Sβ Cζ 0 −Cζ 1 ⎜ ⎟
⎝
= δv ⎠
Cβ Cζ −Sζ Cβ Sβ Sζ r
δw
⎛ ⎞
" # δr
−Sβ (1 + Cβ )/Cβ
2 3
0 1/Cβ 3
r ⎜ ⎟
2 ⎝
+ δv ⎠
Sζ (1 + Cζ2 )/Cβ3 Cζ3 −1/Cβ2 Cζ3 −Sβ Sζ (1 + Cζ2 )/Cβ3 Cζ3 r
δw
⎛ ⎞
" # δr
0 0 0
v ⎜ ⎟
2 ⎝
+ δv ⎠
−(Sζ2 + Cβ2 Cζ2 )/Cβ2 Cζ3 Sζ /Cβ Cζ3 Sβ Sζ2 /Cβ2 Cζ3 r
δw
⎛ ⎞
" # δr
1/Cβ 3
0 −Sβ /Cβ3
w ⎜ ⎟
2 ⎝
+ δv ⎠
−Sβ Sζ (1 + Cζ2 + Cβ2 Cζ2 )/Cβ3 Cζ3 Sβ /Cβ2 Cζ3 Sζ (Sβ2 + Cζ2 )/Cβ3 Cζ3 r
δw
For moderate deflections, it is possible to simplify these relations, consistent with second-order accuracy of the
equations of motion. Hence the second-order approximation for the geometry uses the following expressions.

Sβ = w Cβ = 1 − Sβ2

Sζ = v Cζ = 1 − Sζ2
! !
β̇ ẇ /Cβ
=
ζ̇ v̇ /Cζ
! !
β̈ ẅ /Cβ + β̇ 2 Sβ /Cβ
=
ζ̈ v̈ /Cζ + ζ̇ 2 Sζ /Cζ
! !
δβ δw /Cβ
=
δζ δv /Cζ
! !
β w
∼
=
ζ v
! !
−β −δw
δ ∼
=
Cβ ζ δv
With these approximations, the rotation matrix is still proper, and so the kinematics of connections with other
components remain exact. Further approximations are possible for the equations of motion.
6-3.2 Extension and Torsion Produced by Bending
Bending of the beam (v and w deflection) produces axial and torsional displacements. The extension u and
pitch angle θ of a bent beam are thus nonzero even for large axial and torsional stiffnesses. These variables are
therefore defined as the sum of elastic motion and motion produced by bending:
u = ue + U
θ = θC + φ + Θ
Here ue and φ are quasi-coordinates for the elastic extension motion and elastic torsion motion respectively. For
large axial and torsional stiffnesses, ue and φ approach zero. Bending deflection produces the extension U and
rotation Θ. The first term in θ is the pretwist of the structural principal axes (which can be replaced by θI or zero,
depending on the geometry required). The elastic torsion φ is defined considering the curvature of the beam about
the x-axis:
κx = θ+ + Sβ ζ + = (θC + φ)+
Hence the torsional displacement produced by bending is:
$ r $ x $ x $ x
+
Θ=− Sβ ζ dr = −
+
Sβ ζ r dx = − Sβ ζ dx = − w ζ + dx
0 0 0 0
If there is no elastic extension of the beam, then r = dr/dx = 1, which gives the axial displacement produced by
bending:

ubend = 1 − (v 2 + w 2 ) − 1
Typically therefore the total axial displacement is written
$ x
u = ue + 1 − (v 2 + w 2 ) − 1 dx = ue + U
0
and then
u = ue + 1 − (v 2 + w 2 ) − 1

r 2 = (1 + u )2 + v 2 + w 2 = 1 + ue2 + 2ue 1 − (v 2 + w 2 )
It is simpler (and equivalent to second order) to instead define the elastic extension as r = 1 + ue , so

u = (1 + ue )2 − (v 2 + w 2 ) − 1
$ x
u = ue + (1 + ue )2 − (v 2 + w 2 ) − (1 + ue ) dx = ue + U
0
Evaluation of the geometry requires the following quantities:
θ = θC + φ + Θ u = ue + U r = 1 + ue
θ̇ = φ̇ + Θ̇ u̇ = u̇e + U̇ ṙ = u̇e
θ̈ = φ̈ + Θ̈ ü = üe + Ü r̈ = üe
δθ = δφ + δΘ δu = δue + δU δr = δue
To second order in the displacement (or third order if ue = 0), the extension and torsion produced by bending are
as follows: $
1 x 2
U2 = − (v + w 2 ) dx
2 0
$ x
Θ2 = − w v dx
0
These approximations for U and Θ are used for the second-order and almost-exact geometric models. They are
accurate for moderate deflection, specifically as long as v 2 , w 2 , and ue are small compared to 1. For the exact
geometric model, the extension and torsion produced by bending are written
U = U2 + ΔU
Θ = Θ2 + ΔΘ
The additional terms are as follows.

$ % &
x 1 2
ΔU = (1 + ue )2 − (v 2 + w 2 ) − (1 + ue ) 2
+ (v + w ) dx
0 2
$ x
ΔU̇ = D (1 + ue )u̇e − (v v̇ + w ẇ ) − u̇e + v v̇ + w ẇ dx
0
$ x
ΔÜ = D (1 + ue )üe − (v v̈ + w ẅ ) + u̇e2 − (v̇ 2 + ẇ 2 )
0
2
−D3 (1 + ue )u̇e − (v v̇ + w ẇ ) − üe + v v̈ + w ẅ + v̇ 2 + ẇ 2 dx
$ x
δ(ΔU ) = D(1 + ue ) − 1 δue − (D − 1)v δv − (D − 1)w δw dx
0
⎛ ⎞
δqu
⎜ ⎟
= [ Uu Uv Uw ] ⎝ δqv ⎠
δqw

D−1 = (1 + ue )2 − (v 2 + w 2 )
$ x
ΔΘ = [−Sβ ζ + w v ] dx
0
$ x
ΔΘ̇ = −Sβ ζ̇ − Cβ β̇ζ + w v̇ + ẇ v dx
0
$ x
ΔΘ̈ = −Sβ ζ̈ − 2Cβ β̇ ζ̇ − (Cβ β̈ − Sβ β̇ 2 )ζ + w v̈ + 2ẇ v̇ + ẅ v dx
0
$ " #
x
Sβ δβ
δ(ΔΘ) = − δ(Cβ ζ ) − 2 (Cβ ζ ) + w δv + δw v dx
0 Cβ Cβ
⎛ ⎞
δqu
⎜ ⎟
= [ Θu Θv Θw ] ⎝ δqv ⎠
δqw
⎛ ⎞
r /r
1 ⎜ ⎟
ζ = [ −Sζ Cβ Sβ Sζ ] ⎝ v /r ⎠
Cβ2 Cζ
w /r
⎛ ⎞
(r /r )˙
1 ⎜ ⎟
ζ̇ = [ −Sζ Cβ Sβ Sζ ] ⎝ (v /r )˙ ⎠
Cβ2 Cζ
(w /r )˙
⎛ ⎞
r /r
−2Sβ Sζ β̇/Cβ3 Cζ Sβ β̇/Cβ2 Cζ (1 + Sβ2 )Sζ β̇/Cβ3 Cζ ⎜ ⎟
+ ⎝ v /r ⎠
−ζ̇/Cβ2 Cζ2 +Sζ ζ̇/Cβ Cζ2 2 2
+Sβ ζ̇/Cβ Cζ
w /r
⎛ ⎞
(r /r )¨
1 ⎜ ⎟
ζ̈ = [ −Sζ Cβ Sβ Sζ ] ⎝ (v /r )¨⎠
Cβ2 Cζ
(w /r )¨
⎛ ⎞
(r /r )˙
−2Sβ Sζ β̇/Cβ3 Cζ Sβ β̇/Cβ2 Cζ (1 + Sβ2 )Sζ β̇/Cβ3 Cζ ⎜ ⎟
+2 ⎝ (v /r )˙ ⎠
−ζ̇/Cβ2 Cζ2 +Sζ ζ̇/Cβ Cζ2 2 2
+Sβ ζ̇/Cβ Cζ
(w /r )˙
⎡ −2Sβ Sζ β̈/Cβ3 Cζ Sβ β̈/Cβ2 Cζ (1 + Sβ2 )Sζ β̈/Cβ3 Cζ ⎤

⎛ ⎞
⎢ −2(1 + 2Sβ2 )Sζ β̇ 2 /Cβ4 Cζ +(1 + Sβ2 )β̇ 2 /Cβ3 Cζ +Sβ (5 + Sβ2 )Sζ β̇ 2 /Cβ4 Cζ ⎥ r /r
⎢ ⎥
⎢ ⎥ ⎜ ⎟
+⎢ −4Sβ β̇ ζ̇/Cβ3 Cζ2 +2Sβ Sζ β̇ ζ̇/Cβ2 Cζ2 +2(1 + Sβ2 )β̇ ζ̇/Cβ3 Cζ2 ⎥ ⎝ v /r ⎠
⎢ ⎥
⎣ ⎦
−2Sζ ζ̇ 2 /Cβ2 Cζ3 +(1 + Sζ2 )ζ̇ 2 /Cβ Cζ3 +2Sβ Sζ ζ̇ 2 /Cβ2 Cζ3 w /r
−ζ̈/Cβ2 Cζ2 +Sζ ζ̈/Cβ Cζ2 +Sβ ζ̈/Cβ2 Cζ2
(w /r )˙ = (ẇ − (w /r )ṙ )/r

(w /r )¨ = (ẅ − 2(w /r )˙ ṙ − (w /r )r̈ )/r
The increments ΔU and ΔΘ are evaluated numerically. Using trapezoidal integration with M equally spaced
intervals gives
$ x/ M
x
I= f (ξ) dξ = f (eM M
i ) wi
0 M i=0
where
i x
eM
i =
M
'
wiM = 1/2 i = 0 or i = M
1 otherwise
Now
2M M M
f (e2M 2M
i ) wi = f (eM M
i ) wi + f (oM
i )
i=0 i=0 i=1
where
2i − 1 x
oM
i =
2M
Then doubling the number of integration steps gives

2M
"M M
#
x x 1
I2M = f (e2M
i ) wi
2M
= f (eM M
i ) wi + f (oM
i ) = IM + JM
2M i=0 2M i=0 i=1
2
M
x
JM = f (oM
i )
2M i=1
The process is started with M = 2, and repeated until the integral converges:
|I2M − IM | < (reference)
( is an input tolerance), or until the number of steps reaches a limit. The reference value is the beam length for
ΔU , and one radian for ΔΘ.
6-3.3 Elastic Variables and Shape Functions
The elastic motion of the beam is described by the variables ue , v, w, and φ, as a function of beam axial
station x. This motion is discretized using generalized coordinates q(t) and shape functions h(x), specifically
ue = hTu qu
v = hTv qv
w = hTw qw
φ = hTφ qφ
where q and h are vectors, of length Nu , Nv , Nw , and Nφ . The generalized coordinates q are the elastic degrees of
freedom of the beam. Thus the elastic degree-of-freedom vector of the component has length Nu +Nv +Nw +Nφ .
The number of generalized coordinates used affects the accuracy of the representation of the elastic motion of
the structure. If no elastic degrees of freedom are used, the model reduces to a rigid body (with properties and
geometry defined as for a beam). The analysis also has the capability to suppress any or all of the elastic degrees
of freedom during a part solution process. Note that the axial and torsion variables exclude the motion produced
by bending kinematics. So suppressing ue or φ (either in the component or in the solution process) is equivalent
to the limit of infinite axial or torsional stiffness.
The rigid motion of the entire component is contained in the motion of the body axes, which is the motion at
one end of the beam. The generalized coordinates q represent the elastic motion, measured relative to that rigid
motion. So the model does not use degrees of freedom that represent the total motion (nodal coordinates) for the
other end of the beam. A finite element analysis typically uses Hermite polynomials for the shape functions, so the
degrees of freedom are displacement and rotation at the nodes. Here the shape functions are instead orthogonal
polynomials for the elastic motion.
The shape functions are polynomials in ξ = x/ . Orthogonal polynomials normalized to unit amplitude
at x = are used. Thus the generalized coordinates q have the same units (ft or m, and radians) as the elastic
deflection and rotation. As implemented, the maximum number of shape functions is three for bending:
order h v , hw h 2
h
2
1 ξ 2ξ 2
2 6ξ 3 − 5ξ 2 18ξ 2 − 10ξ 36ξ − 10
3 28ξ 4 − 42ξ 3 + 15ξ 2 112ξ 3 − 126ξ 2 + 30ξ 336ξ 2 − 252ξ + 30
(h(0) = h (0) = 0 to exclude the rigid motion); and four for axial and torsion deflection:
order hu , h φ h 2
h
1 ξ 1 0
2 4ξ 2 − 3ξ 8ξ − 3 8
3 15ξ 3 − 20ξ 2 + 6ξ 45ξ 2 − 40ξ + 6 90ξ − 40
4 56ξ 4 − 105ξ 3 + 60ξ 2 − 10ξ 224ξ 3 − 315ξ 2 + 120ξ − 10 672ξ 2 − 630ξ + 120
(h(0) = 0 to exclude the rigid motion). Typically three shape functions are used for axial deflection, and two shape
functions each for bending and torsion (a fifteen degree-of-freedom component, six rigid and nine elastic); which
gives cubic displacements and quadratic rotation, hence quadratic tension and linear moments along the beam.
The effects of transverse shear are introduced by variables ω and ν that rotate the cross-section (in addition
to the rotation produced by bending). These variables can be nonzero at both ends, hence their shape functions
would be hω = hν = 1, 2ξ − 1, 6ξ 2 − 6ξ + 1, etc. However, by means of a static reduction the structural analysis
accounts for the transverse shear effects in the section elastic constants, so ω and ν do not remain as degrees of
freedom for the component.
Bending of the beam produces axial and torsional displacements. With the second-order approximation, U
and Θ can be expressed as quadratic functions of the bending degrees of freedom qv and qw :
$ x % $ & % $ &
1 2 2 1 x T 1 x T
U2 = − (v + w ) dx = qvT − h h dx qv + qw −
T
h h dx qw
2 0 2 0 v v 2 0 w w
= qvT Hvv qv + qw
T
Hww qw
$ x % $ x &
Θ2 = − w v dx = qw
T
− hw hT
v dx qv
0 0
T
= qw Hwv qv
where the integrals of the shape functions are as follows:

⎡ ⎤
H11 H12 H13
1
Hvv = Hww = − ⎣ H21 H22 H23 ⎦
2
H31 H32 H33
4 3
H11 = ξ
3
324 5 100 3
H22 = ξ − 90ξ 4 + ξ
5 3
22596 5
H33 = 1792ξ 7 − 4704ξ 6 + ξ − 1890ξ 4 + 300ξ 3
5
20
H12 = H21 = 9ξ 4 − ξ 3
3
224 5
H13 = H31 = ξ − 63ξ 4 + 20ξ 3
5
3388 5
H23 = H32 = 336ξ 6 − ξ + 450ξ 4 − 100ξ 3
5
⎡ ⎤
H H12 H13
1 ⎣ 11
Hwv = − 2 H21 H22 H23 ⎦
H31 H32 H33
H11 = 2ξ 2
H12 = 24ξ 3 − 10ξ 2
H13 = 168ξ 4 − 168ξ 3 + 30ξ 2
H21 = 12ξ 3 − 10ξ 2
H22 = 162ξ 4 − 180ξ 3 + 50ξ 2

6048 5
H23 = ξ − 1974ξ 4 + 1020ξ 3 − 150ξ 2
5
H31 = 56ξ 4 − 84ξ 3 + 30ξ 2
4032 5 2340 3
H32 = ξ − 1414ξ 4 + ξ − 150ξ 2
5 3
H33 = 6272ξ 6 − 14112ξ 5 + 11298ξ 4 − 3780ξ 3 + 450ξ 2
The derivatives of U2 and Θ2 are thus:
U̇2 = 2qvT Hvv q̇v + 2qw
T
Hww q̇w
Ü2 = 2qvT Hvv q̈v + 2qw
T
Hww q̈w + 2q̇vT Hvv q̇v + 2q̇w
T
Hww q̇w
δU2 = 2qvT Hvv δqv + 2qw
T
Hww δqw
Θ̇2 = qvT Hwv

T T
q̇w + qw Hwv q̇v
Θ̈2 = qvT Hwv
T T
q̈w + qw Hwv q̈v + q̇vT Hwv
T T
q̇w + q̇w Hwv q̇v
δΘ2 = qvT Hwv
T T
δqw + qw Hwv δqv
from which the derivatives of u and θ can be obtained.
The derivatives of the linear and angular motion of the section can now be expressed in terms of the generalized
coordinates (elastic degrees of freedom):
⎛ ⎞
⎛ ⎞ ⎡ ⎤ q̇u
u̇ hTu + Uu 2qvT Hvv + Uv T
2qw Hww + Uw 0 ⎜ ⎟
⎜ ⎟ ⎢ ⎥ ⎜ q̇v ⎟
⎝ v̇ ⎠ = ⎣ 0 hTv 0 0⎦⎜ ⎜ ⎟
⎟
⎝ q̇w ⎠
ẇ 0 0 hTw 0
q̇φ
⎛ ⎞
⎛ ⎞ ⎡ ⎤ q̇u
T
θ̇ Θu qw Hwv + Θv qvT Hwv
T
+ Θw hTφ ⎜ ⎟
⎜ ⎟ ⎢ ⎥⎜ q̇v ⎟
⎝ β̇ ⎠ = ⎣ −hu Sβ /Cβ r
T
0 hT
w /Cβ r

0 ⎦⎜
⎜ ⎟
⎟
⎝ w⎠
q̇
ζ̇ −hT 2
u Sζ /Cβ Cζ r

hT
v /Cβ Cζ r

hT 2
w Sβ Sζ /Cβ Cζ r

0
q̇φ
(exact); or
⎛ ⎞
⎛ ⎞ ⎡ ⎤ q̇u
T
θ̇ 0 qw Hwv qvT Hwv
T
hTφ ⎜ ⎟
⎜ ⎟ ⎢ ⎥ ⎜ q̇v ⎟
⎝ β̇ ⎠ = ⎣ 0 0 hT
w /Cβ 0 ⎦⎜⎜ ⎟
⎟
⎝ q̇w ⎠
ζ̇ 0 hT
v /Cζ 0 0
q̇φ
(second order). Similar relations give the virtual displacements. These expressions are required for the partial
velocity matrices used in the elastic equations of motion. The structural analysis requires for the curvature:
⎛ ⎧⎡
⎞ T ⎤
θ ⎪ Θu qw Hwv + Θv qvT Hwv
T
+ Θw hTφ
⎨
⎜ ⎟ ⎢ ⎥
δ ⎝ −β ⎠ = ⎣ hTu Sβ /Cβ r

0 −hT
w /Cβ r

0 ⎦
⎪
⎩
Cβ ζ −hT
u Sζ /Cβ Cζ r

hT
v /Cζ r

hT
w Sβ Sζ /Cβ Cζ r

0
⎡ ⎤
0 0 0 0
⎢ ⎥ r
+ ⎣ −hT 2 3
u Sβ (1 + Cβ )/Cβ 0 hT 3
w /Cβ 0⎦ 2
r
hT 2 3 3
u Sζ (1 + Cζ )/Cβ Cζ −hT 2 3
v /Cβ Cζ −hT 2 3 3
w Sβ Sζ (1 + Cζ )/Cβ Cζ 0
⎡ ⎤
0 0 0 0
⎢ ⎥ v
+⎣ 0 0 0 0⎦ 2
r
−hT 2 2 2 2 3
u (Sζ + Cβ Cζ )/Cβ Cζ hT 3
v Sζ /Cβ Cζ hT 2 2 3
w Sβ Sζ /Cβ Cζ 0
⎛ ⎞
⎡ ⎤ ⎫ δqu
0 0 0 0 ⎪
⎬⎜ ⎟
⎢ ⎥ w ⎜ δqv ⎟
+⎣ hT 3
u /Cβ 0 −hT 3
w Sβ /Cβ 0⎦ 2 ⎜ ⎜
⎟
⎟
r ⎪⎭ ⎝ δqw ⎠
−hT 2 2 2 3 3
u Sβ Sζ (1 + Cζ + Cβ Cζ )/Cβ Cζ hT 2 3
v Sβ /Cβ Cζ hT 2 2 3 3
w Sζ (Sβ + Cζ )/Cβ Cζ 0
δqφ
⎛ ⎞
δqu
⎜ ⎟
⎜ δqv ⎟
⎜
= D⎜ ⎟
⎟
⎝ δqw ⎠
δqφ
(exact); or
⎛ ⎞ ⎡ ⎤ ⎛ δqu ⎞
θ 0 0 0 hTφ
⎜ δqv ⎟
δ ⎝ −β ⎠ = ⎣ 0 0 −hT
w 0 ⎦⎝ ⎠
δqw
Cβ ζ 0 hT
v 0 0
δqφ
(second order).
6-3.4 Integration Over Beam Length
The equations of motion require that quantities be integrated over the length of the beam. This integration is
performed numerically, using Gaussian quadrature. Thus
$ $ 1 N
f (x) dx = f x= (ξ + 1) dξ = wi f (xi ) + O(f (2N ) )
0 2 −1 2 2 i=1
where
xi = (ξi + 1)
2
for Gaussian points ξi and weights wi . Quantities in the integrand must be evaluated at the Gaussian integration
points. The shapes functions can be evaluated at an arbitrary point on the beam. The beam properties are specified
in terms of a piecewise-linear distribution in x (a step in the input properties is approximated by a steep ramp,
using two close values of x). Then the properties can be evaluated at an arbitrary point on the beam (including the
Gaussian points) by linear interpolation of the input values. The shape functions are continuous, and the properties
are only used at the Gaussian integration points. Thus a beam having properties that vary rapidly along its length
must be modelled by breaking it into segments, with the major jumps in properties at the nodes.
6-3.5 Beam Cross-Section Motion
The theory requires the motion of a point on the beam cross-section. The geometry has been described above
for the standard locations and connections of the component. For the structural contributions to the equations of
motion, the effects of warp and transverse shear must be considered as well. Hence the position of a cross-section
point relative the body axes is constructed as follows.
a) Constant axial position x;

b) then elastic axial deflection u along the x-axis;
c) then rotation of the cross-section by ν then ω, produced by transverse shear deformation;
d) then elastic bending deflections v then w; this bending produces a rotation of the cross-section (location) axes;
e) then elastic torsion, and a constant rotation θX about the x-axis;
f) then the position relative the bent and rotated cross-section axes, including warp W of the cross-section.
The pitch angle θX = θC for the structural analysis, and θX = θI for the inertial analysis (section principal axes).
The warp displacement W can have three components. Thus the position on the cross-section is:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
x+u 0 W1
r B = ⎝ v ⎠ + C ⎝ η ⎠ + C ⎝ W2 ⎠
w ζ W3
where the variables η and ζ identify the cross-section point, relative the section principal axes at θX . The variables
x, η, and ζ are curvilinear coordinates of the beam. The section is rotated by the matrix C:
C = (Z−ν Yω )C BE = (Z−ν Yω )(Z−ζ Yβ X−θ )
The section warping displacement can in general be described by a set of warping functions Si and scalar amplitudes
Ai : ⎛ ⎞
W1
W = ⎝ W2 ⎠ = Si (η, ζ)Ai (x)
W3 i
(for example, ref. 7). From the virtual displacement δW , differential equations (in x) are obtained for the amplitudes
Ai (static equations if the inertial effects of warping are neglected). Here it is assumed that these equations are
solved to eliminate the warping variables, so the effects of warp are accounted for in the section elastic constants.
For an isotropic beam with an elastic axis, St.Venant’s torsional warping function can be used: W1 = λφ . For
simplicity, this expression may also be used in the equations presented here for an anisotropic beam, although the
analysis used to obtain the section elastic constants must fully consider the effects of warp.
6-3.6 Hamilton’s Principle
The component theory is based on finite element technology (refs. 11 and 12). The equations of motion are
obtained using Hamilton’s principle:
$ t2 $ t2
δ L dt = δ (T − U + W ) dt = 0
t1 t1
where the terms in the Lagrangian L are the kinetic energy, the strain energy, and the work of external loads.
In a finite element analysis, the Lagrangian is discretized. Here a component Lagrangian rather than a system
Lagrangian is considered, and the external loads include all constraint forces. Let u be the component displacement,
represented by generalized coordinates a using u = N a (N are the shape functions). The strain energy can be
written as an integral over the structure volume:
$
δU = δT σ dΩ
The material properties define the stress σ as a function of the strain ; and the geometry gives σ from u:
= Lu = LN a = Ba
σ = D + σ0
Hence the strain energy becomes:

%$ $ &
T T T
δU = δa B DB dΩ a + B σ0 dΩ
The generalization to nonlinear material properties and geometry is straightforward (σ() and (u)). The work of
the external loads can be written as integrals of the body forces b, surface forces t, and discrete force F :
$ $
δW = δuT b dΩ + δuT t dΓ + δuT F
%$ $ &
T T T T
= δa N b dΩ + N t dΓ + N F
For this component the surface forces are assumed to be discretized; and the only body force is gravity, which is
handled with the inertial loads. So the only work terms needed are those from the discrete forces acting on the
structure (and similar terms for the discrete moments).
The equations of motion can be obtained from Hamilton’s principle as the coefficient of δaT inside an integral
over time. The kinetic energy can be written as an integral over the structure volume:
$
1 2
δT = δ ρv dΩ
2
where ρ and v are the material density and velocity. Hence δ ȧ terms arise from the kinetic energy. These can
be eliminated by integrating by parts (in time), with the assumption δa = 0 at t1 and t2 . The integration and δ
operation are complicated for the nonlinear and time-varying case however. The appearance of δ ȧ can be avoided,
and so the integration over time ignored, by using the d’Alembert approach: inertial acceleration is treated as a
body force. So the kinetic energy can be written as:
$
δT = δuT (−r̈ + g)ρ dΩ
%$ &
T T
= δa N (−N ä + g)ρ dΩ
where for the linear case the acceleration relative inertial space is r̈ = ü = N ä. The gravity force acting on the
body has been included here. Thus
$ $ $ $ $
N T N ρ dΩ ä + B T DB dΩ a = N T ρ dΩ g + N T t dΓ + N T F − B T σ0 dΩ
are the equations of motion for the linear problem. The derivation of the equations of motion for this component
follows the above general approach.
6-3.7 Strain Energy
Evaluation of the strain energy begins with the analysis of strain (refs. 13 to 15). The Green-Lagrange strain
tensor is obtained from the metric tensors of the undistorted and distorted beams (gmn and Gmn ). In terms of
curvilinear coordinates ym = (x, η, ζ), the undistorted and distorted position vectors are:
⎛ ⎞ ⎛ ⎞
x 0
r = ⎝ 0 ⎠ + X−θC ⎝ η ⎠
0 ζ
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
x+u 0 W1
R = ⎝ v ⎠ + C ⎝ η ⎠ + C ⎝ W2 ⎠
w ζ W3
Here x is the distance along the straight beam axis, while η and ζ specify a position on the cross-section plane
(parallel to the structural principal axes, but the origin is not necessarily at the tension center). Assuming small
strain, the section loads can be expressed as linear combinations of the moment strain measure κ and force strain
measure γ:
κ=K −k
= CT C
K

k = XθC X−θ C
T
k = ( θC 0 0)
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 + u 1 11
γ = C T ⎝ v ⎠ − ⎝ 0 ⎠ = ⎝ 2 12 ⎠
w 0 2 13

(ref. 2). It will be shown that that Kx = θC + φ , so κx = φ ; and γx = 11 = ue . Writing C = Cs C BE , the
curvature K is the sum of shear and bending/torsion terms:
= C T C = C EB (C T C )C BE + C EB (C BE ) = C EB K
K s C BE + K
EB/E
s s
Hence to second order:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 φ φ
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
κ = Xθ ⎝ −ω ⎠ + Xθ ⎝ −β ⎠ = Xθ ⎝ −w − ω ⎠
ν Cβ ζ v + ν
⎡ ⎤⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 v + ν w + ω 1 + u 1 ue
⎢ ⎥⎜ ⎟ ⎜ ⎟ ⎜ ⎟
γ = Xθ ⎣ −(v + ν) 1 0 ⎦⎝ v ⎠ − ⎝ 0 ⎠ = Xθ ⎝ −ν ⎠
−(w + ω) 0 1 w 0 −ω
with θ ∼
= θC + φ here. Without the shear deformation, the second-order moment strain measure is:
⎛ ⎞
φ
⎜ ⎟
κ = ⎝ −Cθ w + Sθ v ⎠
Sθ w + Cθ v
⎛ ⎞
⎛ ⎞ ⎡ ⎤ δqu
δφ 0 0 0 hT
φ ⎜ ⎟
⎜ ⎟ ⎢ ⎥⎜
⎜
δqv ⎟
⎟
δκ = Xθ ⎝ −δw + v δφ ⎠ = Xθ ⎣ 0 0 −hT
w v hTφ ⎦⎜ ⎟
⎝ δqw ⎠
δv + w δφ 0 hT
v 0 w hTφ
δqφ
⎛ ⎞
⎡ ⎤ δqu
0 0 0 hT
φ ⎜ ⎟
⎢ ⎥⎜
⎜
δqv ⎟
⎟
= ⎣0 Sθ hT
v −Cθ hT
w κz hTφ ⎦⎜ ⎟
⎝ δqw ⎠
0 Cθ hT
v Sθ hT
w −κy hTφ
δqφ
For the geometrically exact model, rotation of the deformed section gives
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
θ θ + Sβ ζ θC + φ
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
K = R ⎝ β ⎠ = Xθ ⎝ −β ⎠ = Xθ ⎝ −β ⎠
ζ Cβ ζ Cβ ζ
From Θ = −Sβ ζ , there follows Kx = θ + Sβ ζ = θC

+ φ , and hence κx = φ exactly. Then the moment
strain measure is:
⎛ ⎞
φ
⎜ ⎟
κ = Xθ ⎝ −β ⎠
Cβ ζ
⎛ ⎞ ⎡ ⎤ ⎛ ⎞ ⎛ ⎞
0 0 0 θ δφ
! δφ !
δκ = Xθ ⎝ −β Cβ ζ ⎠ = Xθ ⎢
⎣ Cβ ζ 1
⎥ ⎜ ⎟ ⎜ ⎟
0 ⎦ δ ⎝ −β ⎠ + ⎝ 0 ⎠
δ + δθ
Cβ ζ β β 0 1 Cβ ζ 0
⎛ ⎞
⎡ ⎤ δqu
⎛ T ⎞
0 0 0 ⎜ ⎟ hφ
⎢ ⎥ ⎜ δqv ⎟ ⎜
⎟+⎝ 0 ⎟
= ⎣ κz Cθ Sθ ⎦ D ⎜
⎜ ⎟ ⎠ δqφ
⎝ w⎠
δq
−κy −Sθ Cθ 0
δqφ
using the matrix D as defined above. There follows:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
δue ⎡ hT u 0 0 0 ⎤ δqu δqu
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜ δφ ⎟ ⎢ 0 0 0 hT ⎥ ⎜ δqv ⎟ ⎜ δq ⎟
⎜ ⎟ = ⎢" φ
# ⎥⎜ ⎟=C⎜ v ⎟
⎜ ⎟ ⎣ κ C S ⎦⎜ ⎟ ⎜ ⎟
⎝ δκy ⎠ z θ θ ⎝ δqw ⎠ ⎝ δqw ⎠
D
δκz −κy −Sθ Cθ δqφ δqφ
(exact); or
⎛ ⎞ ⎡ ⎤⎛ ⎞ ⎛ ⎞
δue hT
u 0 0 0 δqu δqu
⎜ ⎟ ⎢ ⎥⎜ ⎟ ⎜ ⎟
⎜ δφ ⎟ ⎢ 0 0 0 hT ⎥ ⎜ δqv ⎟ ⎜ δq ⎟
⎜ ⎟=⎢ φ
⎥⎜ ⎟=C⎜ v ⎟
⎜ ⎟ ⎢ ⎥⎜ ⎟ ⎜ ⎟
⎝ δκy ⎠ ⎣ 0 Sθ hT
v −Cθ hT
w κz hTφ ⎦ ⎝ δqw ⎠ ⎝ δqw ⎠
δκz 0 Cθ hT
v Sθ hT
w −κy hTφ δqφ δqφ
(second order). An approximation for the pitch angle:
Cθ ∼
= cos(θC + φ) ∼
= cos θC − φ sin θC
= sin(θC + φ) ∼
Sθ ∼ = sin θC + φ cos θC
is consistent with the second-order model.
The basis vectors of the undistorted and distorted beam are gm = ∂r/∂ym and Gm = ∂R/∂ym respectively.
Then the metric tensors are gmn = gm ·gn and Gmn = Gm ·Gn ; and the Green-Lagrange strain tensor is obtained
from
1
fmn = (Gmn − gmn )
2
The basis vector g1 is tangent to the line described by constant η and ζ, which is a helix for a beam with pretwist

(θC = 0). So g1 is not perpendicular to g2 and g3 . Using the strain γmn = fmn is equivalent to assuming
that the axial stress follows the basis vectors in the twisted beam. It is generally more appropriate to assume
that the constitutive relation is defined in local rectangular Cartesian coordinates zk . The unit vectors of zk are
ek = (i, g2 , g3 ). Thus the strain tensor γmn is related to fmn by
∂zk ∂zl
fmn = γkl
∂ym ∂yn
where
∂zk
= e k · gm
∂ym
Here the basis vectors are:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 + u 0 1 0
G1 = ⎝ v ⎠ + C ⎝ η ⎠ + C W + CW g1 = ⎝ 0 ⎠ + X−θ

C
⎝η⎠
w ζ 0 ζ
⎛ ⎞ ⎛ ⎞
0 0
G2 = C ⎝ 1 ⎠ + CWη g2 = X−θC ⎝ 1 ⎠
0 0
⎛ ⎞ ⎛ ⎞
0 0
G3 = C ⎝ 0 ⎠ + CWζ g3 = X−θC ⎝ 0 ⎠
1 1
or
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 + u −Kz η + Ky ζ 1 0
C T G1 = C T ⎝ v ⎠ + ⎝ −Kx ζ + W
⎠ + KW XθC g1 = ⎝ 0 ⎠ + ⎝ −θC

ζ⎠

w Kx η 0 θC η
⎛ ⎞ ⎛ ⎞
0 0
C T G2 = ⎝ 1 ⎠ + W η X θ C g2 = ⎝ 1 ⎠
0 0
⎛ ⎞ ⎛ ⎞
0 0
C T G3 = ⎝ 0 ⎠ + W ζ X θ C g3 = ⎝ 0 ⎠
1 1
and the transformation ⎡ ⎤
% & 1 0 0
∂zk ⎢ ⎥
= ⎣ −θC ζ 1 0⎦
∂ym

θC η 0 1
gives γ11 = f11 + 2θC

(ζf12 − ηf13 ). With the assumption of small strain, γmn ∼
= mn , where mn is linear in
the strain measures. Thus
1
11 = (G11 − g11 ) + 2θC (ζ12 − η13 )
2
1
∼
= (1 + u )2 + v 2 + w 2 − 1 + 2(−Kz η + Ky ζ) + (Kx2 − θC
2
)(η 2 + ζ 2 ) + W1
2

+ 2θC (ζ12 − η13 )
∼
= ue − κz η + κy ζ + (θC
φ + 1/2φ 2 )(η 2 + ζ 2 ) + W1
2

+ 2θC (ζ 12 − η 13 ) − θC φ (ζ + η 2 ) + θC ζ(W1η + W2 ) − η(W1ζ + W3 )
∼
= ue − κz η + κy ζ + 1/2φ 2 (η 2 + ζ 2 ) + 2θC

(ζ 12 − η 13 ) + θC φ (ζλη − ηλζ )
212 = G12 − g12

∼
= 2 12 − (Kx − θC
)ζ + W1η + W2
∼ 2 12 + (λη − ζ)φ
=
213 = G13 − g13

∼
= 2 13 + (Kx − θC
)η + W1ζ + W3
∼
= 2 13 + (λζ + η)φ
is the required strain. The nonlinear term producing coupling between extension and torsion is conventionally
retained in 11 . In the final expression for each strain, the representative warping function W1 = λφ has been
used. It is understood that the complete effects of warp must be considered when the section elastic constants are
evaluated. Thus
⎛ ⎞
ue
⎜ ⎟
⎛ ⎞ ⎡ ⎤⎜
⎜
2 12 ⎟
⎟
11 1 θC ζ −θC η θC (ζλη − ηλζ ) + 1/2φ (η 2 + ζ 2 ) ζ −η ⎜ ⎟
⎜ ⎟ ⎢ ⎜
⎥⎜ 2 13 ⎟
⎝ 212 ⎠ = ⎣ 0 1 0 λη − ζ 0 0 ⎦⎜ ⎟
⎟
⎜ φ ⎟
213 0 0 1 λζ + η 0 0 ⎜⎜
⎟
⎟
⎝ κy ⎠
κz
⎛ ⎞
ue
⎜ ⎟
⎛ ⎞ ⎡ ⎤ ⎜
⎜
2 12 ⎟
⎟
11 1 θC ζ −θC η θC (ζλη − ηλζ ) + φ (η 2 + ζ 2 ) ζ −η ⎜ ⎟
⎜ ⎟ ⎢ ⎥ ⎜
⎜
2 13 ⎟
⎟
δ ⎝ 212 ⎠ = ⎣ 0 1 0 λη − ζ 0 0 ⎦δ⎜ ⎟
⎜ φ ⎟
213 0 0 1 λζ + η 0 0 ⎜ ⎟
⎜ ⎟
⎝ κy ⎠
κz
are the relations between the strain and the strain measures of the beam.
From Hamilton’s principle, the strain energy is the integral over the structure of the product of the stress and
strain:
$ $
T
δU = δ σ dΩ = (σ11 δ11 + σ22 δ22 + σ33 δ33 + 2σ12 δ12 + 2σ13 δ13 + 2σ23 δ23 ) dΩ
The stress is obtained from the strain by the constitutive law σij = Eijkl kl . Beam theory assumes that only the
stresses acting the plane perpendicular to the beam axis are important. So σ22 , σ33 , and σ23 are neglected, and the
constitutive law reduces to: ⎛ ⎞ ⎡ ⎤⎛ ⎞
σ11 Q11 Q15 Q16 11
⎝ σ12 ⎠ = ⎣ Q51 Q55 Q56 ⎦ ⎝ 212 ⎠
σ13 Q65 Q65 Q66 213
The strain energy can now be written in terms of the section loads:
$$ $
δU = T
δ σ dA dx = Fx δue + Fy 2δ12 + Fz 2δ13 + Mx δφ + My δκy + Mz δκz dx
0
The section loads are obtained from the stress, and hence from the section strain measures, as follows.
⎛ ⎞ ⎡ ⎤
Fx 1 0 0
⎜ ⎟ ⎢ ⎥
⎜ Fy ⎟ ⎢ θC ζ 1 0 ⎥⎛
⎜ ⎟ ⎢ ⎥ σ11 ⎞
⎜ ⎟ $ ⎢ ⎥
⎜ Fz ⎟ ⎢ −θC η 0 1 ⎥⎜
⎜ ⎟= dA ⎢ ⎥ ⎝ σ12 ⎟
⎠
⎜ ⎟ ⎢ 2 ⎥
⎜ Mx ⎟ ⎢ θC (ζλη − ηλζ ) + φ (η + ζ ) λη − ζ
2
λζ + η ⎥
⎜ ⎟ ⎢ ⎥ σ13
⎜ ⎟ ⎢ ⎥
⎝ My ⎠ ⎣ζ 0 0 ⎦
Mz −η 0 0
⎛ ⎞
ue
⎜ ⎟
⎜ 2 12 ⎟
⎜ ⎟
⎜ ⎟
⎜ 2 13 ⎟
⎜
=T⎜ ⎟
⎟
⎜ φ ⎟
⎜ ⎟
⎜ ⎟
⎝ κy ⎠
κz
At this point the effects of transverse shear are statically eliminated, reducing the 6 × 6 matrix T to the 4 × 4
matrix S. Generally it is appropriate to neglect the shear force (not the shear strain), so T is inverted, the Fy and
Fz rows and columns of T −1 are eliminated to produce S −1 , and the inverse of the resulting matrix gives S. At a
constrained end, assuming zero shear strain might be more appropriate; then S is simply obtained by eliminating
the Fy and Fz rows and columns of T . Including the nonlinear terms coupling extension and torsion, the section
loads are: ⎛ ⎞ ⎡ ⎤⎛ ⎞
Fx Suu Suφ + 1/2φ Suu kP2 Suw Suv ue
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎜ Mx ⎟ ⎢ Sφu + φ Suu kP2 Sφφ Sφw Sφv ⎥ ⎜ φ ⎟
⎜ ⎟=⎢ ⎥⎜ ⎟
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎝ My ⎠ ⎣ Swu Swφ Sww Swv ⎦ ⎝ κy ⎠
Mz Svu Svφ Svw Svv κz
Using the beam theory for anisotropic or composite materials including transverse shear deformation, the section
elastic constants S are the input quantities required by this component. Note that S is symmetric. With
⎛ ⎞ ⎛ ⎞ ⎡ ⎤⎛ ⎞
δue δqu CfT δqu
⎜ ⎟ ⎜ ⎟ ⎢ ⎥⎜ ⎟
⎜ δφ ⎟ ⎜ δq ⎟ ⎢ C T ⎥ ⎜ δqv ⎟
⎜ ⎟=C⎜ v ⎟=⎢ x ⎥⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎢ ⎥⎜ ⎟
⎝ δκy ⎠ ⎝ δqw ⎠ ⎣ CyT ⎦ ⎝ δqw ⎠
δκz δqφ CzT δqφ
the strain energy is

$
δU = Fx δue + Mx δφ + My δκy + Mz δκz dx = δquT Ku + δqvT Kv + δqw
T
Kw + δqφT Kφ = δq T K
0
Thus the structural terms for the equations of motion of the elastic degrees of freedom are:
⎡ ⎤
$ Ku

⎢K ⎥
K= Cf Fx + Cx Mx + Cy My + Cz Mz dx = ⎣ v ⎦
0 Kw
Kφ
The result is
$
Ku = hu Fx dx
0
$
Kv = hv (Sθ My + Cθ Mz ) dx
0
$
Kw = hw (−Cθ My + Sθ Mz ) dx
0
$

Kφ = hφ Mx + hφ (κz My − κy Mz ) dx
0
for the second-order model. The linear, small-angle result can be evaluated as required.
Finally, the strain energy is obtained using Euler-Bernoulli theory for a beam of isotropic materials with an
elastic axis. Transverse shear effects are neglected, and the elastic axis is the beam axis, which is straight within
the component. The constitutive law is now:
⎛ ⎞ ⎡ ⎤⎛ ⎞
σ11 E 0 0 11
⎝ σ12 ⎠ = ⎣ 0 G 0 ⎦ ⎝ 212 ⎠
σ13 0 0 G 213
Then the section loads are obtained from the stress, and hence from the section strain measures, as follows.
⎛ ⎞ ⎡ ⎤
Fx 1 0 0
⎜ ⎟ $ ⎢ ⎥
⎜ Mx ⎟ ⎢ θC (ζλη − ηλζ ) + φ (η 2 + ζ 2 ) λη − ζ λζ + η ⎥
⎜ ⎟= dA ⎢ ⎥
⎜ ⎟ ⎢ ⎥
⎝ My ⎠ ⎣ζ 0 0 ⎦
Mz −η 0 0
⎛ ⎞
⎡ ⎤ ue

E EθC (ζλη − ηλζ ) + 1/2Eφ (η 2 + ζ 2 ) Eζ −Eη ⎜ ⎟
⎢ ⎥⎜ φ ⎟
⎣0 G(λη − ζ) 0 0 ⎦⎜
⎜ ⎟
⎟
⎝ κy ⎠
0 G(λζ + η) 0 0
κz
⎡ ⎤⎛ ⎞
EA θC EA kT2 + 1/2φ EA kP2 EAzC −EAyC ue
⎢ ⎥⎜ ⎟
⎢ θC EA kT2 + φ EA kP2 2
GJ + θC EIpp
θC EIzp
−θC EIyp ⎥ ⎜ φ ⎟
=⎢
⎢
⎥⎜ ⎟
⎣ EAzC
θC EIzp / zz
EI / yz ⎥
−EI
⎜ ⎟
⎦ ⎝ κy ⎠
−EAyC
−θC EIyp / yz
−EI / yy
EI κz
The section integrals are here evaluated relative the elastic axis:
$
EIyp = Ey(zλy − yλz ) dA
$
EIzp = Ez(zλy − yλz ) dA
$
EIpp = E(zλy − yλz )2 dA
$ $
/ yy =
EI Ey 2 dA = E(y − yC )2 dA + EAyC
2 2
= EIyy + EAyC
$ $
/ zz =
EI Ez 2 dA = E(z − zC )2 dA + EAzC
2 2
= EIzz + EAzC
$ $
/ yz =
EI Eyz dA = E(y − yC )(z − zC ) dA + EAyC zC = EAyC zC
but are conventionally defined relative the tension center instead. The higher-order section integrals EIxp , EIzp ,
and EIpp are seldom available, and so are neglected for this model. The result for the section loads is thus:
⎛ ⎞ ⎡ ⎤⎛ ⎞
Fx EA θC EA kT2 + 1/2φ EA kP2 EAzC −EAyC ue
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎜ Mx ⎟ ⎢ θC EA kT + φ EA kP GJ
2 2
0 0 ⎥⎜ φ ⎟
⎜ ⎟=⎢ ⎥⎜ ⎟
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎝ My ⎠ ⎣ EAzC 0 EIzz + EAzC 2
−EAyC zC ⎦ ⎝ κy ⎠
Mz −EAyC 0 −EAyC zC 2
EIyy + EAyC κz
So using beam theory for isotropic materials with an elastic axis, the required section structural properties (modulus
weighted) are $
EAyC = Ey dA
$
EAzC = Ez dA
$
EAkP2 = E(y 2 + z 2 ) dA
$
EA = E dA
$
EAkT2 = E(zλy − yλz ) dA
$

GJ = G (λy − z)2 + (λz + y)2 dA
$
EIyy = E(y − yC )2 dA
$
EIzz = E(z − zC )2 dA
The integral is over the beam cross-section, in structural principal axes with origin at the beam axis. Note that the
section loads are approximately:
Fx ∼
= EA(ue + θC
2
φ kT − yC κz + zC κy )
Mx = (GJ + Fx kP2 )φ + Fx kT2 θC

My ∼
= EIzz κy + Fx zC
Mz ∼
= EIyy κz − Fx yC
This form is not used here, but it illustrates the value of using the tension center for the beam axis, since then the
bending moments My and Mz do not depend on the accuracy of the tension force Fx .
6-3.8 Extension/Torsion Coupling
In the above structural analysis, it was assumed that the constitutive relation is defined in local rectangular
Cartesian coordinates. The consequence of this assumption is a distinction between the torsion moments produced
by elastic torsion φ and by pretwist θC

, in the presence of a tension force:
Mx = (GJ + Fx kP2 )φ + Fx kT2 θC

(ref. 16). For a circular cross-section, kT must be zero, since pretwist is not then meaningful with isotropic
materials. If instead the constitutive law is applied in the curvilinear coordinates, the axial strain is
1
11 = (G11 − g11 ) ∼
= ue − κz η + κy ζ + (θC

φ + 1/2φ 2 )(η 2 + ζ 2 ) + W1
2
so the relations between the strain and the strain measures of the beam become:
⎛ ⎞
ue
⎜ ⎟
⎛ ⎞ ⎡ 1/2φ )(η 2 ⎤⎜⎜
2 12 ⎟
⎟
11 1 0 0 (θC + 2
+ζ ) ζ −η ⎜ ⎟
⎜ ⎟ ⎢ ⎥ ⎜ 2 13 ⎟
⎝ 212 ⎠ = ⎣ 0 1 0 λη − ζ 0 0 ⎦⎜⎜ ⎟
⎟
⎜ φ ⎟
213 0 0 1 λζ + η 0 0 ⎜ ⎜
⎟
⎟
⎝ κy ⎠
κz
The structural properties for an isotropic beam, including the extension/torsion coupling term, are then
$
EIyp = Ey(y 2 + z 2 ) dA
$
EIzp = Ez(y 2 + z 2 ) dA
$
EIpp = E(y 2 + z 2 )2 dA
$
EAkT2 = E(y 2 + z 2 ) dA = EAkP2
and the torsion moment is:

Mx = (GJ + Fx kP2 )φ + Fx kP2 θC

This result often has been obtained (as in ref. 17) by explicitly making the assumption that the axial stress is tangent
to the line described by constant η and ζ, which is a helix for a pretwisted beam. Such an assumption might be
appropriate for an anisotropic beam. In general, the extension/torsion coupling is defined by the section constant
kT for an isotropic beam, or Suφ = Sφu for an anisotropic beam.
6-3.9 Kinetic Energy
Using the d’Alembert approach, in which inertial acceleration is treated as a body force, the virtual work of
the inertial and gravitational forces is:
$$
δT = (δrI )T (r̈I − g I ) dm dx
The integration is over the section mass and then the beam length. The acceleration relative the inertial frame is
obtained from the position as follows:
rI = xBI/I + C IB rB
BI/B rB + C IB ṙB
ṙI = ẋBI/I + C IB ω

= C IB v BI/B + ω BI/B rB + ṙB

r̈I = C IB v̇ BI/B + ω̇ BI/B rB + 2 BI/B v BI/B + ω
ω BI/B ṙB + r̈B + ω BI/B ω
BI/B rB
The virtual displacement can be expressed in terms of the generalized coordinates:
δrI = δ(xBI/I + C IB rB )
= δxBI/I + δC IB rB + C IB δrB
= δxBI/I − C IB rB δψ BI/B + C IB δrB
⎛ ⎞
δxBI/B
⎜ δψ BI/B
⎟
⎜ ⎟
⎜ δq u ⎟
= C IB [ I −
rB RuT RvT T
Rw RφT ] ⎜ ⎟
⎜ δqv ⎟
⎝ ⎠
δqw
δqφ
= C IB RT δq
The motion of the B axes is the sum of the motion of B relative the component frame F, and the motion of F relative
the inertial frame I:
xBI/I = C IF xBF/F + xF I/I
C BI = C BF C F I
For a frame component, there is no BF motion, so virtual displacements of FI equal the virtual displacements of
BI. Otherwise the rigid degrees of freedom are the motion BF; but there are no virtual displacements of FI then,
so:
δxBI/I = C IF δxBF/F
δψ BI/B = δψ BF/B
Hence the rigid body equations of motion can be derived considering virtual displacements of BI in all cases. Then
the virtual work is
$$
δT = δq T RC BI (r̈I − g I ) dm dx
= δq T M
= (δxBI/B )T Mx + (δψ BI/B )T Mψ + δquT Mu + δqvT Mv + δqw
T
Mw + δqφT Mφ
with M the inertial terms in the beam equations of motion. By using virtual displacements δxBI/B and δψ BI/B
in body axes (instead of inertial axes), the rigid body equations are obtained in B axes. The displacement, velocity,
acceleration, and virtual displacement are required for a point on the beam, relative the component body axes:
rB = r0 + ry ηb + rz ζb
ṙB = ṙ0 + ṙy ηb + ṙz ζb
r̈B = r̈0 + r̈y ηb + r̈z ζb
δrB = δr0 + δry ηb + δrz ζb
The variables η and ζ identify the cross-section point, relative the section principal axes at θI . Relative section
axes that are bent but not twisted: % &
ηb Cθ −Sθ η
=
ζb Sθ Cθ ζ
with θ = θI + φ + Θ. Note that r0 is the position at the beam axis (η = 0 and ζ = 0). Also used are the position
at the center-of-gravity and the displacement from the beam axis to the center-of-gravity:
rI = r0 + ry ηIb + rz ζIb
ΔrI = rI − r0 = ry ηIb + rz ζIb
The virtual displacement of the beam can also be written:

⎛ ⎞
δqu
⎜ δq ⎟
δrB = [ RuT RvT T
Rw RφT ] ⎝ v ⎠
δqw
δqφ
so ⎡ ⎤
I
⎢ r ⎥
B
⎢ ⎥
⎢R ⎥
R = ⎢ u ⎥ = R0 + Ry ηb + Rz ζb
⎢ Rv ⎥
⎣ ⎦
Rw
Rφ
RI = R0 + Ry ηIb + Rz ζIb
where RI gives the virtual displacement at the center-of-gravity. Thus the inertial terms in the beam equations of
motion are evaluated from:
$$
M= RC BI (r̈I − g I ) dm dx
$$
= R v̇ BI/B + ω BI/B v BI/B − C BI g I

+ r̈B + ω̇ BI/B rB + 2 BI/B ω
ω BI/B ṙB + ω BI/B rB dm dx
$ %
= v − Cg)m
RI (v̇ + ω
0
0 + 2
+ RI (r̈0 + ω̇r ω
ω ṙ0 + ω r0 )m
I + 2
+ R0 (Δr̈I + ω̇Δr ω
ω ΔṙI + ω ΔrI )m
y + 2
+ Ry (r̈y + ω̇r ω
ω ṙy + ω ry )Iηη
z + 2
+ Rz (r̈z + ω̇r ω
ω ṙz + ω rz )Iζζ
z + 2
+ Ry (r̈z + ω̇r ω
ω ṙz + ω rz )Iηζ &

+ Rz (r̈y + ω̇ry + 2 ω
ω ṙy + ω ry )Iηζ dx
⎡ ⎤
Mx
⎢ Mψ ⎥
⎢ ⎥
⎢M ⎥
=⎢ u⎥
⎢ Mv ⎥
⎣ ⎦
Mw
Mφ
where
$
1
Iηη = ηb2 dm = (Iθ + IP C2θ ) − yI zI mS2θ
2
$
1
Iζζ = ζb2 dm = (Iθ − IP C2θ ) + yI zI mS2θ
2
$
1
Iηζ = ηb ζb dm = IP S2θ + yI zI mC2θ
2
with C2θ = cos 2θ = Cθ2 − Sθ2 and S2θ = sin 2θ = 2Sθ Cθ . The required section inertial properties (mass
weighted) are
$
m= dm
$
myI = y dm
$
mzI = z dm
$
Iθ = (y 2 + z 2 ) dm
$
IP = (y 2 − z 2 ) dm
The integral is over the beam cross-section, in inertial principal axes with origin at the beam axis. Note that for a
circular cross-section, IP = 0; while for a cross-section with small z dimension, IP ∼ = Iθ .
The motion of a point on the beam cross-section is required, relative the component body axes: rB , ṙB , r̈B ,
and δrB . The variables η and ζ identify the cross-section point, relative the section principal axes at θI . The
inertial effects of warp and transverse shear are neglected. Thus the position on the cross-section is:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 0 0
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
rB = xEB/B + C BE ⎝ η ⎠ = xEB/B + (C BE Xθ ) ⎝ ηb ⎠ = xEB/B + (Y−β Zζ )T ⎝ ηb ⎠
ζ ζb ζb
with θ = θI + φ + Θ. Note that

! !
η̇b −ζb
= θ̇
ζ̇b ηb
For the geometrically exact model, the required motion is evaluated as follows.
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
x+u −Sζ −Sβ Cζ
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
rB = ⎝ v ⎠ + ⎝ Cζ ⎠ ηb + ⎝ −Sβ Sζ ⎠ ζb
w 0 Cβ
⎛ ⎞ ⎡ ⎤⎛ ⎞ ⎡ ⎤⎛ ⎞
u̇ −Sβ Cζ 0 −Cζ θ̇ Sζ −Cβ Cζ Sβ Sζ θ̇
⎜ ⎟ ⎢ ⎥⎜ ⎟ ⎢ ⎥⎜ ⎟
ṙB = ⎝ v̇ ⎠ + ⎣ −Sβ Sζ 0 −Sζ ⎦ ⎝ β̇ ⎠ ηb + ⎣ −Cζ −Cβ Sζ −Sβ Cζ ⎦ ⎝ β̇ ⎠ ζb
ẇ Cβ 0 0 ζ̇ 0 −Sβ 0 ζ̇
⎛ ⎧⎡⎞ ⎤⎛ ⎞ ⎛ ⎞⎫
ü ⎪ −Sβ Cζ 0 −Cζ θ̈ Sζ (ζ̇ 2 + θ̇2 ) − 2Cβ Cζ β̇ θ̇ + 2Sβ Sζ ζ̇ θ̇ ⎪
⎨ ⎬
⎜ ⎟ ⎢ ⎥⎜ ⎟ ⎜ ⎟
r̈ = ⎝ v̈ ⎠ + ⎣ −Sβ Sζ 0 −Sζ ⎦ ⎝ β̈ ⎠ + ⎝ −Cζ (ζ̇ + θ̇ ) − 2Cβ Sζ β̇ θ̇ − 2Sβ Cζ ζ̇ θ̇ ⎠ ηb
B 2 2
⎪
⎩ ⎪
⎭
ẅ Cβ 0 0 ζ̈ −2Sβ β̇ θ̇
⎧⎡ ⎤⎛ ⎞ ⎛ ⎞⎫
⎪ Sζ −Cβ Cζ Sβ Sζ θ̈ Sβ Cζ (β̇ 2 + ζ̇ 2 + θ̇2 ) + 2Cβ Sζ β̇ ζ̇ + 2Cζ ζ̇ θ̇ ⎪
⎨ ⎬
⎢ ⎥⎜ ⎟ ⎜ ⎟
+ ⎣ −Cζ −Cβ Sζ −Sβ Cζ ⎦ ⎝ β̈ ⎠ + ⎝ Sβ Sζ (β̇ 2 + ζ̇ 2 + θ̇2 ) − 2Cβ Cζ β̇ ζ̇ + 2Sζ ζ̇ θ̇ ⎠ ζb
⎪
⎩ ⎪
⎭
0 −Sβ 0 ζ̈ −Cβ (β̇ 2 + θ̇2 )
⎛ ⎞ ⎡ ⎤⎛ ⎞ ⎡ ⎤⎛ ⎞
δu −Sβ Cζ 0 −Cζ δθ Sζ −Cβ Cζ S β Sζ δθ
⎜ ⎟ ⎢ ⎥⎜ ⎟ ⎢ ⎥⎜ ⎟
δrB = ⎝ δv ⎠ + ⎣ −Sβ Sζ 0 −Sζ ⎦ ⎝ δβ ⎠ ηb + ⎣ −Cζ −Cβ Sζ −Sβ Cζ ⎦ ⎝ δβ ⎠ ζb
δw Cβ 0 0 δζ 0 −Sβ 0 δζ
⎛ ⎞
δqu
⎜ ⎟
⎜ δqv ⎟
= [ RuT RvT T
Rw RφT ] ⎜
⎜
⎟
⎟
⎝ w⎠
δq
δqφ
With the second-order model, the position on the cross-section is approximated as follows:
⎛ ⎞ ⎡ ⎤⎛ ⎞ ⎛ ⎞
x+u 1 −v −w 0 x + u − v ηb − w ζb
⎜ ⎟ ⎢ ⎥⎜ ⎟ ⎜ ⎟
rB ∼
=⎝ v ⎠+⎣v 1 0 ⎦ ⎝ ηb ⎠ ∼
=⎝ v + ηb ⎠
w w 0 1 ζb w + ζb
It is consistent to use as well

= cos(θI + φ) ∼
Cθ ∼ = cos θI − φ sin θI
Sθ ∼
= sin(θI + φ) ∼
= sin θI + φ cos θI
so
! !
η̇b −ζb
= φ̇
ζ̇b ηb
Then the required motion is evaluated as follows:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
x+u −v −w
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
rB = ⎝ v ⎠ + ⎝ 1 ⎠ ηb + ⎝ 0 ⎠ ζb
w 0 1
⎛
⎞ ⎛ ⎞ ⎛ ⎞
u̇ −v̇ − w φ̇ −ẇ + v φ̇
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
ṙB = ⎝ v̇ ⎠ + ⎝ 0 ⎠ ηb + ⎝ −φ̇ ⎠ ζb
ẇ φ̇ 0
⎛
⎞ ⎛ ⎞ ⎛ ⎞
ü −v̈ − 2ẇ φ̇ − w φ̈ −ẅ + 2v̇ φ̇ + v φ̈
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
r̈B = ⎝ v̈ ⎠ + ⎝ 0 ⎠ ηb + ⎝ −φ̈ ⎠ ζb
ẅ φ̈ 0
and ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
δu −δv − w δφ −δw + v δφ
δrB = ⎝ δv ⎠ + ⎝ 0 ⎠ ηb + ⎝ −δφ ⎠ ζb
δw δφ 0
⎛ ⎞
δqu
⎜ δq ⎟
= [ RuT RvT RwT
RφT ] ⎝ v ⎠
δqw
δqφ
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
Ru hu 0 0 0 0 0 0 0 0
⎢ Rv ⎥ ⎢ 2Hvv qv hv 0 ⎥ ⎢ −hv 0 0 ⎥ ⎢ 0 0 0⎥
⎣ ⎦=⎣ ⎦+⎣ ⎦ ηb + ⎣ ⎦ζ
Rw 2Hww qw 0 hw 0 0 0 −hw 0 0 b
Rφ 0 0 0 −hφ w 0 hφ hφ v −hφ 0
is the virtual displacement of the beam.
The equations for the rigid body motion are obtained from the virtual work of the inertial and applied forces.
The linear and angular equations, in body axes, are:
Mx = F B
Mψ = M B
where F B and M B are the total force and moment acting on the body, about the origin of B, in B axes. These
equations of motion are written in terms of the velocities v BI/B and ω BI/B , which are then represented in terms
of linear degrees of freedom q̇ and angular degrees of freedom ṗ. So in order to maintain symmetry, the equations
must be multiplied by the transpose of the appropriate partial velocity matrices:
∂v BI/B
vq =
∂ q̇
for the linear equations, and

∂ω BI/B
ωp =
∂ ṗ
for the angular equations. These matrices are identified in the description of the conventions for representing rigid
motion, for the system response piece. The inertial terms in the linear and angular rigid body equations of motion
are as follows, for comparison with the rigid body component.
$$
Mx = C BI (r̈I − g I ) dm dx
BI/B v BI/B − C BI g I )
= M (v̇ BI/B + ω
BI/B + ω
+ (ω̇ BI/B ω
BI/B )M zCM
B
ω BI/B M żCM
+ 2 B B
+ M z̈CM
$$
Mψ = rB C BI (r̈I − g I ) dm dx
BI/B I B ω BI/B + 2I˙B ω BI/B + I¨B

= I B ω̇ BI/B + ω
B
+ (M zCM BI/B v BI/B − C BI g I )
)(v̇ BI/B + ω
B
where the component mass M , center of gravity location zCM , and moment of inertia matrix I B are:
$$ $
M= dm dx = m dx
0
$$ $
B B
M zCM = r dm dx = rIB m dx
0
$$ $
B
M żCM = ṙB dm dx = ṙIB m dx
0
$$ $
B
M z̈CM = r̈B dm dx = r̈IB m dx
$$ 0
I =−
B
rB rB dm dx
$$
I˙B = − rB
ṙB dm dx
$$
I¨B = − rB
r̈B dm dx
With no elastic motion, the center of gravity location and moment of inertia matrix are:
⎛ ⎞
$ x
B
M zCM = ⎝ yIb ⎠ m dx
0 zIb
⎡ ⎤
$ Iηη + Iζζ −xyIb m −xzIb m
⎢ ⎥
IB = ⎣ −xyIb m mx2 + Iζζ −Iηζ ⎦ dx
0
−xzIb m −Iηζ mx2 + Iηη
and their derivatives are zero.
The equations for the elastic motion are obtained from the virtual work of the structural, inertial, and applied
forces. The equations for the elastic generalized coordinates are:
⎛ ⎞
Mu + K u
⎜ Mv + K v ⎟
⎝ ⎠= vqT F E + ωqT M E
Mw + K w
Mφ + K φ
Structural damping is now introduced, by adding

⎛ ⎞
Cu q̇u
⎜ Cv q̇v ⎟
⎝ ⎠
Cw q̇w
Cφ q̇φ
√
to the left-hand-side of the equation. The diagonal damping matrices are evaluated as C = g M K from the
linear mass and spring matrices, where the structural damping constant g is twice the critical damping ratio. The
reactions F E and M E are evaluated from all interface loads acting on all the connections of a location (calculated
as described for the standard structural dynamic components). The summation is over all the locations of the
component. The partial velocities
⎛ ⎞
u̇
∂v EB/E ∂ ⎝ v̇ ⎠
vq = = C EB
∂ q̇ ∂ q̇
ẇ
⎛ ⎞
EB/E θ̇
∂ω ∂ ⎝ ⎠
ωq = =R β̇
∂ q̇ ∂ q̇
ζ̇
are obtained from the expressions for the linear and angular velocities in terms of the generalized coordinates.
6-3.12 Sensors
dynamic component: beam reaction by deflection method, and beam reaction by force balance method. The
section load can also be calculated at a node using the standard sensor for the structural dynamic load. The section
load consists of the section torsion and bending moments, the axial tension, and the section shear forces (the shear
forces are not available with the deflection method):
a) Torsion moment: Mx (positive for positive slope φ ).

b) Bending moments: My and Mz (positive for negative curvature w and positive curvature v respectively).
c) Axial tension force: Fx (positive for positive extension ue ).
d) Shear forces: Fy and Fz .
The sensor quantity is the section load at location Q. The load is the moment or force at the axial station xL of the
specified location. The load acts on the beam segment extending from 0 to xL , at the tension center, in structural
principal axes.
The deflection method obtains the section load from the elastic motion and structural coefficients. The
structural analysis provides expressions for the reactions at the beam axis, from which bending moments at the
tension center can be obtained:
MyT C = My − Fx zC
MzT C = Mz + Fx yC
Thus the section load from the deflection method is
⎛ ⎞ ⎡ ⎤⎛ ⎞
Fx Suu Suφ + 1/2φ Suu kP2 Suw Suv ue
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎜ Mx ⎟ ⎢ Sφu + φ Suu kP2 Sφφ Sφw Sφv ⎥⎜ φ ⎟
⎜ ⎟=⎢ ⎥⎜ ⎟
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎝ MyT C ⎠ ⎣ Swu − Suu zC Swφ − (Suφ + 1/2φ Suu kP2 )zC Sww − Suw zC Swv − Suv zC ⎦ ⎝ κy ⎠

MzT C Svu + Suu yC Svφ + (Suφ + 1/2φ Suu kP2 )yC Svw + Suw yC Svv + Suv yC κz
(anisotropic); or
⎛ ⎞ ⎡ ⎤⎛ ⎞
Fx EA θC EA kT2 + 1/2φ EA kP2 EAzC −EAyC ue
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎜ Mx ⎟ ⎢ θ C EA kT2 + φ EA kP2 GJ 0 0 ⎥⎜ φ ⎟
⎜ ⎟=⎢ ⎥⎜ ⎟
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎝ MyT C ⎠ ⎣ 0 −(θC EA kT2 + 1/2φ EA kP2 )zC EIzz 0 ⎦ ⎝ κy ⎠

MzT C 0 (θC EA kT2 + 1/2φ EA kP2 )yC 0 EIyy κz
(isotropic). Note that at a step in the stiffness there should be a corresponding step in the curvature or slope, so the
load remains continuous. With only a small number of shape functions it is not possible to simulate such a step
well, so the results for the reaction will not be accurate near a step in stiffness.
The force balance method obtains the section load from the difference between the applied forces and inertial
forces acting on the beam segment to one side of xL . The position of the tension center (or beam axis) at span
B
station xL (location Q), from the origin of the body frame, is rL . The orientation of the bent cross-section axes is
BE
given by C . The difference between the applied forces and the inertial forces, acting on the segment of beam
outboard of xL is:
" $ $ #
FL+ = C EB
C BS
F − S
C BI
(r̈ − g ) dm dx
I I
x>xL xL
" $ $ #

ML+ = C EB xCB/B − rL
C BS M S + ( B
)C BS F S − rB − rL
( B
)C BI (r̈I − g I ) dm dx
x>xL xL
The difference between the applied forces and the inertial forces, acting on the segment of beam inboard of xL is:
" $ $ #
xL
FL− = −C EB
C BS
F − S
C BI
(r̈ − g ) dm dx
I I
x<xL 0
" $ $ #
xL
ML− = −C EB
C BS S
M + (
x CB/B
− rL
B
)C BS F S − r −
B
( rL
B
)C BI (r̈I − g ) dm dx
I
x<xL 0
The summation of the applied forces is over all loads (with axes S) on all connections (at xCB/B ) outboard or
inboard of xL (the end of the beam at x = 0 is considered inboard of all xL ; the end of the beam at x = is
considered outboard of all xL ). The section loads calculated using the forces on either side of xL are combined,
so that this sensor gives at the beam ends the same result as the nodal reaction. Thus
FL = (xL / )FL+ + (1 − xL / )FL−

ML = (xL / )ML+ + (1 − xL / )FM −
or
% $$ &
FL = C EB
WC BS
F − S
WC BI
(r̈ − g ) dm dx
I I
% $$ &
ML = C EB CB/B C BS F S −
W C BS M S + x W rB C BI (r̈I − g I ) dm dx − C EB rL
B
FL
with the weighting function W 0

xL / x > xL
W =
xL / − 1 x ≤ xL
As a function of x given xL , W = xL / − 1 + u(xL ), where u(xL ) is the unit step function at x = xL . As a
function of xL given x, W = xL / − u(x), where u(x) is the unit step function at xL = x. The force balance
result for the section load can be written as follows:
! " # B !
FL C EB 0 F1 − M/x
=
ML −C EB rL
B
C EB /B − M
M /ψ
with FL = (Fx , Fy , Fz ) and ML = (Mx , My , Mz ). The vector on the right can be evaluated like the rigid body
equations of motion, with the addition of the weighting function W . However, Gaussian integration of the inertial
forces does not treat the step in W accurately, so there would be a jump in the load as the sensor location xL passes
a Gaussian integration point xi . The integrated inertial load is continuous with xL if the integrand at xL is handled
analytically. So the quadrature becomes
$ $ $
N

F = W f (x) dx = W f (x) − f (xL ) dx + W f (xL ) dx = wi W (xi ) f (xi ) − f (xL )
0 0 0 2 i=1
where xi = (ξi + 1)( /2), for Gaussian points ξi and weights wi . The force balance method can capture the steps
in the section load produced by discrete loads on the beam. For structural dynamic interfaces and applied load
interfaces, such steps are appropriate (although the shape functions are not consistent with discrete loads except
at the beam ends). However, the loads from aerodynamic interfaces are actually distributed forces. The step in
the section load associated with a discrete aerodynamic load at xF can be eliminated by assuming that the load is
uniformly distributed over xF − δ1 to xF + δ2 :
⎧
⎪ 0 if x < xF − δ1
⎪
⎪
⎪
⎨ 1/2δ1 if x < xF
D(x) =
⎪
⎪ 1/2δ2 if x < xF + δ2
⎪
⎪
⎩
0
where δ1 and δ2 are the minimum of half a specified range δ, or the distance to the end of the beam. Then the
weighting function for distributed aerodynamic forces is:
$ $ xL $ $ xL
WA = W D dx = (xL / − 1)D dx + (xL / )D dx = xL / − D dx
0 0 xL 0
⎧
⎪ 0 if xL < xF − δ1
⎪
⎪
⎪
⎨1− 1
2δ1 (xF − xL ) if xL < xF
= xL / − 2
⎪
⎪ 2 − − xL )
1 1
⎪
⎪ 2δ2 (xF if xL < xF + δ2
⎩
1
⎧
⎪ 1 if xF ≤ xL − δ2
⎪
⎪
⎪
⎨1− 1
2δ2 (xF − xL ) if xF ≤ xL
= xL / − 2
⎪
⎪ 2 − − xL ) if xF ≤ xL + δ1
1 1
⎪
⎪ 2δ1 (xF
⎩
0
So it is equivalent to use a ramp from xL − δ2 to xL + δ1 instead of a step in W as a function of x. The weighting

function WA is used for the aerodynamic forces in F1B and the aerodynamic moments in M /B . A distributed force
produces a moment on the beam:
$ $ $
W (x − xL )D dx = W xD dx − xL W D dx
0 0 0
which gives
$ $ xL $ $ $ xL
xF WF = W xD dx = (xL / − 1)xD dx + (xL / )xD dx = xL / xD dx − xD dx
0 0 xL 0 0
⎧
⎪
⎪ xF + δ2 −δ 1
if xF ≤ xL − δ2
⎪ 4

⎪
⎪
⎨ 1 − 1 (xF − xL ) 1 (xF + xL − δ2 ) + δ2 −δ1
δ2 − δ1 if xF ≤ xL
= xL / xF + − 2 2δ2
2 4
4 ⎪ 1
⎪
2 − 2δ1 (xF − xL ) 2 (xF + xL − δ1 ) if xF ≤ xL + δ1
1 1
⎪
⎪
⎪
⎩
0
/B .
The weighting function WF is used for the aerodynamic forces in M
The beam reaction can also be evaluated by using the standard sensor for structural dynamic load. This sensor
provides the total load acting at a location or a connection Q, about a reference R, in R axes. The nodal reactions
can thus be evaluated by identifying Q as the location or connection at the end of the beam, on which the structural
dynamic interfaces act. The reference R should be a connection C defined at the tension center (yC and zC , at
axial station x), in structural principal axes.
6-3.13 Cross-Section Structural Properties
To use the beam model for anisotropic or composite materials, it is necessary to obtain the cross-section
structural properties, including all terms coupling extension, bending, and torsion. The following two computer
programs can calculate the required elastic constants:
a) NABSA (Nonhomogeneous Anisotropic Beam Section Analysis), developed by Borri of

the Politecnica Di Milano and his colleagues (ref. 18).
b) VABS (Variational-Asymptotical Beam Sectional Analysis), developed by Hodges of Geor-

gia Institute of Technology and his colleagues (refs. 7, 8, and 19; which also provide compar-
isons of NABSA and VABS results).
Both programs are two-dimensional, finite-element based cross-section analyses for a general nonhomogeneous
and anisotropic beam section, including the effects of warp and initial twist. They obtain 6 × 6 matrices of elastic
constants, including the transverse shear terms, which can be reduced to the 4 × 4 matrices required by this
component.
6–4 References
1) Hodges, D.H.; Ormiston, R.A.; and Peters, D.A. “On the Nonlinear Deformation Geometry of Euler-Bernoulli
Beams.” NASA TP 1566, April 1980.
2) Hodges, D.H. “Nonlinear Equations for Dynamics of Pretwisted Beams Undergoing Small Strains and Large
Rotations.” NASA TP 2470, May 1985.
3) Hodges, D.H. “A Mixed Variational Formulation Based on Exact Intrinsic Equations for Dynamics of Moving
Beams.” Int. J. Solids Structures, Volume 26, Number 11, 1990.
4) Bauchau, O.A., and Hong, C.H. “Nonlinear Composite Beam Theory.” Journal of Applied Mechanics, Volume
55, Number 1, March 1988.
5) Smith, E.C., and Chopra, I. “Aeroelastic Response, Loads, and Stability of a Composite Rotor in Forward
Flight.” AIAA Journal, Volume 31, Number 7, July 1993.
6) Yuan, K.-A.; Friedmann, P.P.; and Venkatesan, C. “Aeroelastic Behavior of Composite Rotor Blades with Swept
Tips.” American Helicopter Society Forum, June 1992.
7) Hodges, D.H.; Atilgan, A.R.; Cesnik, C.E.S.; and Fulton, M.V. “On a Simplified Strain Energy Function for
Geometrically Nonlinear Behavior of Anisotropic Beams.” Composites Engineering, Volume 2, Numbers 5–7,
1992.
8) Cesnik, C.E.S., and Hodges, D.H. “VABS: A New Concept for Composite Rotor Blade Cross-Sectional
Modeling.” American Helicopter Society Forum, May 1995.
9) Smith, E.C., and Chopra, I. “Formulation and Evaluation of an Analytical Model for Composite Box-Beams.”
Journal of the American Helicopter Society, Volume 36, Number 3, July 1991.
10) Hodges, D.H., and Dowell, E.H. “Nonlinear Equations of Motion for the Elastic Bending and Torsion of
Twisted Nonuniform Rotor Blades.” NASA TN D-7818, December 1974.
11) Bathe, K.-J. Finite Element Procedures in Engineering Analysis. Prentice-Hall, Inc., Englewood Cliffs, New
Jersey, 1982.
12) Zienkiewicz, O.C. The Finite Element Method. McGraw-Hill Book Company, Limited, London, 1977.
13) Bisplinghoff, R.L.; Mar, J.W.; and Pian, T.H.H. Statics of Deformable Solids. Addison-Wesley Publishing
Company, Inc., Reading, Massachusetts, 1965.
14) Washizu, K. Variational Methods in Elasticity and Plasticity. Second Edition, Pergamon Press, New York,
1975.
15) Washizu, K. “Some Considerations on a Naturally Curved and Twisted Slender Beam.” Journal of Mathematics
and Physics, Volume 43, Number 2, June 1964.
16) Hodges, D.H. “Torsion of Pretwisted Beams Due to Axial Loading.” Journal of Applied Mechanics, Volume
47, Number 2, June 1980.
17) Houbolt, J.C., and Brooks, G.W. “Differential Equations of Motion for Combined Flapwise Bending, Chord-
wise Bending, and Torsion of Twisted Nonuniform Rotor Blades.” NACA Report 1346, 1958.
18) Giavotto, V.; Borri, M.; Mantegazza, P.; and Ghiringhelli, G. “Anisotropic Beam Theory and Applications.”
Computers and Structures, Volume 16, Number 1–4, 1983.
19) Cesnik, C.E.S., and Hodges, D.H. “Variational-Asymptotical Analysis of Initially Curved and Twisted Com-
posite Beams.” Applied Mechanics Review, Volume 46, Number 11, Part 2, November 1993.
Chapter 7
ROD/CABLE COMPONENT
7–1 Description
A rod/cable component is a structural dynamic component. It has the standard features, and the elastic motion
is described by a set of flexible rods or cables, each connecting two points. For this component, the rigid motion
describes the motion of central body, which contains all the mass. One end of a rod/cable is connected to a location
on the central body. The elastic degrees of freedom are the position of the other end of the rod/cable, measured
relative to the body location (hence relative to the rigid motion). The ends of the rod/cable are assumed to be
pinned, so it does not support moments. Thus a spring/damper/actuator is used to connect the rod/cable ends,
transmitting a force that is always directed between the two points. A cable is modelled by assuming that the
spring can not support a compression load. Bending dynamics of the cable are neglected.
A rod/cable does not support moments, so only pinned structural dynamic interfaces should be used at its
location. In general, no interface can produce a moment on a rod/cable. No joints can be used at a rod/cable
location. Because the spring/damper/actuator model is used in the rod/cable model, this component can have no
joints at all. The component-specific sensors can measure the following quantities: motion, length, reaction, power
(of rod/cable).
7–2 Theory
The rigid body equations of motion are the same as for the rigid body component. The elastic motion is
included however when evaluating the point of application of forces and moments on the body.
The elastic motion affects the position of a location relative to the body origin (EB). EB is the sum of elastic
degrees of freedom and a constant position (EB = EL + LB). The constant position of the location (the position
for zero elastic motion) relative the body axes is specified by z LB/B and C LB (L relative B, in B axes). Then the
elastic motion occurs, described by xEL/L and C EL (and their derivatives). Locations can be defined that are not
used by a rod/cable. At such a location, there is no elastic motion (EB = LB), and the position is constant relative
the body axes. A location used by a rod/cable has elastic motion. Only one (or no) rod/cable can be defined at each
location. A rod/cable is a flexible connection between the points L and E. Figure 1 illustrates the configuration.
86 ROD/CABLE COMPONENT
The elastic motion consists of the vector from L to E. Thus:
xEL/L =
C EL = I
and the sum of the elastic and constant motions is
xEB/B = C BL xEL/L + z LB/B

C EB = C EL C LB = C LB
Generally there is little reason to not use C LB = I.
The elastic degree-of-freedom vector consists of the vectors for all rod/cables. Hence with N rod/cables,
the length of the degree-of-freedom vector is 3N . For each rod/cable there are three elastic equations of motion,
obtained from equilibrium of loads:
fspring = F E
where is a unit vector in the direction of the rod/cable:
= /| |
This equation follows from the assumption that the rod/cable does not support moments: there is no moment M E
at E; and F E , the moment of F E about L, is always zero (since = /| | = 0). So the total reaction force at E
must be aligned with the rod/cable, in the direction of . Note that if there is only one rod/cable, the net force on
the body must be equal and opposite to F E , hence also aligned with the cable.
The reaction F E is evaluated from all interface loads acting on all connections of the rod/cable location
(calculated as described for standard structural dynamic components). The moment M E is zero by assumption.
The tension force (scalar) is obtained from the standard spring/damper/actuator model
as a function of the length of the rod/cable:
θ=| |
d
θ̇ = | | = T ˙
dt
The offset in this model is the nominal length of the rod/cable:
θ0 = 0
The bias c0 is the preload, the compression force when the length equals 0. So if c0 = 0, then 0 is the unloaded
length. A cable is modelled using the option for no compressive loads. For an offset actuator, the controlled
quantity is the unloaded length. For a force actuator, the controlled quantity is the preload. A rigid link can be
modelled using a large stiffness, which will produce | | → 0 . Requiring that | | = 0 exactly would give a static
constraint equation to be solved; this approach is not implemented. A rigid link can also be modelling using a
rigid body component.
ROD/CABLE COMPONENT 87
spring/
damper/ E
actuator
Figure 7-1 Rod/cable configuration.

88 ROD/CABLE COMPONENT
7-2.4 Sensors
dynamic component.
7-2.4.1 Rod/Cable Motion
The sensor quantity is the rod/cable motion : the position of location E, relative the central body at L, in E
axes. The location is that of rod/cable Q. Note that the E and L axes are parallel, since C EL = I. The following
motion can be obtained:
a) Displacement: = xEL/L
b) Velocity: ˙
c) Acceleration: ¨
d) Angles: cos−1 = cos−1 ( /| |)
The angles are the direction cosines of the vector , in E axes. The standard sensor for the elastic degrees of
freedom also gives access to this motion. This component-specific sensor identifies the quantity by rod/cable,
rather than by degree-of-freedom element.
7-2.4.2 Rod/Cable Length
The sensor quantity is the length | |, for rod/cable Q. The following motion can be obtained:
√
a) Displacement: | |= T
b) Velocity: d| |/dt = T ˙
7-2.4.3 Rod/Cable Reaction
The sensor quantity is the magnitude of the reaction acting on rod/cable Q. The tension force fspring is
evaluated from the elastic motion, using the spring/damper/actuator model.
7-2.4.4 Rod/Cable Power
The sensor quantity is the power of the reaction acting on rod/cable Q. The definition of the power (rate of
work performed by the forces) is
P = fspring θ̇
where θ̇ is the rod/cable length, and the tension force fspring is evaluated from the elastic motion.
Chapter 8
TRANSMISSION COMPONENT
8–1 Description
A transmission component models a subsystem that transmits rotational motion and torques. Figure 1
illustrates the configuration considered. There is a root and one or more branches (or no branches), connected
through a gear train. The transmission has rigid rotation θ0 . The root motion is rigid. The branch motion can
include elastic rotation θi , measured relative to the rigid rotation. Gear ratios relate the rotation of the root and
branch to the degrees of freedom. The root and each branch can have inertial, damping, and applied loads. The
transmission component is an approximation to the actual device.
A transmission component is not a true structural dynamic component, lacking rigid body motion or frames.
It shares some of the standard features of a structural dynamic component however, including:
a) mass, hence inertial (rotational only);

b) structural dynamic interfaces (torque only);
c) applied load interfaces (torque only);
d) controls and sensors (component-specific only).
but excluding joints, gravitational forces, and aerodynamic interfaces. The standard spring/damper/actuator model
is also used, for the elastic branches. If a drive train model (consisting of transmission and structural dynamic
components) has fewer masses than elastic degrees of freedom, the resulting mass matrix will be singular. The
solution procedure must be able to handle the static equations that are present in such a case. Alternatively,
additional masses can be included, to eliminate the singularity.
Figure 2 illustrates the functionality of the transmission component.
Degrees of Freedom
The component can have up to two degree-of-freedom vectors: rigid motion, and elastic motion. The rigid degree
of freedom is the motion of the root. The rigid motion vector is absent for prescribed or controlled rigid motion.
The elastic degrees of freedom are a subset of the branches.
Component Input
a) Structural dynamic interface: the input is torque.
b) Controls: for use by actuators, or applied loads.
90 TRANSMISSION COMPONENT
ROOT BRANCHES
g
sprin
Ri
Q0, I0, D0 R0
φ 0 = a θ0
0
a0
ai
Qi, Ii, Di
φ i = a (θ0+θi)
i
Figure 8-1 Configuration of transmission component.

TRANSMISSION COMPONENT 91
degrees of
controls freedom sensors
(for actuator, rigid
aux force) elastic
torque
rotation
structural dynamic
interface (torque)
Figure 8-2 Functionality of transmission component.

Component Output
a) Structural dynamic interface: the output is rotation.
b) Sensors.
The transmission component has a set of nodes at which there is rotational motion, connected through a gear
train. One node is identified as the root, and the remaining (if any) are the branches. At each node there is:
a) rotation φ;
b) torques produced by rotational inertia I and rotational damping D;
c) interface torque Q (structural dynamic interface or control torque).
The transmission has motion described by the variable θ0 that is rigid (without elastic restraint in the component).
The motion of the root node is identified as the rigid motion, except for a root gear ratio a0 :
φ0 = a0 θ0
The branch motion can include elastic rotation θi , measured relative to the rigid rotation. So with a branch gear
ratio ai , the motion of a branch node is:
φi = ai (θ0 + θi )
Each node (root or branch) is a location, and a connection, and possibly has an interface. The variable θ0 describing
the rigid motion is one of the following (analogous to the joint options):
a) Degree of freedom.
b) Prescribed: the motion can be constant velocity, rotating variable, or constant position.
c) Controlled: the motion can be obtained from a control vector, implementing a displacement
actuator. The control can be just the displacement (an approximation if the motion is time
varying), or include the velocity and acceleration.
The motion of a branch can be rigid, or elastic with degree of freedom θi . Each elastic degree of freedom has an
equation of motion obtained by equilibrium of loads on the branch:
fspring = finterfaces
where the reaction fspring is obtained from a spring/damper/actuator model, and finterfaces = Q is obtained from
the interface load acting on the branch.
8-3.2 Interfaces
The root and each branch can have an interface: structural dynamic interface, applied load interface, or none
(but not both). The interface provides a torque Q that acts on the root or branch node. The structural dynamic
interface is of torque kind. The component input is the torque Q, and the component output is the rotation φ. The
applied load interface implements a control torque. The torque is a gain a times the value f of some element of a
control vector (component input): Q = af .
8-3.3 Controls
A set of controls is available to the component. These vectors are component input, for connection to a system
input piece or to an input/output interface. The component can use the controls for the rigid motion, actuators, or
applied load interfaces. A particular control vector and element can be used once, more than once, or not at all.
8-3.4 Sensors
to a system output piece or to an input/output interface. The sensors can measure the following quantities:
a) motion (degrees of freedom, torque interface);

b) structural dynamic interface constraint (torque);
c) reaction (branch);
d) power (branch);
e) governor.
The value of any sensor can be multiplied by a scale factor. Automatic scaling produces as appropriate a value in
degrees or horsepower.
8–4 Theory
8-4.1 Rigid Motion
For prescribed rigid motion, the root variable can be specified as constant velocity or rotating variable:
θ0 = Ωt + ψ0
θ̇0 = Ω
or constant position:
θ0 = ψ0
Here the rotational speed Ω is a constant parameter for the constant velocity case, or obtained from a specified
period for the rotating variable case. The offset ψ0 is a constant input parameter. For controlled rigid motion, the
root variable is obtained from a control vector (component input):
⎛ ⎞ ⎛ ⎞
θ0 c + ψ0
⎝ θ̇0 ⎠ = ⎝ ċ ⎠
θ̈0 c̈
including a constant offset ψ0 . This implements a displacement actuator. The control vector must consist of the
displacement, velocity, and acceleration (length 3), in order to fully evaluate the transmission motion. If the control
vector provides only the displacement, then
⎛ ⎞ ⎛ ⎞
θ0 c + ψ0
⎝ θ̇0 ⎠ = ⎝ 0 ⎠
θ̈0 0
is used, which is an approximation if c is time-varying. A system input piece does not provide the derivatives of
the input variables. A component can be constructed to provide the derivatives for a time-varying control. Note
that a force actuator can be implemented using an applied load interface.
8-4.2 Rigid Equation of Motion
Equilibrium of torques on the root and each branch produces the following equations:
I0 φ̈0 + D0 φ̇0 = Q0 + R0
Ii φ̈i + Di φ̇i = Qi + Ri
where Q is the interface torque at the node, and R is the reaction at the gear train. It is assumed that there are no
power losses in the gear train (where the motion is entirely described by θ0 ):
P = Rj aj θ̇0 = 0
j=0
(in the summation, j equal zero means the root). Thus the reactions R must be satisfy the relation:
Rj aj = 0
j=0
Substituting for Rj gives the equation for rigid motion of the transmission:
aj Ij φ̈j + aj Dj φ̇j = aj Qj
j=0 j=0 j=0
or
a2j Ij θ̈0 + a2j Dj θ̇0 + a2j Ij θ̈j + a2j Dj θ̇j = aj Qj
j=0 j=0 j=1 j=1 j=0
For the elastic motion of the i-th branch, the reaction torque R = −f is provided by a spring/damper/actuator.
Thus the equation for elastic motion is
Ii φ̈i + Di φ̇i + fi = Qi
or
a2i Ii (θ̈i + θ̈0 ) + a2i Di (θ̇i + θ̇0 ) + ai fi = ai Qi
The elastic motion between the branch node and the gear train is Δφi = ai θi . The standard spring/damper/actuator
model of structural dynamic components is used. For each branch degree of freedom, a scalar reaction torque is
generated:
fi = fspring = f (Δφi , Δφ̇i , c)
where c is an optional control (an element of a specified control vector) for an actuator. The force is evaluated as
the sum of spring, damper, and bias terms. The spring is modelled by linear and elastomeric terms. The damper
is modelled by linear, elastomeric, and hydraulic terms. Optionally the device can support only compression, or
only tension. A force actuator or offset actuator can be implemented.
8-4.4 Sensors
The following sensor quantities can be defined as component output (none of the sensors of a standard
structural dynamic component is used).
8-4.4.1 Motion of Degrees of Freedom
The sensor quantity consists of the degrees of freedom of the vector Q. The degree of freedom Q can be
elastic or rigid (Q=EB). The displacement, velocity, or acceleration can be obtained.
8-4.4.2 Motion of Torque Interface
The sensor quantity is the rotation angle φ at a root or branch Q. The displacement, velocity, or acceleration
can be obtained.
8-4.4.3 Structural Dynamic Interface Torque
The sensor quantity is the constraint torque Q at a root or branch Q. It is also possible to define a system
output piece for the interface variable fl = Q.
8-4.4.4 Branch Reaction
The sensor quantity is the reaction torque on a branch Q (elastic only). The reaction is evaluated from the
elastic motion, using the spring/damper/actuator reaction fspring .
8-4.4.5 Branch Power
The sensor quantity is the power of a branch Q (elastic only). The definition of the power (rate of work
performed by the forces) is
P = fspring φ̇
where the spring/damper/actuator reaction fspring is evaluated from the elastic motion Δφ.
8-4.4.6 Governor
The sensor quantity is a rigid rotation error of the component motion. The rigid motion is measured relative
some reference motion:
ψe = θ0 − θref = θ0 − (Ωt + ψ0 )
ψ̇e = θ̇0 − θ̇ref = θ̇0 − Ω
where the reference frequency Ω is obtained from a specified period. Then
θg = KP ψ̇e + KI ψe
is the value of the sensor. Proportional (rotational speed error) and integral (rotation angle error) terms are included,
with constant gains.
rigid ξR elastic ξE torque control
rigid motion 2 2 yes yes
elastic motion 2 2 yes (a) yes
torque 0 0 (a) no no (b)
sensor:
motion or spring 2 2 no yes
reaction 2 2 yes yes
degree of freedom 2 2 no no (c)
governor 2 2 no no (b)
“torque” means a torque structural dynamic interface.

sensor = motion or spring: QUANT = 64–66, 68–69
sensor = reaction: QUANT = 67
sensor = degree of freedom: QUANT = 61–63
sensor = governor: QUANT = 70
Notes:
a) no if interface at root
b) yes if root is controlled variable
c) yes for rigid if root is controlled variable
Figure 8-3 Functionality of transmission component.

Chapter 9
REFERENCE FRAME COMPONENT
9–1 Description
A reference frame component provides the analysis with access to the motion of a specified frame. The
component can also be used to sum and transform vector quantities, typically loads or motion. While related to
structural dynamic components, the reference frame component does not model a physical entity.
Figure 1 illustrates the functionality of the reference frame component.
Component Input
a) Loads: force, moment, or force and moment (in F axes, about origin F).
b) Motion vectors (in F or P axes).
c) Rotation vector (from F to T axes).
d) Gust velocity (in inertial axes).
Component Output
a) Sensors.
9-3.1 Functions
The reference frame component provides information regarding a specified frame. It performs three primary
functions:
a) provide access to frame motion;

b) sum and transform load vectors;
c) sum and transform motion vectors.
The component frame is the subject reference frame F. A parent frame P can be specified. Axes T can also be
defined, by a rotation from frame F (TF can be constant or variable).
9-3.2 Sensors
98 REFERENCE FRAME COMPONENT
loads frame
motion sensors
rotation vector
gust velocity
Figure 9-1 Functionality of reference frame component.

REFERENCE FRAME COMPONENT 99
a) motion (motion of frame axes, velocity relative air);

b) reaction (total load);
c) vector (total motion);
d) unit vectors;
e) relation to ground.
g’s, degrees, or Mach number.
9–4 Theory
9-4.1 Definition of Axes T
Axes T can be defined by a constant or variable rotation from frame F. The T axes have the same origin as
the F frame. A constant rotation is defined by an input matrix C T F . A variable rotation is defined by a rotation
vector θ: ⎛ ⎞
θx
θ = ⎝ θy ⎠
θz
(component input). The elements of the rotation vector are interpreted in one of the following ways.
a) Aircraft Euler angles: C T F = Xθx Yθy Zθz

b) Rodrigues parameters: C T F = C(p), p = θ
c) Arbitrary Euler angles: C T F = Uθx Vθy Wθz , for rotations U V W = ±XY Z
d) Small angles: C T F = I − θ
9-4.2 Sensors
The following sensor quantities can be defined as component output: motion of frame axes; velocity relative
air; reaction; vector (reference frame to parent); vector (parent to reference frame); unit vectors; relation to ground.
9-4.2.1 Motion of Frame Axes
The sensor quantity is the axes motion at the origin of Q, relative S, in Q axes (except in S axes for displacement).
The Q axes can be the reference or rotated frame (Q=F or T). The S axes are a parent frame (S=P). A frame that
is not a parent of the component frame can not be handled (because of the frame functionality). The following
motion can be obtained:
a) Displacement: xF P/P
b) Velocity: C F P ẋF P/P = v F P/F
c) Acceleration: F P/F v F P/F
C F P ẍF P/P = v̇ F P/F + ω
d) Load factor: F P/F v F P/F − C F I g I
n = v̇ F P/F + ω
e) Angular velocity: ω F P/F
e) Angular acceleration: ω̇ F P/F
f) Rodrigues displacement: p from C F P
g) Rodrigues velocity: ṗ from ω F P/F
h) Rodrigues acceleration: p̈ from ω̇ F P/F

i) Rotational displacement: uψ from C F P
j) Euler angles: φ, θ, ψ from C QR
where FP is the reference frame motion. For Q=T, the motion is multiplied by C T F (the origin of F and T are the
same). The Rodrigues parameters are defined by the relation p = 2u tan ψ/2, for rotation by angle ψ about an
axis in the direction of the unit vector u. Hence u = p/|p| and ψ = 2 tan−1 |p|/2 give the rotational displacement
sensor:
p |p| tan−1 |p|/2
uψ = 2 tan−1 =p
|p| 2 |p|/2
The Euler angles are obtained from C QR as for the structural dynamic components.
9-4.2.2 Velocity Relative Air
The sensor quantity is the velocity relative the air at the origin of F, in Q axes. The axes Q can be the reference
or rotated frame (Q=F or T). For Q=T, the motion is multiplied by C T F . The following motion can be obtained:
air velocity and air angles.
a) Air velocity: the velocity of the frame origin relative to the air, in F axes
v F = vB
F
− vW
F
− vG
F
= v F I/F − C F I vW
I
− C F I vG
I
where FI is the reference frame motion. The air velocity is calculated as for the aerodynamic interface of a standard
structural dynamic component, but without any interference velocity terms. The gust velocity term is omitted if
the required component input is not defined.
b) Air angles: longitudinal velocity fraction, sideslip angle, and angle of attack
⎛ ⎞ ⎛ ⎞
u/V v1 /|v|
⎝ β ⎠ = ⎝ sin−1 v2 /|v| ⎠
α tan−1 v3 /v1
calculated from the components of the air velocity vF .
9-4.2.3 Reaction
The reference frame component can sum and transform force and moment vectors: from frame F to a parent
frame P; or to axes T. The sensor quantity is the total load (force and moment) about the origin of Q, in Q axes.
Q can be the rotated frame, the reference frame, or a parent frame (Q=T or S, S=F or P). The loads summed
(component input) are about the origin of F, in F axes. For the output in P axes:
F P = CP F FF
M P = CP F F P/P C P F
MF + x FF
where FP is the reference frame motion. For the output in T axes:
F T = CT F FF
M T = CT F MF
9-4.2.4 Vector, Reference Frame to Parent
The reference frame component can sum and transform motion vectors: from frame F to a parent frame P; or
to axes T. The summation operation can be be replaced by average, or weighted-summation. The sensor quantity
is the total motion in Q axes. Q can be the rotated frame, the reference frame, or a parent frame (Q=T or S, S=F
or P). The vectors summed (component input) are in F axes. For the output in P axes:
vP = C P F Gv F
vT = C T F Gv F
The gain factor G is 1 for summation; 1/N for average (N equal the number of vectors); or an input value for
weighted-summation.
9-4.2.5 Vector, Parent to Reference Frame
The reference frame component can sum and transform motion vectors: from a parent frame P to frame F; or
to axes T. The summation operation can be be replaced by average, or weighted-summation. The sensor quantity
is the total motion in Q axes. Q can be the reference frame or the rotated frame (Q=F or T). The vectors summed
(component input) are in S axes. S is a parent frame (S=P). For the output in F axes:
vF = C F P Gv P
vT = C T F C F P Gv P
The gain factor G is 1 for summation; 1/N for average (N equal the number of vectors); or an input value for
weighted-summation.
9-4.2.6 Unit Vectors
The sensor quantity is a unit vector of the Q frame, in S axes. Q can be the reference frame or the rotated
frame (Q=F or T). S is a parent frame (S=P). For unit vectors of the F frame:
P
i j k = CP F
F
while for unit vectors of the T frame: P

i j k = CP F CF T
T
where FP is the reference frame motion. These unit vectors can be evaluated as a function of time, and then
averaged by a filter component. The resulting averaged vectors will not have unit length. Thus any component
that uses this sensor should not assume that the vectors have unit length.
9-4.2.7 Relation to Ground
The sensor quantity is the ground normal or the height above the ground. The ground normal is a unit vector
in Q axes. Q can be the reference frame or the rotated frame (Q=F or T). Thus
kgF = C F I kgI
or
kgT = C T F C F I kgI
where FI is the reference frame motion. The height above ground is the distance the origin of the reference frame
is above ground level:
zg = hIAGL − (xF I/I )T kgI
where hIAGL is the constant height of the origin of the inertial frame above ground level.
frame β rotation forces motion gust

vector velocity
sensor:
motion (rigid or air) yes (a) yes (b) no no no (c)
unit vectors
relation to ground
sensor:
reaction (total load) 0 yes yes no no
sensor:
vector (total motion) 0 yes no yes no
Notes:
a) order depends on quantity
b) no if height above ground
c) yes if velocity relative air
Figure 9-2 Functionality of reference frame component.

Chapter 10
FILTER COMPONENT
10–1 Description
A filter component implements a filter operation on an input. It can function as a low-pass filter, as a resolver,
or as a differentiator. The operations available are:
a) mean
b) minimum/maximum: minimum, maximum, mid value, half peak-to-peak
c) square: mean square, rms, amplitude, cyclic product
d) harmonic: cosine, sine, magnitude, phase
e) derivative: velocity, acceleration
The filter component is frequently required to obtain the mean value of a quantity. Any component can apply an
implicit filter operation when it evaluates a component input (as described for the response system piece). It is
more flexible to use a separate filter component to build the system.
Figure 1 illustrates the functionality of the filter component.
Degrees of Freedom
The component can have one degree of freedom vector, if the differential equation model is used.
Component Input
a) Filter input.
b) Azimuth angle (harmonic operations only).
Component Output
a) Filter output.
The filter operation can be implemented by an implicit model, a differential equation model, or a direct
model. Different model kinds can be used for the trim, transient, and flutter tasks. The implicit model operates
on the response solution within a window. It can not be used for the flutter task. This model utilizes the standard
capability of the analysis to apply a filter when evaluating component input (as described for the response system
piece). The differential equation model uses a first-order lag to filter the input. It is only available for the mean,
106 FILTER COMPONENT
input degrees of
freedom
output
azimuth
Figure 10-1 Functionality of filter component.

FILTER COMPONENT 107
cosine, and sine operations. The direct model performs no operation on the input. It is available for cases where a
filter is not required or not appropriate for all three tasks.
The mean, cosine, and sine operations are linear; all other filter operations are nonlinear. Which model
options can be used depends on the operation kind and on the analysis task, as follows:
task linear operations nonlinear operations

trim all models implicit
transient all models implicit
flutter differential equation or direct direct
For nonlinear operations, the filter operation is not in fact being applied in the flutter task (and the results will
probably not be meaningful). If the model is not implicit, analytical matrices are available.
10–4 Theory
Let f be the component input, and x be the component output. For the differential equation model, the
equation of motion is:
τ ξ˙ + ξ = g
where ξ is the degree of freedom vector. The component output is evaluated from:
⎧
⎨ GF (f ) implicit
x = Gξ differential equation
⎩
Gg direct
where G is a constant gain value. For the implicit model, the standard implicit filter operation F (f ) is utilized
when evaluating the component input. The window over which the filter is applied is defined directly, or calculated
from a specified period. Time derivatives are evaluated by one of the standard finite difference options.
For a differential equation or direct model, the quantity g has the following definition:
a) Mean: g =f
b) Cosine: g = f 2 cos nψ
c) Sine: g = f 2 sin nψ
d) Other: g =f
For the cosine and sine operations, n is a specified harmonic, and ψ the azimuth angle. The azimuth angle can be
prescribed:
ψ = Ωt + ψ0
where the frequency Ω is obtained from a specified period. Alternatively, the azimuth angle can be a variable,
obtained from a component input θ:
ψ = θ + ψ0
(not for the implicit model). Thus for the cosine and sine operations, the differential equations are time-varying;
but still linear with a prescribed azimuth.
It would be possible to define a differential equation model for the nonlinear operations as well. Consider for
example:
mean square: g = f2 x = Gξ

f
rms: g= x = G ξ2 − ξ12
f2

f 2 cos nψ
magnitude: g= x=G ξ12 + ξ22
f 2 sin nψ
However these equations are nonlinear; more degrees of freedom are required for the rms and magnitude; and
the direct model is still not correct. Therefore a differential equation model is not implemented for the nonlinear
operations.
FILTER COMPONENT 109
ξ input azimuth
motion 1 yes yes
output 0 yes (a) no (b)
Notes:
a) no if differential equation model for trim, transient,
and flutter
b) yes if direct model for trim, transient, and flutter
Figure 10-2 Functionality of filter component.

Chapter 11
REFERENCE PLANE COMPONENT
11–1 Description
A reference plane component identifies the orientation of a plane. The orientation is defined by a small
rotation from some nominal plane. The nominal plane (x–y plane in T axes) has a fixed orientation relative the
component frame (F axes). The identification method can be one of the following:
a) plane through a set of position vectors;

b) harmonics of normal displacement of position vectors;
c) harmonics of sensor(s) that measure normal displacement.
The orientation is restricted to a small rotation from the nominal plane since large angles would require a nonlinear
identification method. The identified orientation can be filtered in this component, to produce the mean position.
The filter operation can be implemented by an implicit model, a differential equation model, or a direct model.
Figure 1 illustrates the functionality of the reference plane component.
Degrees of Freedom
The component can have one degree of freedom vector, if the differential equation model is used.
Component Input
a) Measurements of position (kind depends on identification method).
b) Measurements of rate.
Component Output
a) Rotation angles (relative T axes).
b) Rotation vector (relative F axes).
c) Rotation rates (relative T axes).
112 REFERENCE PLANE COMPONENT
degrees of
measurements rotation angles
freedom
position rotation vector
rate rotation rates
Figure 11-1 Functionality of reference plane component.

REFERENCE PLANE COMPONENT 113
11-3.1 Geometry
The orientation of the T axes relative the component frame F is defined by the matrix
⎡ ⎤
T1
⎢ ⎥
C T F = ⎣ T2 ⎦
T3
where is a unit vector of the T axes, relative the F frame. The reference plane orientation is defined by its normal
vector n. It is assumed that n is found by a small rotation from the z-axis in T: rotation by φx about the x-axis,
and rotation by φy about the y-axis. Thus
n = C3 ∼ 3
= (I + φ)
where
φx
φ = φx 1 + φy 3 = 1 2
φy
is the rotation vector. The component output consists of the rotation vector φ, which defines the orientation of the
reference plane relative the component frame F; and the rotation angles and rotation rates:

φx φ̇x
φy φ̇y
which define the orientation relative the nominal plane axes T. The rates are only available from the harmonic
identification methods.
11-3.2 Identification Method
The following methods can be used to identify the rotation angles, from an appropriate set of measurements:
plane through position, harmonics of normal displacement, or harmonics of sensors.
a) Plane Through Position: The measurements are a set of position vectors r (from
the origin of F, F axes). A plane is passed through these locations.
b) Harmonics of Normal Displacement: The measurements are a set of position

vectors r. It is assumed that the motion of r consists primarily of rotation about the
3 axis (which goes through the origin of the F frame). A plane is defined by the
harmonics of the normal displacement z = rT 3 . The rotation rates are obtained
from ż = ṙT 3 .
c) Harmonics of Sensors: The measurements are a set of scalar sensors z. It is

assumed that z measures the normal displacement from the reference plane. A
plane is defined by the harmonics of z. The rotation rates are obtained from ż.
The required measurements can be obtained from structural dynamic components. The harmonic methods assume
that a rotation about the z-axis of T generates the measurements. That rotation can be counter-clockwise (about
the positive 3 axis), or clockwise (about the negative 3 axis). In order to identify an instantaneous plane using the
differential equation or direct model, at least three measurements are required. The implicit model uses information
at past times to define the plane, so may not require three measurements.
11-3.3 Filter Model
The filter operation can be implemented by an implicit model, a differential equation model, or a direct model.
Different model kinds can be used for the trim, transient, and flutter tasks. The implicit model operates on the
response solution within a window. It can not be used for the flutter task. This model utilizes the standard capability
of the analysis to apply a filter when evaluating component input (as described for the response system piece). The
differential equation model uses a first-order lag to filter the input. The direct model performs no operation on the
input. It is available for cases where a filter is not required or not appropriate for all three tasks.
11–4 Theory
11-4.1 Component Equations
Let f be the component input, and x be the component output. For the differential equation model, the
equation of motion is:
τ ξ˙ + ξ = g
where ξ is the degree of freedom vector:

ξx
ξ=
ξy
The component output is evaluated from:
⎧
⎨ F (f ) implicit
φx
x1 = = ξ differential equation
φy ⎩
g direct
and

x2 = φ = 1 2 x1
where F (f ) is an implicit operation and g an explicit operation on the component input f . These operations
depend on the identification method. Similar equations are constructed for the rotation rate degrees of freedom.
11-4.2 Geometry
The reference plane is defined by the normal n, and one point r0 on the plane (from the origin of F, in F axes).
Then all vectors r to points on the reference plane must satisfy the relation
(r − r0 )T n = 0
Writing c = r0T n (constant) and substituting
n∼
= 3 − 2 φx + 1 φy
gives ⎛ ⎞
φx
z = rT 3 = c + rT 2 φx − rT 1 φy = rT 2 rT 1 1 ⎝ −φy ⎠
c
11-4.3 Plane Through Position Identification Method
The measurements are a set of position vectors r, that are assumed to lie on the reference plane, hence satisfy
the relation ⎛ ⎞
φx
rT 2 rT 1 1 ⎝ −φy ⎠ = rT 3
c
or x = Θy. Concatenating all the measurements (in space and/or time) gives the equations
X = ΘY
which has the least-squared-error solution
Θ = (X T X)−1 X T Y
The required matrices are ⎡ ⎤

rT 2
XT X = ⎣ rT 1 ⎦ rT 2 rT 1 1
1
⎡ T ⎤
2 (rrT )2 T2 (rrT )1 T2 r
= ⎣ T1 (rrT )2 T1 (rrT )1 T1 r ⎦
rT 2 rT 1 1
⎡ ⎤
rT 2
XT Y = ⎣ rT 1 ⎦ rT 3
1
⎡ T ⎤
2 (rrT )3
= ⎣ T1 (rrT )3 ⎦
rT 3
where the sum is over all measurement values. For the differential equation and direct models, the measurement
values are the set of positions at one time. The explicit operation g is the solution for φx and φy from Θ. For the
implicit model, the measurement values are the set of positions, at all times in the filter window. The sum over
time is accomplished by replacing r and rrT with their mean values, obtained using the standard implicit filter
option when evaluating the component input. The mean of rrT is obtained from the cyclic product:
⎛ ⎞
zx
mean ⎝ xy ⎠
yz
Then the remaining sum is over the set of positions. The implicit operation F is the solution for φx and φy from
Θ.
11-4.4 Harmonic Identification Methods
If the measurements are a set of position vectors r that lie on the reference plane, they satisfy the relation
z = rT 3 = c + rT 2 φx − rT 1 φy
It is assumed that the motion of r consists primarily of rotation about the 3 axis, so
rT 2 = R sin ψ
rT 1 = R cos ψ
for some constant gain R. The azimuth angle is prescribed: ψ = ±(Ωt + ψ0 ), where the frequency Ω is obtained
from a specified period. The minus sign is used for clockwise rotation (rotation about the negative 3 axis), which
changes the sign of the identified φx . Thus the relation
z = rT 3 = c + R(φx sin ψ − φy cos ψ)
must be satisfied by the measurements r. If the measurements are a set of scalar sensors z, it is assumed that they
satisfy the relation
z = c + R(φx sin ψ − φy cos ψ)
for some constant gain R. The time derivative is

ż = R (φ̇x ± Ωφy ) sin ψ − (φ̇y ∓ Ωφx ) cos ψ)
using ψ̇ = ±Ω; and ż = ṙT 3 if the measurements are the normal displacement.
For both harmonic identification methods then, the reference plane can be found from the harmonics of the
normal displacement z:
φx 1 zs /R
=
φy N −zc /R

φ̇x φy 1 żs /R
±Ω =
φ̇y −φx N −żc /R
where the sum is over all N measurement values. The gain R is specified for each input, so z/R is a measure of
φ. The reference azimuth ψ0 must also be specified for each input, so z/R measures φy at ψ = 0.
For the differential equation and direct models, the measurement values are the set of positions or sensors at
one time. Thus
φx 1 (2 sin ψ)z/R
=
φy N −(2 cos ψ)z/R

φ̇x φy 1 (2 sin ψ)ż/R
±Ω =
φ̇y −φx N −(2 cos ψ)ż/R
The explicit operation g is this solution for φx and φy . For the implicit model, the measurement values are the set
of positions or sensors, at all times in the filter window. Thus the standard implicit filter option for evaluating the
component input is used to obtain the harmonics zs and zc . The harmonics are averaged over the set of positions
or sensors. The implicit operation F is this solution for φx and φy .
ξ position rate
measurements measurements
motion 1 yes yes
rotation angles 0 yes (a) X
rotation vector 0 yes (a) X
rotation rates 0 yes (a) yes (a)
Notes:
a) no if differential equation model for trim, transient, and flutter
Figure 11-2 Functionality of reference plane component.

Chapter 12
DIFFERENTIAL EQUATION COMPONENT
12–1 Description
A differential equation component implements static, first order, or second order differential equations. The
component can also function to simply add quantities. The differential equation input is a combination of the
component input. The options are to concatenate, sum, average, or sum with gain the component input vectors.
The component output is a linear combination of the degrees of freedom and input. Optionally a nonlinear
transformation of the output variables can be implemented. Analytical matrices are available (if the output
transformation is not used).
Figure 1 illustrates the functionality of the differential equation component.
Degrees of Freedom
The component can have one degree of freedom vector, for the first order or second order differential equation
model.
Component Input
a) Control vectors.
Component Output
a) Output vector.
Let ξ be the degree of freedom vector, x the component output vector, and v the differential equation input
vector. The static equations have the form
x = Ds v
The input form can be static, first order, or second order. Optionally Ds = I can be used. The first order differential
equations have the form
ξ˙ = Aξ + Bv
x = Cξ + Dv
120 DIFFERENTIAL EQUATION COMPONENT
degrees of
controls freedom output
(input from control)
Figure 12-1 Functionality of differential equation component.

DIFFERENTIAL EQUATION COMPONENT 121
The input form can be first order or second order. The second order differential equations have the form
A2 ξ¨ + A1 ξ˙ + A0 ξ = B0 v
x = C2 ξ¨ + C1 ξ˙ + C0 ξ + D0 v
The input form must be second order. The differential equation is strictly proper (the number of poles is greater
than the number of zeros) if the output equation depends only on the degrees of freedom: D = 0 for first order
equations; or D0 = 0 for second order equations. An output transformation can be defined: x = F (xlin ), where
xlin is calculated from the linear equations.
12-3.2 Differential Equation Input
A set of control vectors f (component input) is available to the component. The differential equations input
v is calculated from f , by one of the following methods.
a) Identify by element: each element of v is an element of some vector f . This option can be used to concatenate
elements or vectors.

b) Sum: v = f , summed over N vectors.

c) Average: v = (1/N ) f , summed over N vectors.

d) Sum with gain (weighted sum): v = Gf , summed over N vectors, with gain G.
12-3.3 Static Equation
The static equation x = Ds v can be used to add the component input, in general with a matrix of weights. This
component implements scalar addition of quantities; the reference frame component is used for vector addition of
loads and motion. With Ds = I, the output x = v is simply the quantity v calculated from the control vectors f .
Alternatively, defining v as the concatenated control vectors, and Ds as the corresponding concatenated matrices,
gives
⎛ . ⎞
.
⎜ . ⎟
x = Ds v = · · · Di · · · ⎝ fi ⎠
..
.
which is a common application of this component.
12–4 Theory
12-4.1 Alternate Input Form
The static equations can be input in first order form. Without the time derivatives, the first order equations
are:
0 = Aξ + Bv
x = Cξ + Dv
Solving for ξ gives

x = C(−A−1 B) + D = Ds v
assuming that A can be inverted. The static equations can be input in second order form. Without the time
derivatives, the second order equations are:
A0 ξ = B0 v
x = C0 ξ + D0 v

x = C0 (A−1
0 B0 ) + D0 = Ds v
assuming that A0 can be inverted.
The first order equations can be input in second order form. Let ζ be the second order degrees of freedom:
A2 ζ̈ + A1 ζ̇ + A0 ζ = B0 v
x = C2 ζ̈ + C1 ζ̇ + C0 ζ + D0 v
The first order degrees of freedom (state vector) are

ζ̇
ξ=
ζ
Then the matrices of the first order differential equations are:

% &
−A−1
2 A1 −A−1
2 A0
A=
I 0
% −1 &
A2 B0
B=
0

C = C1 − C2 A−1
2 A1 C0 − C2 A−1
2 A0
D = D0 + C2 A−1
2 B0
12-4.2 Output Transformation
Optionally a nonlinear transformation of the output variables can be implemented: x = F (xlin ), where xlin
is calculated from the linear equations. The transformation is defined by a set of values F = Fn at xn , n = 1 to
N . Then F (x) is evaluated by linear interpolation:
xn − x x − xn−1
F = Fn−1 + Fn
xn − xn−1 xn − xn−1
where xn−1 < x ≤ xn . There is no extrapolation, so F = F1 for x < x1 and F = FN for x > xN .
DIFFERENTIAL EQUATION COMPONENT 123
ξ controls
motion yes (a) yes
output yes (b) yes (c)
Notes:
a) first order (1) if equations first order; second order (2) if equations second order
b) static (0) if equations first order; second order (2) if equations second order
c) no if strictly proper differential equation
Figure 12-2 Functionality of differential equation component.

Chapter 13
PROGRAMMABLE COMPONENT
13–1 Description
A programmable component implements user-defined calculations. The component equations can be static,
first order, or second order. The basic version of this component is equivalent to the differential equation component.
Other versions are constructed by modifying subroutines called by the component. The differential equation input
is a combination of the component input. The options are to concatenate, sum, average, or sum with gain the
component input vectors. One or more component output vectors are calculated from the degrees of freedom and
input.
Figure 1 illustrates the functionality of the programmable component.
Degrees of Freedom
The component can have one degree of freedom vector, for the first order or second order differential equation
model.
Component Input
a) Control vectors.
Component Output
a) Output vectors.
Let ξ be the degree of freedom vector, x a component output vector, and v the differential equation input
vector. A set of control vectors f is available to the component; v is obtained from f . The static equations have
the form
F = Ds v
x = UB (F, v, f, t)
where UB is a user-defined function. The basic version of the component gives UB = F . The input form of the
matrix can be static, first order, or second order. The first order equations have the following form, where UA and
126 PROGRAMMABLE COMPONENT
degrees of
controls freedom output
(input from control)
Figure 13-1 Functionality of programmable component.

PROGRAMMABLE COMPONENT 127
UB are user-defined functions:

E = Aξ + Bv − ξ˙
F = Cξ + Dv
˙ ξ, v, f, t)
0 = UA (E, ξ,
˙ ξ, v, f, t)
x = UB (F, ξ,
The basic version gives UA = E and UB = F . The input form of the matrices can be first order or second order.
The second order equations have the following form, where UA and UB are user-defined functions:
E = B0 v − (A2 ξ¨ + A1 ξ˙ + A0 ξ)
F = C2 ξ¨ + C1 ξ˙ + C0 ξ + D0 v
¨ ξ,
0 = UA (E, ξ, ˙ ξ, v, f, t)
¨ ξ,
x = UB (F, ξ, ˙ ξ, v, f, t)
The basic version gives UA = E and UB = F . The input form of the matrices form must be second order.
The differential equation is strictly proper (the number of poles is greater than the number of zeros) if the output
equation depends only on the degrees of freedom: D = 0 and x = UB (F, ξ, ˙ ξ, t) for first order equations; or
¨ ξ,
D0 = 0 and x = UB (F, ξ, ˙ ξ, t) for second order equations.
More than one output vector can be defined. The basic version of the component gives x = F for the first
output vector, and x = 0 for all others. Thus the matrices have the dimension of the first output vector. Optionally,
the matrices (and hence vectors E and F ) can be ignored. Then the equations of motion and the output equations
are entirely user-defined.
13-3.2 Differential Equation Input
A set of control vectors f (component input) is available to the component. The differential equations input
v is calculated from f , by one of the following methods.
a) Identify by element: each element of v is an element of some vector f . This option can be used to concatenate
elements or vectors.

b) Sum: v = f , summed over N vectors.

c) Average: v = (1/N ) f , summed over N vectors.

d) Sum with gain (weighted sum): v = Gf , summed over N vectors, with gain G.
Both v and f are available to the user-defined functions UA and UB .
13-3.3 User-Defined Calculations
An input parameter specifies the kind of user-defined calculations to be performed. Hence more than one
version can be implemented simultaneously. The basic version of the component is equivalent to the differential
equation component. User-defined calculations are implemented by modifying subroutines that are called by the
programmable component. The following subroutines are available to the user:
subroutine operation
UPGMRD read parameters
UPGMWT write parameters
UPGMTB tables required
UPGMIN initialize
UPGMEN element names
UPGMB0 function UB (static)
UPGMB1 function UB (first order)
UPGMB2 function UB (second order)
UPGMA1 function UA (first order)
UPGMA2 function UA (second order)
Subroutine names and common names beginning with the letter “U” are reserved for the user. The exception is
that common UNITCM is not available to the user.
An array of component parameters is available. Input and print of these parameters can be in user-defined
form, but the data must be stored in this array for proper handling by the analysis. The user-defined calculations
can use tables. An input parameter specifies the number of tables required (in addition to the table for equation
matrices). An array of saved quantities can be defined, for use during the component calculations. The analysis
subroutines have an argument that signals when the component is being executed in order to obtain equation
matrices (by numerical perturbation). Thus if necessary a special model can be implemented to produce the
appropriate linearized matrices of a nonlinear system. Utilities provide access to all component input vectors.
13–4 Theory
13-4.1 Transformation from Input Form
The matrix of the static equations can be input in first order form. Without the time derivatives:
0 = Aξ + Bv
x = Cξ + Dv

x = C(−A−1 B) + D = Ds v
assuming that A can be inverted. The matrix of the static equations can be input in second order form. Without
the time derivatives:
A0 ξ = B0 v
x = C0 ξ + D0 v

x = C0 (A−1
0 B0 ) + D0 = Ds v
The matrices of the first order equations can be input in second order form. Let ζ be the second order degrees
of freedom:
A2 ζ̈ + A1 ζ̇ + A0 ζ = B0 v
x = C2 ζ̈ + C1 ζ̇ + C0 ζ + D0 v
PROGRAMMABLE COMPONENT 129
The first order degrees of freedom (state vector) are

ζ̇
ξ=
ζ
Then the matrices of the first order differential equations are:

% &
−A−1
2 A1 −A−1
2 A0
A=
I 0
% −1 &
A2 B0
B=
0

C = C1 − C2 A−1
2 A1 C0 − C2 A−1
2 A0
D = D0 + C2 A−1
2 B0

ξ controls
motion yes (a) yes
output yes (b) yes (c)
Notes:
a) first order (1) if equations first order; second order (2) if equations second order
b) static (0) if equations first order; second order (2) if equations second order
c) no if strictly proper differential equation
Figure 13-2 Functionality of programmable component.

Chapter 14
TRANSFER FUNCTION COMPONENT
14–1 Description
A transfer function component implements first order differential equations that are defined in terms of a
single (scalar) transfer function H: w = Hv. The transfer function input is a combination of the component input.
The component output is a gain times the transfer function output. Analytical matrices are available.
Figure 1 illustrates the functionality of the transfer function component.
Degrees of Freedom
The component can have one degree of freedom vector. The number of degree of freedom elements equals the
number of poles.
Component Input
a) Control vectors.
Component Output
a) Output vector.
14-3.1 Transfer Function
The transfer function H is a product of poles, zeros, gains, and lags. A time lag is here modelled as a
combination of one pole and one zero. The poles must be unique. If Np is the number of poles and Nz the number
of zeros, then
Np ≥ Nz ≥ 0
is required. For Np = Nz = 0 the equations reduce to a static gain; it is simpler to use the differential equations
component for such cases.
14-3.2 Transfer Function Input and Output
The transfer function defines a relation between scalar input v, and scalar response w. A set of control vectors
f (component input) is available to the component. The transfer function input v (scalar) is calculated from f as
132 TRANSFER FUNCTION COMPONENT
degrees of
controls freedom output = gain * H * input
(input from sum of
elements with gain)
Figure 14-1 Functionality of transfer function component.

TRANSFER FUNCTION COMPONENT 133
a weighted sum of the elements of all the control vectors:
v= KvT f
The component output vector x is calculated from the transfer function output w (scalar) as:
x = Kx w
So each element of x equals a constant gain times w.
14–4 Theory
14-4.1 Transfer Function
The transfer function H is defined as a product of poles, zeros, gains, and lags:
2Nz
(s − zm )
H = K 2m=1 Np
n=1 (s − pn )
The poles and zeros, pn and zm , can be real or complex. The algorithm used to produce time domain equations
requires that the poles be unique. A time lag is approximated by a combination of one pole, one zero, and a gain:
1 − τ s/2 s − 2/τ
e−τ s ∼
= =−
1 + τ s/2 s + 2/τ
The definition of H can include more than one gain factor, and more than one time lag. In terms of the residues,
the transfer function can be written:
Np
an
H = a0 +
n=1
(s − pn )
where 2
(pi − zm )
ai = lim (s − pi )H = K 2
n=i (pi − pn )
s→pi
a0 = lim H = K
s→∞
(a0 = 0 if Nz < Np ).
14-4.2 Differential Equation
A set of time domain differential equations equivalent to the transfer function must be constructed. The
component has a degree of freedom vector ξ, with each element ξn corresponding to a pole. Define the equations
of motion as
1
ξn = v
s − pn
Then w = Hv with H in the form
Np
an
H = a0 +
n=1
(s − pn )
gives
Np
w = a0 v + an ξn
n=1
The equivalent differential equations are:
⎡ ⎤
1
.
(s − Λ)ξ = .. ⎦ v
⎣
1

w = · · · an · · · ξ + a0 v
where Λ is the diagonal matrix of the poles pn . For a pair of complex poles, equations with real coefficients are
obtained from the real and imaginary parts of the above equations:
% &
s − pR pI ξR 1
= v
−pI s − pR ξI 0

ξR
Δw = 2aR −2aI
ξI
Thus the transfer function can be transformed to a first order differential equation:
ξ˙ = Aξ + Bv
w = Cξ + Dv
for scalar v and w, and degree of freedom vector ξ (length Np ). The matrices
⎡ ⎤
..
⎢ . ⎥
⎢ pn ⎥
⎢ % & ⎥
A=⎢
⎢ pR −pI ⎥
⎥
⎢ pI pR ⎥
⎣ ⎦
..
.
⎡ . ⎤
.
⎢ . ⎥
⎢ 1 ⎥
⎢% &⎥
B=⎢⎢ 1 ⎥
⎥
⎢ 0 ⎥
⎣ ⎦
..
.

C = · · · an [ 2aR −2aI ] · · ·
D = a0
are evaluated from the poles and residues of the transfer function. The relations
v= KvT f
x = Kx w
complete the equations of the component.

TRANSFER FUNCTION COMPONENT 135
ξ controls
motion 1 yes
output 0 yes if a0 = 0
Figure 14-2 Functionality of transfer function component.

Chapter 15
FOURIER SERIES COMPONENT
15–1 Description
A Fourier series component generates a time history from harmonics. The output is a scalar quantity, but can
include time derivatives. The azimuth angle can be a variable or prescribed.
Figure 1 illustrates the functionality of the Fourier series component.
Component Input
a) Harmonics vector.
b) Azimuth angle.
Component Output
a) Time history vector.
The component evaluates a quantity c in the time domain, from harmonics:

" M
#
c = G f0 + (fnc cos nψ + fns sin nψ)
n=1
where G is a constant gain value. The component input f is a vector of the harmonics:
f T = ( f0 · · · fnc fns ···)
The derivatives of f may also be available, as separate component input vectors:
f˙T = ( f˙0 · · · f˙nc f˙ns ···)

¨T
f = ( f¨0 · · · f¨nc f¨ns ···)
The component input vectors contains M harmonics. Optionally c can be calculated using a subset of the harmonics,
over the range n = NB to NE . The component output x consists of the scalar c, and possibly its derivatives:
xT = ( c )
or
xT = ( c ċ )
138 FOURIER SERIES COMPONENT
harmonics
output
azimuth
Figure 15-1 Functionality of Fourier series component.

FOURIER SERIES COMPONENT 139
or
xT = ( c ċ c̈ )
In calculating the derivatives of c, the derivatives of f are assumed to be zero if they are not available. The azimuth
angle ψ can be a variable or prescribed. For a prescribed azimuth:
ψ = Ωt + ψ0
ψ̇ = Ω
where the frequency Ω is obtained from a specified period. Alternatively, the azimuth angle can be a variable,
obtained from a component input θ:
ψ = θ + ψ0
ψ̇ = θ̇
ψ̈ = θ̈
The component input vector for the variable azimuth may contain just the displacement θ, or it may contain the
derivatives as well. In calculating the derivatives of c, the derivatives of ψ are required (the derivatives are assumed
to be zero if they are not available).
15–4 Theory
The component evaluates a quantity c in the time domain, from harmonics:

c = G f0 + (fnc cos nψ + fns sin nψ)
Then

ċ = G ( − fnc sin nψ + fns cos nψ) nψ̇ + G f˙nc cos nψ + f˙ns sin nψ
c̈ = G ( − fnc cos nψ − fns sin nψ) (nψ̇)2 + G ( − fnc sin nψ + fns cos nψ) nψ̈

+G − f˙nc sin nψ + f˙ns cos nψ 2nψ̇ + G f¨nc cos nψ + f¨ns sin nψ
are the derivatives of c.

140 FOURIER SERIES COMPONENT
harmonics harmonics harmonics azimuth

displacement velocity acceleration
output yes yes (a) yes (b) yes
Notes:
a) no if output velocity and acceleration not required
b) no if output acceleration not required
Figure 15-2 Functionality of Fourier series component.

Chapter 16
PRESCRIBED CONTROL COMPONENT
16–1 Description
A prescribed control component generates a transient time history. The output is a scalar quantity, but can
include time derivatives. The following functions are available to produce the time history: step; ramp; cosine
impulse; sine doublet; square impulse; square doublet; triangular impulse; triangular doublet; general piecewise
linear; random. The random time history can be generated with a uniform distribution, a normal distribution, or
filtered.
Figure 1 illustrates the functionality of the prescribed control component.
Component Output
a) Time history vector.
The component evaluates a quantity c from a prescribed function of time: c = GF (t − tB ), where G is a

constant gain value, and tB the time at which the transient begins. The function F (τ ) is the selected transient
shape. The displacement c can be also constructed from the derivative or integral of the function F . Hence the
options are
⎧
⎪ GF (t − tB )
⎪
⎪
⎪
⎪ GF (t − tB )
⎪
⎨
c = GF (t − tB )
⎪
⎪ 3
⎪
⎪ G F (t − tB )
⎪
⎪
⎩ 33
G F (t − tB )
The component output vector x can consist of the scalar c, and possibly its derivatives:
⎧
⎨(c)
xT = ( c ċ )
⎩
( c ċ c̈ )
Alternatively, the component output can be a vector of displacements:
x = Sc
142 PRESCRIBED CONTROL COMPONENT
output
Figure 16-1 Functionality of prescribed control component.

PRESCRIBED CONTROL COMPONENT 143
where S is a constant vector of scale factors (so all elements of x have the same shape in time). The derivatives
and integrals are not available for a random time history. The following functions F (τ ) are available to produce
the time history. In all cases, F = 0 for τ < 0.
a) Step: F =1
b) Ramp: F =τ
c) Cosine impulse: F = 1/2(1 − cos 2πτ /T ) for 0 < τ ≤ nT
d) Sine doublet: F = sin 2πτ /T for 0 < τ ≤ nT
e) Square impulse: F =1 for 0 < τ ≤ T
f) Square doublet: F =1 for 0 < τ ≤ T /2
F = −1 for T /2 < τ ≤ T
g) Triangular impulse: F = 2τ /T for 0 < τ ≤ T /2
F = 2 − 2τ /T for T /2 < τ ≤ T
h) Triangular doublet: F = 4τ /T for 0 < τ ≤ T /4
F = 2 − 4τ /T for T /4 < τ ≤ 3T /4
F = 4τ /T − 4 for 3T /4 < τ ≤ T
i) General piecewise linear: F = Fi at τi , i = 1 to I
F0 = 0 at τ0 = 0, and F = FI for τ > τI
j) Random: F =R
For a step occurring at t0 , the convention is that the value just before t0 is used at t0 .
16–4 Theory
16-4.1 Derivatives and Integrals
The component evaluates a quantity c from a prescribed function of time F (τ ). For the available shape
functions, the derivatives and integrals are as follows. In all cases, the quantities are zero for τ < 0.
a) Step:
F =1
F = δ(τ )
F = δ (τ )
3
F =τ
33
F = 12 τ 2
b) Ramp:
F =τ
F = 1
F = δ(τ )
3
F = 12 τ 2
33
F = 16 τ 3
c) Cosine impulse (ω = 2π/T ):

for 0 < τ ≤ nT for τ > nT
F = 1
2 (1 − cos ωτ ) F =0
F = 12 ω sin ωτ F = 0
F = 12 ω 2 cos ωτ F = 0
3 3
F = 12 (τ − ω1 sin ωτ ) F = 12 nT
33 33
F = 12 12 τ 2 − ω12 (1 − cos ωτ ) F = 14 (nT )2 + 12 nT (τ − nT )
d) Sine doublet (ω = 2π/T ):

for 0 < τ ≤ nT for τ > nT
F = sin ωτ F =0
F = ω cos ωτ F = 0
F = −ω 2 sin ωτ + ωδ(τ ) − ωδ(τ − nT ) F = 0
3 3
F = ω1 (1 − cos ωτ ) F =0
33 33
F = ω1 (τ − ω1 sin ωτ ) F = ω1 nT
e) Square impulse:
for 0 < τ ≤ T for τ > T
F =1 F =0

F = δ(τ ) − δ(τ − T ) F = 0
F = δ (τ ) − δ (τ − T ) F = 0
3 3
F =τ F =T
33 33
F = 12 τ 2 F = 12 T 2 + T (τ − T )
f) Square doublet:
for 0 < τ ≤ T /2 for T /2 < τ ≤ T
F =1 F = −1

F = δ(τ ) − 2δ(τ − T /2) + δ(τ − T ) F = δ(τ ) − 2δ(τ − T /2) + δ(τ − T )
F = δ (τ ) − 2δ (τ − T /2) + δ (τ − T ) F = δ (τ ) − 2δ (τ − T /2) + δ (τ − T )
3 3
F =τ F =T −τ
33 33
F = 12 τ 2 F = − 14 T 2 + T τ − 12 τ 2
for τ > T
F =0
F = 0
F = 0
3
F =0
33
F = 14 T 2
g) Triangular impulse:
for 0 < τ ≤ T /2 for T /2 < τ ≤ T
F = 2
T τ F = (2 − 2
T τ)

F = 2
T F = − T2
F = T2 δ(τ ) − 4
T δ(τ − T /2) + 2
T δ(τ − T ) F = T2 δ(τ ) − 4T δ(τ − T /2) + T2 δ(τ − T )
3 3
F = T1 τ 2 F = − 12 T + 2τ − T1 τ 2
33 33
F = 3T 1 3
τ F = 12 T − 12 T τ + τ 2 − 3T
1 2 1 3
τ
for τ > T
F =0
F = 0
F = 0
3
F = 12 T
33
F = − 14 T 2 + 12 T τ
h) Triangular doublet:
for 0 < τ ≤ T /4 for T /4 < τ ≤ 3T /4
F = 4
T τ F =2− 4
T τ

F = 4
T F = − T4
F = 4
T δ(τ ) − 8
T δ(τ − T /4) F = 4
T δ(τ ) − 8
T δ(τ − T /4)
+ T8 δ(τ − 3T /4) − 4
T δ(τ − T ) + T8 δ(τ − 3T /4) − 4
T δ(τ − T )
3 3
F = T2 τ 2 F = − 14 T + 2τ − T2 τ 2
33 33
2 3
F = 3T τ F = 48 T − 14 T τ + τ 2 −
1 2 2
3T τ3
for 3T /4 < τ ≤ T for τ > T
F = 4
T τ −4 F =0
F = 4
T F = 0
F = 4
T δ(τ ) − 8
T δ(τ − T /4) F = 0
+ T8 δ(τ − 3T /4) − 4
T δ(τ − T )
3 3
F = 2T − 4τ + τ 2
T
2
F =0
33 33
F = − 13
24 T + 2T τ − 2τ +
2 2 2
3T τ3 F = 18 T 2
Here δ(t − t0 ) is an impulse at t = t0 , and δ (t − t0 ) is a doublet at t = t0 . Derivatives and integrals are not
available for a random time history. Higher derivatives are nonzero only for the cosine impulse and sine doublet.
The impulse and doublet can not be used directly in the analysis, since the component output x must be finite
and evaluated at times specified by the solution procedures. The impulses and doublets can be ignored, or the
system can be constructed such that derivatives requiring impulses or doublets are not required. The component
can also implement an approximation for the impulse and doublet. The impulse δ(t) is modelled as a triangular
impulse Δ(t) of width 2Δt:
a) Impulse: Δ = (Δt + t)/Δt2 for −Δt < t < 0

Δ = (Δt − t)/Δt 2
for 0 < t < Δt
b) Doublet: Δ = 1/Δt2 for −Δt < t < 0

Δ = −1/Δt2 for 0 < t < Δt
If necessary, Δ is shifted in time so that F = 0 for τ < 0 is maintained. Thus the following approximations can
be used:
σ
δ(t − t0 ) ∼= Δ(t − t1 ) = (if σ > 0)
Δt2
and
sign(t − t1 )
δ (t − t0 ) ∼
= Δ (t − t1 ) = − (if σ > 0 and t − t1 = 0)
Δt2
where t1 = max(t0 , Δt) and σ = Δt − |t|.
16-4.2 General Time History
The general time history is a piecewise linear function, defined by the points F = Fi at τi , for i = 1 to I; with
the convention that F0 = 0 at τ0 = 0 (i = 0), and F = FI for τ > τI . Optionally the specification of the general
time history (τi , Fi ) can be obtained from a table. The general time history is evaluated by linear interpolation:
τi − τ τ − τi−1
F = Fi−1 + Fi
τi − τi−1 τi − τi−1
where τi−1 < τ ≤ τi . Let Δi = τi − τi−1 and u = (τ − τi−1 )/Δi . Then for τi−1 < τ ≤ τi (0 < u ≤ 1):
F = uFi + (1 − u)Fi−1
1
F = Fi − Fi−1
Δi
% &
1 1
F = Fi+1 − Fi − Fi − Fi−1 δ(τ − τi )
Δi+1 Δi
% 2 &
3 u u2
F = Gi−1 + Δi Fi + u − Fi−1
2 2
% 3 u2 &
33 2 u u3
F = Hi−1 + Δi Gi−1 u + Δi Fi + − Fi−1
6 2 6
and for τ > τI
F = FI
F = 0
F = 0
3
F = GI + (τ − τI )FI
33
F = HI + (τ − τI )GI + 12 (τ − τI )2 FI
where
1
i
Gi = Δj Fj + Fj−1
j−1
2
1
i
Hi = Δj Gj−1 + Δ2j Fj + 2Fj−1
j−1
6
16-4.3 Random Time History
The random time history can be generated with a uniform distribution, a normal distribution, or filtered. The
Park and Miller minimal standard method is used to generate a random series RU (τ ) with amplitude uniformly
distributed over the range −1 to 1. This method uses a linear congruential generator of a random number uniformly
distributed from 0 to 1:
si+1 = (75 si ) mod (231 − 1)
implemented in a form that avoids integer overflow. The seed s scaled to the range 0 to 1 is a random number X,

and then RU = 2X − 1. The standard deviation of the component output c equals 1/3 = 0.57735 times the
gain G. The series is generated at the time steps of the transient part solution procedure, beginning with an initial
seed value. Thus the same series is generated whenever the part solution is executed.
The Box-Muller method is used to generate a random series RN (τ ) with a normal (Gaussian) amplitude
distribution of zero mean and unit standard deviation. The standard deviation of the component output c equals
the gain G. This method calculates two values of the random number RN :

RN 1 = X1 (−2 ln z)/z

RN 2 = X2 (−2 ln z)/z
where z 2 = X12 + X22 , and X1 and X2 are two random numbers with uniform distribution over the range −1 to
1, generated as above.
A Markov process is used to generate a random series RM with a normal (Gaussian) amplitude distribution
of zero mean and unit standard deviation, and an exponential autocorrelation function. The standard deviation of
the component output c equals the gain G. A continuous process is obtained from a low-pass filter of the Gaussian
white noise W :
√
Ṙ + ωR = 2ω σ W
where ω is the filter frequency (time constant 1/ω), and σ is the rms value of R. Thus the spectrum of R is:
σ2 ω 1
S(Ω) =
π Ω + ω2
2
The corresponding discrete process is obtained from a difference equation:

σ
Rk+1 = e−ω Δt Rk + 1 − e−ω Δt Wk
ω Δt/2
The discrete input Wk has a normal distribution of zero mean and unit standard deviation, generated as above. The
rms value of the discrete series is
2 tanh ω Δt/2 2
σR = σ
ω Δt/2
Since σR = 1 here, the difference equation becomes
1
Rk+1 = e−ω Δt Rk + 1 − e−ω Δt Wk
tanh ω Δt/2

= e−ω Δt Rk + (1 − e−ω Δt ) (1 + e−ω Δt ) Wk
The series is generated at the time steps of the transient part solution procedure. The difference equation gives an
accurate simulation of the continuous filter if Δt is small enough (ωΔt ≤ 1 or so). A maximum value of ωΔt is
specified, and then Δt is chosen to meet this criterion if the time step of the transient part is too large.
Chapter 17
GUST COMPONENT
17–1 Description
A gust component generates an aerodynamic gust velocity. The gust velocity is calculated at collocation
points on the system. In general, the gust velocity is a function of time, and of position in space. The position is
the collocation point location relative some origin, perhaps convected by the wind. The following models for the
spatial dependence are implemented:
a) elementary (uniform, angular, or quadratic);

b) prescribed (function of one position coordinate);
c) tabular (function of two, three, or four of the four space/time coordinates).
Each model includes a gust amplitude factor obtained from component input. Hence the possible sources for the
time dependence of the gust velocity are the amplitude factor, a moving collocation point, convection, and the table
data. For an aircraft it is conventional to apply Taylor’s hypothesis: that the time dependence of the gust velocity
is produced by motion of the system relative a gust field which varies in space but not time (a frozen field).
Figure 1 illustrates the functionality of the gust component.
Component Input
a) Gust amplitude (velocity, angular velocity, gradients).
b) Location of origin of gust field.
c) Location of collocation point.
Component Output
a) Gust velocity at collocation point.
b) Gust angular velocity at collocation point.
17-3.1 Geometry
The component output is the gust velocity v I and angular velocity ω I . These quantities are the perturbation
velocities of the air, in inertial axes I. The inertial axes are used since they are a parent frame for any other
component. A structural dynamic component requires the gust velocity (as an aerodynamic interface) to be in
150 GUST COMPONENT
gust amplitude
gust velocity
at collocation point
location of origin
of gust field
gust angular velocity
at collocation point
location of
collocation point
Figure 17-1 Functionality of gust component.

GUST COMPONENT 151
parent frame axes, not wind/gust axes. The gust amplitude in wind/gust axes G is v G and ω G (component input).
The vectors ⎛ ⎞
uG
v G = ⎝ vG ⎠
wG
⎛ ⎞
pG
ω G = ⎝ qG ⎠
rG
are the perturbation velocities of the air, in negative G axes. Vectors of the gradients are similarly defined. Figure
2 illustrates the definition and sign conventions. In general, a positive gust value is from the corresponding axis,
hence in the direction of the negative axis. Using aircraft conventions for the wind/gust axes, uG is positive aft,
vG is positive from the right, and wG is positive up. The orientation of the wind/gust axes relative inertial axes
(GI) is defined by the wind system piece.
The component input includes the location of the collocation points rCI/I and the origin of the gust field
rOI/I . These quantities must be the position relative the origin of I, in I axes. Typically rOI/I is the origin of the
inertial frame, or the origin of the base frame of the system (hence moves with the system).
17-3.2 Gust Models
The elementary gust models are identified as uniform, angular, and quadratic. For the uniform model, the
gust velocity equals the gust amplitude v G , hence varies with time but is uniform throughout space. The uniform
model does not depend on the collocation point position. For the angular model, the gust velocity is generated by
linear (v G ) and angular (ω G ) gust amplitude terms. Only the angular model produces nonzero angular velocity ω I
from this component. For the quadratic model, the gust velocity is generated by gradient gust amplitude terms,
including uniform, linear, and quadratic variation in space.
For the prescribed gust model, the gust velocity equals the gust amplitude v G times a spatial shape factor
F . The shape factor depends on only one of the three position coordinates in the wind/gust axes. The following
functions are available to produce the gust shape: step; ramp; cosine impulse; sine doublet; square impulse; square
doublet; triangular impulse; triangular doublet; general piecewise linear.
For the tabular gust model, the gust velocity equals the gust amplitude v G times a shape factor F for each of the
three gust velocity components. The shape factors depend on two, three, or four of the four space/time coordinates
in the wind/gust axes. They are obtained from a two-dimensional, three-dimensional, or four-dimensional table.
17–4 Theory
17-4.1 Geometry
The gust field depends on the collocation point position rG = (x, y, z)T in wind/gust axes. This position is
measured from the origin of the gust field:
rG = C GI (rCI/I − rOI/I )
Optionally the gust field can be convected by the wind velocity W (in the x-axis direction). Then
⎛ ⎞
Wt
rG = C GI (rCI/I − rOI/I − vW I
t) = C GI (rCI/I − rOI/I ) + ⎝ 0 ⎠
0
152 GUST COMPONENT
wind/gust axes wind gust
wG
x uG
W
vG
Figure 17-2 Velocity of air produced by wind and gust.

GUST COMPONENT 153
I
where vW is the reference wind velocity (any variation of the wind in a ground boundary layer is neglected here).
For a free body, the wind/gust axes are typically aligned with the velocity axes, in which case the x-coordinate of
rG measures distance along the flight path.
Figure 3 illustrates the geometry of a convected gust. The gust field defines the velocity as a function of
position in space, convected by the wind velocity W . A free body moves through this field with flight velocity V .
For the prescribed model, the gust velocity is proportional to the shape factor F (d − dB ), where dB is the point at
which the gust begins (F (ξ) = 0 for ξ < 0). Figure 3 shows the case in which the distance d is the x-coordinate of
the wind/gust axes. Hence ξ = d − dB is measured in the positive x-axis direction of the wind/gust axes. At time
t = 0 (not necessarily the beginning of the transient), the origin of the inertial and wind/gust axes is at ξ = −dB ,
so the gust shape starts at d = dB . The gust field is being convected in the negative x-axis direction by the wind
velocity W . For a free body, the system is moving at speed V in the positive x-axis direction (of the velocity axes,
which may not be parallel to the wind/gust axes). The distance dB should be chosen so at the start of the transient
(t = tB ) the most forward part of the system has not yet encountered the gust, hence so
CI/I
ξ = (iG )T (C GI rfwd ) + W tB − dB = (xfwd + V tB ) + W tB − dB
is negative, or
dB > xfwd + (V + W )tB
where xfwd is the most forward position relative the velocity axes (a constant; assuming that the velocity axes are
aligned with the wind/gust axes). So dB should be forward of the entire system, at the start of the transient.
17-4.2 Elementary Gust Models
The elementary gust models are identified as uniform, angular, and quadratic. For the uniform model, the
input gust amplitude v G is transformed from wind/gust axes to inertial axes:
v I = −C IG v G
(independent of the collocation point location). For the angular model, the input linear and angular gust amplitudes
produce both linear and angular gust velocities at the collocation point:
v I = −C IG (v G + ω
G rG )
ω I = −C IG ω G
Only the angular model produces nonzero angular velocity ω I from this component. For the quadratic model, the
input amplitudes produce the gust velocity at the collocation point:

v I = −C IG v G + vxG x + vyG y + vzG z + vxx
G 2 G 2
x + vyy G 2
y + vzz G
z + vxy G
xy + vxz G
xz + vyz yz)
with linear and quadratic gradient terms as well as the uniform amplitude v G .
17-4.3 Prescribed Gust Model
For the prescribed gust model, the gust velocity is the product of the input gust amplitude v G and a shape
factor F (ξ):
v I = −C IG v G F (d − dB )
154 GUST COMPONENT
collocation point
wG y y
ξ x x
W V
vG
z z
wind/gust axes velocity axes
(same origin as
inertial axes)
Figure 17-3 Geometry and sign conventions for convected gust.

GUST COMPONENT 155
where d is one of the three position coordinates in the wind/gust axes (x, y, or z). The gust begins at d = dB
(F (ξ) = 0 for ξ < 0). Typically for a convected gust the amplitude vG is constant, so the gust field varies only
along one direction in space. Time variation of the gust at a collocation point is produced by motion of the system
relative the gust field, or by convection of the gust field relative the system. In general, v G can also be a function
of time. The following functions F (ξ) are available to produce the gust shape. In all cases, F = 0 for ξ < 0.
a) Step: F =1
b) Ramp: F =ξ
c) Cosine impulse: F = 1/2(1 − cos 2πξ/L) for 0 < ξ < nL
d) Sine doublet: F = sin 2πξ/L for 0 < ξ < nL
e) Square impulse: F =1 for 0 < ξ < L
f) Square doublet: F =1 for 0 < ξ < L/2
F = −1 for L/2 < ξ < L
g) Triangular impulse: F = 2ξ/L for 0 < ξ < L/2
F = 2 − 2ξ/L for L/2 < ξ < L
h) Triangular doublet: F = 4ξ/L for 0 < ξ < L/4
F = 2 − 4ξ/L for L/4 < ξ < 3L/4
F = 4ξ/L − 4 for 3L/4 < ξ < L
i) General piecewise linear: F = Fi at ξi , i = 1 to I
F0 = 0 at ξ0 = 0, and F = FI for ξ > ξI
For a step occurring at ξ0 , the convention is that the value just before ξ0 is used at ξ0 . The general gust shape is
evaluated by linear interpolation:
ξi − ξ ξ − ξi−1
F = Fi−1 + Fi
ξi − ξi−1 ξi − ξi−1
where ξi−1 < ξ ≤ ξi ; with the convention that (ξ0 , F0 ) = (0, 0), and F = FI for ξ > ξI . A general one-
dimensional shape function can also be defined using the tabular model.
17-4.4 Tabular Gust Model
For the tabular gust model, the gust velocity is the product of the input gust amplitude v G and a shape factor
F for each of the three velocity components:
⎡ ⎤
Fx 0 0
v I = −C IG ⎣ 0 Fy 0 ⎦ vG
0 0 Fz
where each factor F depends on two, three, or four of the four space/time coordinates in the wind/gust axes
(positions x, y, z, and time t). The factors Fx , Fy , and Fz are obtained from a two-dimensional, three-dimensional,
or four-dimensional table.
156 GUST COMPONENT
gust amplitude origin location

v G , ω G , gradients rOI/I rCI/I
gust velocity yes (a) (b) yes (c) yes (c) (d)
Notes:
a) angular velocity yes only for angular gust model
b) gradients yes only for gradient gust model
c) no for uniform gust model
d) yes only if collocation points of output and input are same
Figure 17-4 Functionality of gust component.

Chapter 18
AERODYNAMIC COMPONENTS
18–1 Description
Aerodynamic components include wings and wakes. Wing and wake components can be separate or combined,
depending on the model and solution procedure. Wake geometry and wake-induced velocity can be calculated
by separate components also. If the aerodynamic solution is implemented using separate components, then the
models and parameters of corresponding wing, wake, and wake geometry components must be consistent. A wake
component identifies the parent wing components. A wake geometry component identifies the parent wake and
wing components. The child component can get parameters from the parent components, so duplicate input is
minimized.
Usually several components are required model a system, for both the structural dynamics and the aerody-
namics. While these components are separate entities, they must be able to work together. Hence the component
interfaces must be consistent throughout the system, so the input required by one component is available as output
from another component. Figure 1 illustrates the typical relation between aerodynamic components. The compo-
nent descriptions provide details about the options for each component input and component output, including axes.
In particular there is a standard aerodynamic interface in all structural dynamic components. The aerodynamic
model usually requires reference frame components as well, to sum vectors, transform axes, and provide the frame
motion.
18–2 Wings and Bodies
A wing or body is a surface moving through the air. A wing is a thin surface, which can be defined in terms
of the mid-surface and thickness (or upper and lower surfaces). The boundary conditions of the aerodynamic
problem involve the velocity at the surface, and the pressure loads acting on the surface. The interfaces between
the structure and the air occur at the surface of a wing or body. The interface is discretized, consisting of a set of
collocation points as required by the aerodynamic theory. These collocation points must be defined as connection
points on structural dynamic components. Generally the interface between structural dynamic and aerodynamic
components will be in terms of the velocity and the force (discretized) at the collocation points.
18–3 Inflow and Wake
A wake component essentially solves for the motion of the air. In principle, there is complete mutual influence
between the wings and bodies of the system. This mutual influence can be viewed as arising from the behavior
of the wakes. In practice, some paths of influence may be neglected. There is one wake per wing. A component
may deal with all wakes for a wing set. By convention, the mutual interference between the wings in a set is
158 AERODYNAMIC COMPONENTS
structural force and moment aerodynamic

dynamic FM or wing
or body
position
rr
velocity relative air

vqv ω
circulation
position r aero load
ΓF M
interference interference
velocity vA wake velocity vA
geometry
wake collocation
geometry point
wake
or inflow
or flow field
Figure 18-1 Typical relation between aerodynamic components.

AERODYNAMIC COMPONENTS 159
accounted for completely. In particular, an empirical model may require that all the wings in the set be considered
together. The mutual interference between aerodynamic components (including wing sets) may be accounted for
completely or not, depending on how the interfaces (in terms of interference velocity) between the components
are constructed.
An inflow or wake component calculates the wake-induced velocity at a collocation point. The collocation
points can be on the wing generating the wake, on another aerodynamic component, or at any other point in the
flow field, as required. The structural dynamic components refer to this velocity as an interference velocity vA ,
in the standard aerodynamic interface. The wake-induced velocity is calculated at time t, from the loading at that
and all past times. Some models may require only the loading at the current time t. In general, the inertial frame
I is the only common parent frame for the wake geometry and all collocation points. Thus the induced velocity is
calculated in I axes, from the wake geometry in I axes.
The model used can be uniform inflow (“inflow” components) or nonuniform inflow (“wake” components).
A vortex wake model is typically used to calculate the nonuniform induced velocity on the wings. Alternatively,
an approximate wake model based on ideal-wing theory can be used. The solution from such an approximate
model is called uniform inflow, although the induced velocity might still have a simple variation over the surface
of the wing or wing set.
Nonuniform inflow components calculate the induced velocity from an integral equation, based on a discretized
vortex wake model. The integral equation arises from integration over the wake age, hence the dependence on the
loading at past times. The trim and transient solution methods can evaluate the induced velocity by integrating
over the wake, but an integral equation can not be linearized for the flutter task. In order to use a nonuniform
inflow model in the flutter task, a set of differential equations equivalent to the integral equation is required.
Uniform inflow components are global, quasistatic representations of the wake for a set of wings, using
empirical models based on momentum or vortex theory. The approximations of uniform inflow models usually
require additional corrections in the wing component (such as for the three-dimensional flow effects at the wing
tips). A quasistatic model is designed to be used with the mean (filtered) values of the wing set loading and velocity.
Hence it can be used in all solution tasks, by using the appropriate filters for trim, transient, and flutter models.
Inflow components can also include dynamic inflow models, which are global, low frequency representations
of the wake. These are finite-state models of the unsteady aerodynamic effects of the wake, relating parameters
defining the induced velocity and aerodynamic loading distributions on the wing sets, by means of ordinary
differential equations. Typically they are low order models (in time and space) for the perturbations from the
trim solution, based on simplified representations of the wing set aerodynamics. A quasistatic version of dynamic
inflow can be used, but frequently the time lags are important.
The uniform inflow solution can be used to initialize the nonuniform inflow solution, for better convergence.
This is accomplished using a successive substitution loop solution, with stages or levels:
a) level 1 is uniform inflow;

b) level 2 is nonuniform inflow with rigid or prescribed wake geometry;
c) level 3 is nonuniform inflow with free wake geometry.
In this approach, there are separate inflow and wake components, both modelling the same wing set. The induced
velocity from the inflow and wake components must be summed for use by the wings. Depending on the loop
level, either the inflow component or the wake component must give zero velocity, while the other component
provides the solution. A computational fluid dynamics calculation can be a fourth level, usually initialized by the
nonuniform inflow solution.
The model used to calculate the induced velocity can also depend on the location of the collocation point.
The collocation points locations can be categorized as on this wing set, or off this wing set. For collocation points
on the wing set, the wing and the span station can be identified.
18–4 Wake Geometry
The wake geometry describes the position of the wake vorticity in space. The undistorted geometry is obtained
from the motion of the wing: a wake element is convected by the wind, from the position in the air at which it
was created. This geometry is distorted by the self-induced velocity of all the wakes in the system, as well as the
influence of wings and bodies on the air velocity. In axes moving with the wing, the wake geometry also reflects
the past history of the wing position. The distorted wake geometry is calculated for a set of wings (not necessarily
the same set as for calculation of the induced velocity). Specifying the set defines what mutual interaction between
wakes is accounted for in the wake geometry. If the set does not consist of all wakes in the system, then some
interaction is being neglected.
The wake geometry model can be rigid, prescribed, or free. A rigid model calculates the wake geometry
distortion from the mean induced and interference velocity at the wing. A prescribed wake geometry is obtained
from an empirical model, based on measurements. A free wake geometry is obtained by calculation. The wake
geometry model may depend on the loop solution level.
18–5 Solution Procedure
18-5.1 Aerodynamic Solution Partition
The aerodynamic solution is often implemented by using separate wing, wake, and wake geometry compo-
nents. Separate components allow different wing sets to be used for different parts of the calculations, depending
on the models available and the mutual interaction accounted for. The calculations of a single component can
be performed in separate steps. Separate steps allow quantities to be updated at different rates in the solution.
Separate components or separate steps also allow different solution procedures to be used for the quantities re-
quired. Typically the wake geometry components obtain the wake position from the motion and loading. The
wake components calculate the influence coefficients from the wake geometry, and then the induced velocity from
the influence coefficients and loading. The wing components obtain the loading from the induced velocity. Thus
the trim and transient tasks can use a partitioned solution procedure, in which parts solve the equations and loops
iterate between part solutions. A typical solution procedure for systems involving aerodynamic components may
include a wake geometry loop and/or a circulation loop.
18-5.2 Wake Geometry Loop
For efficiency it is important to move computationally intensive calculations outside inner loops (if allowed
by weak coupling). An important case is the calculation of the wake geometry and influence coefficients for a
nonuniform inflow model, which can be moved outside the trim iteration. This approach is possible when the
AERODYNAMIC COMPONENTS 161
coupling between the wake geometry and the rest of the solution is relatively weak, which is typically so as long
as the wing set is trimmed to a specified speed, orientation, and lift.
The wake geometry loop divides the system into a calculation of the wake geometry and influence coefficients,
and the solution of the rest of the equations. The loop is a successive substitution iteration, with stages or levels:
a) level 1 is uniform inflow;

b) level 2 is nonuniform inflow with rigid or prescribed wake geometry;
c) level 3 is nonuniform inflow with free wake geometry.
So for better convergence, the uniform inflow solution is used to initialize the nonuniform inflow solution, and the
rigid wake geometry solution is used to initialize the free wake geometry solution. Whether the analysis requires
the results of level 3 or level 2 depends on the importance of the free wake geometry or nonuniform inflow to the
problem. A computational fluid dynamics calculation can be a fourth level, usually initialized by the nonuniform
inflow solution.
At each level, iteration will be necessary if there is anything being calculated by the inner loops that will
change the wake geometry (and hence the influence coefficients): such as the wing speed; the mean inflow; the
wing set lift or circulation; or inboard circulation peak locations. With a trim solution inside the wake geometry
loop, it may not in practice be necessary to iterate on the wake geometry at each level. However, if a wing is being
analyzed at a specified angle of attack (rather than trimmed to a specified lift), then the overall geometry of the
wake is not known in advance, only after the wing loading has been calculated. If the wake geometry significantly
influences the wing loading, then iteration will be required in the wake geometry loop. In such cases, a relaxation
factor on the wake geometry is also frequently required.
Consider a rigid wake geometry depending on the mean inflow velocity v. The relaxation can be applied to
the solution for v:
v n = f vn + (1 − f )v n−1
The wake geometry iteration can be represented as an operation
vn = A(v n−1 )
where v is the relaxed inflow. Requiring that the solution is converged:
vn − v n−1 = A(v n−1 ) − v n−1 = 0
defines a nonlinear function of v, with the Newton-Raphson solution
A(v n−1 ) − A v n−1 vn − A v n−1

vn =
= = f vn + (1 − f )v n−1
1−A 1 − A
which gives f = 1/(1 − A ), the ideal relaxation factor for convergence of the successive substitution iteration.
For example, the wake geometry loop requires iteration and a relaxation factor if the system is a rotor or propeller
operating at fixed pitch and zero flight speed (hover). Momentum or vortex theory gives for a hovering rotor
A(λ) = κCT /2λ, hence A = −κCT /2λ2 = −1. So a relaxation factor of 1/2 is required on the wake geometry
iteration.
18-5.3 Circulation Loop
A circulation loop implements an iteration between the wake and inflow components (which calculate the
induced velocity from the loading) and the components of the rest of the system (which calculate the loading for
fixed induced velocity). Such a partition can be used in order to solve the wake components (integral equations)
and structural dynamic components (differential equations) by appropriate methods, and to reduce the number of
variables being solved in one part. The loop is a successive substitution iteration. A relaxation factor of f = 0.10
or 0.05 is often required for nonuniform inflow models.
Consider an inflow solution depending on the wing set lift L. The relaxation can be applied to L:
Ln = f vn + (1 − f )Ln−1
The circulation iteration can be represented as an operation
Ln = A(Ln−1 )
where L is relaxed lift. Requiring that the solution is converged:
Ln − Ln−1 = A(Ln−1 ) − Ln−1 = 0
defines a nonlinear function of L, with Newton-Raphson solution
A(Ln−1 ) − A Ln−1 Ln − A Ln−1
Ln = = = f Ln + (1 − f )Ln−1
1 − A 1 − A
which give f = 1/(1 − A ), the ideal relaxation factor for convergence of the successive substitution iteration.
With the wing set thrust a function of the inflow v (the inner loop solution):
∂A ∂A ∂v ∂v
A = = =c
∂L ∂v ∂L ∂L
then
1
f=
1 + c (∂v/∂L)
(like a lift-deficiency function). In terms of the operation A(L), the successive substitution iteration will converge
if
|1 − f + f A | < 1
For any finite A , a relaxation factor can be found that produces convergence, but the method fails if A = ∞. In
such a case, the iteration typically oscillates about the correct solution, the magnitude of the oscillation decreasing
as f approaches zero. But at f = 0 the iteration is turned off, so the correct solution can never be found. To
analyze the system in such a case, it is necessary to change the definition of the problem: either change the order
that parts are solved in the loop; or change the physical model that is the source of the sensitivity of A to L.
For example, a rotor has ∂A/∂λ ∼

= σa/4, where σ the solidity ratio and a the blade lift-curve slope, so
1
f=
1 + σa
4 (∂λ/∂C T)
Convergence of the circulation iteration is a problem for a hovering rotor at low thrust. Uniform inflow gives for
hover
1/2
CT = c1 θ − c2 λ = c1 θ − c3 CT = A(CT )
c3 −1/2
A (CT ) = − CT
2
so A = ∞ at zero thrust, and the iteration does not converge. This problem can be avoided by fixing the wake
geometry in the calculation of the hovering rotor induced velocity at low thrust (see the component description).
Chapter 19
RIGID AIRFRAME AERODYNAMICS COMPONENT
19–1 Description
A rigid airframe aerodynamics component calculates the aerodynamic forces and moments acting on a rigid
body, consisting of a wing-body, horizontal tail, and vertical tail. The “body” is typically a single structural
dynamic component. The aerodynamic model is quasistatic, designed to be used with the mean (filtered) values
of the component input. The following models are implemented:
a) nonlinear;
b) stability derivatives (linearized).
The trim task can only use the nonlinear model. The analysis uses four collocation points on the airframe: wing-
body, horizontal tail, vertical tail, and stability derivative. The nonlinear model can have separate wing-body,
horizontal tail, and vertical tail properties (allowing the use of separate aerodynamic interference effects at the
three collocation points), or the tail loads can be included in the wing-body properties. The stability derivative
point is only required if a task uses the stability derivatives model. The stability derivatives model does not use
separate tail properties. For consistency, when the stability derivatives model is used for the flutter or transient
task, the nonlinear model for trim can not use separate tail properties, and the wing-body point must be at the same
location as the stability derivative point.
The component implements a quasistatic model of the airframe aerodynamics, using a small number of
collocation points. It does not include a wake model or noncirculatory loads that would be required for high
frequency aerodynamics. Thus it is intended for the static and low frequency motion of an airframe. Elastic motion
often involves localized, high frequency oscillations. In such cases, this component calculates aerodynamic loads
that are not consistent with elastic motion of the airframe. Hence the velocity at the collocation points should be
calculated without the elastic motion (which might be accomplished by the filtering), and the aerodynamic loads
should not be included in the elastic equations of motion of the airframe.
Figure 1 illustrates the functionality of the rigid airframe aerodynamics component.
Component Input
a) Velocity of airframe relative air at the collocation points: velocity, dynamic pressure, rate of change of velocity,
and angular velocity (only velocity and dynamic pressure for tail points).
b) Controls: for flaperon, elevator, aileron, and rudder.
164 RIGID AIRFRAME AERODYNAMICS COMPONENT
velocity body relative air aerodynamic force

at collocation point at collocation point
airframe lift/q
controls sensors
Figure 19-1 Functionality of rigid airframe aerodynamics component.

RIGID AIRFRAME AERODYNAMICS COMPONENT 165
Component Output
a) Aerodynamic loads at the collocation points: force and moment (only force for tail points).
b) Airframe lift/q (L/q at wing-body, horizontal tail, and vertical tail).
c) Sensors (for each collocation point).
19-3.1 Geometry and Axes
The analysis uses the body axes B of the rigid airframe. It is assumed that the velocity relative the air
(component input) is in the B axes, and the calculated loads (component output) are thus also in the B axes. The
required motion relative the air includes the following quantities at the four collocation points:
a) velocity v B ;
b) dynamic pressure q;
c) rate of change of velocity v̇ B ;
d) angular velocity ω B .
Only v B and q are required at the tail collocation points. Time varying quantities must be averaged, so the mean
dynamic pressure will be:
q = 1/2ρv 2 = 1/2ρ(v)2 + 1/2ρσv2
where σv2 is the mean square velocity perturbation. Thus the input must include the averaged dynamic pressure as
a separate quantity. The required aerodynamic loads include the following quantities at the four collocation points:
a) force F B ;
b) moment M B .
Only F B is required at the tail collocation points. The position of the collocation point is the point of action of
the force, and the reference for the moment. The analysis assumes that the same point is used for the velocity
collocation point and the force collocation point. Typically the B axes are the rigid body axes of a structural
dynamic component. For an aircraft, the axes are the system base frame (frame degrees of freedom), with origin
at the center of gravity.
The aerodynamic properties of the airframe are defined in velocity axes A. The angle of attack and sideslip
angle of the airframe (α and β) define the orientation of the A axes relative the B axes. The transformation between
B and A axes is performed by this component as required. Figure 2 illustrates the conventions for the velocity axis
forces. The wing-body moments act about the axes shown. The linear and angular motion are measured relative
the corresponding B axes. In particular, angle of attack and sideslip are positive for linear velocity in the z-axis
and y-axis directions respectively; while yaw is positive for rotation about the z-axis.
The horizontal tail can have a cant angle φHT , defined positive to the left (so at 90 degrees it is equivalent
to a vertical tail). The vertical tail can have a cant angle φV T , defined positive to the right (so at 90 degrees it is
equivalent to a horizontal tail).
19-3.2 Aerodynamic Models
The aerodynamic load Q can be calculated using a nonlinear model (QNL ) or a stability derivatives model
velocity wing-body horizontal vertical

of air tail tail
y LHT
L
Y
DHT
x D DVT
LVT
Figure 19-2 Velocity axis loads.

(linearized, QL ). For the trim task, only the nonlinear model can be used:
Q = QNL
This model can also be used in the transient and flutter tasks. The nonlinear loads can be calculated from simple
equations and/or tables. Separate properties can be defined for the wing-body, horizontal tail, and vertical tail.
The linearized model evaluates the loads from stability derivatives. This model can only be used in the transient
and flutter tasks. The properties are defined for the complete system, at the stability derivative collocation point.
The total velocity is composed of the trim velocity and a perturbation velocity. From the perturbation velocity,
perturbation loads QL are calculated. Then the total aerodynamic loads are composed of the trim loads and the
perturbation loads:
Q = QNL trim + QL
The component must thus save the trim velocity and trim loads if the linearized model is to be used.
19-3.3 Nonlinear Model
The nonlinear aerodynamic model considers the velocity axis loads acting on the wing-body (WB), horizontal
tail (HT), and vertical tail (VT). The following loads are included.
a) Wing-body drag, side, and lift forces: D, Y , and L (along the x, y, and z axes respectively,
acting at the wing-body collocation point).
b) Wing-body roll, pitch, and yaw moments: Mx , My , and Mz (about the x, y, and z axes
respectively, with origin at the wing-body collocation point).
c) Horizontal tail drag and lift forces: D and L.
d) Vertical tail drag and lift forces: D and L.
The loads are defined in terms of force or moment divided by dynamic pressure q. There are thus ten loads in the
model (6 WB, 2 HT, 2 VT). Each load can be obtained from one of the following sources:
a) Equations.
b) Two-dimensional table, as a function of α or β, and Mach number M . Equations are still needed for the rate,
flaperon, aileron, and flap terms; and for the elevator and rudder terms.
c) Three-dimensional or four-dimensional tables, as a function of α, β, M , flaperon, elevator, aileron, rudder, or
flap (any combination of three or four independent variables). Equations are still needed for the rate terms, and
perhaps for the control and flap terms.
19-3.4 Stability Derivatives Model
The stability derivatives model considers the velocity axis loads acting on the entire airframe. The following
loads are included.
a) Linearized drag, side, and lift forces: X, Y , and Z (along the x, y, and z axes respectively,
acting at the stability derivative collocation point).
b) Linearized roll, pitch, and yaw moments: L, M , and N (about the x, y, and z axes
respectively, with origin at the stability derivative collocation point).
The loads are defined in terms of force or moment divided by the reference dynamic pressure. The model has 33
derivatives. The required derivatives are all obtained from one of the following sources:
a) Two-dimensional table, as a function of α and Mach number.

b) Constant coefficients.
19-3.5 Airframe Lift/q
The airframe lift divided by dynamic pressure can be calculated (component output). The vector contains
L/q for the wing-body, horizontal tail, and vertical tail. When the stability derivative model is used, or separate
properties for the tail loads are not used, only the wing-body term is nonzero.
19-3.6 Controls
to a system input piece or to an input/output interface. A particular control vector and element can be used once,
more than once, or not at all. The following airframe aerodynamic controls are identified in these control vectors:
a) flaperon δf (positive for down deflection, increasing wing-body lift);

b) elevator δe (positive for down deflection, increasing horizontal tail lift);
c) aileron δa (positive for down deflection on the right wing, causing a negative roll moment);
d) rudder δr (positive for down deflection, increasing vertical tail lift).
19-3.7 Sensors
Sensor vectors are available corresponding to each collocation point. The sensor vector contains the input
velocity, controls, aerodynamic environment quantities, and the calculated loads (divided by dynamic pressure).
19–4 Theory
19-4.1 Nonlinear Model
The nonlinear aerodynamic model considers the loads acting on the wing-body (WB), horizontal tail (HT),
and vertical tail (VT). The wing-body forces and moments are
⎛ ⎞
−D/q
B
FNL = qC BA ⎝ Y /q ⎠
−L/q
⎛ ⎞
Mx /q
B
MNL = qC BA ⎝ My /q ⎠
Mz /q
The horizontal tail force is ⎛ ⎞

−D/q
B
FNL = qRH C BA ⎝ 0 ⎠
−L/q
where RH = XφHT . The vertical tail force is

⎛ ⎞
−D/q
B
FNL = qRV C BA ⎝ 0 ⎠
−L/q
B
where RV = X90−φV T . For the nonlinear model, the loads at the stability derivative point are set to zero: FNL =0
B
and MNL = 0. The angle of attack and sideslip angle provide the transformation between body axes and velocity
axes:
C BA = Yα Z−β
This definition corresponds to yaw-then-pitch of the airframe body axes relative the air velocity vector. By
definition, the velocity is along the x-axis in the A axes:
⎛ ⎞
V
v B
=C BA ⎝0⎠
0
from which the angle of attack and sideslip in terms of the components of v B are obtained. Thus at the wing-body
collocation point, the following quantities define the aerodynamic environment.
α = tan−1 v3B /v1B

β = sin−1 v2B /|v B |
v B v̇ B − v3B v̇1B
α̇ = 1B 32
(v1 ) + (v3B )2
⎛ ⎞
p
⎝ q ⎠ = ωB
r
The Mach number M = |v B |/cs is calculated at each collocation point.
The aerodynamic model includes the effects of aerodynamic interference from the wing-body on the tail, in
terms of an angle of attack change (positive for increasing angle of attack at the tail); and a sideslip change σ
(positive for increasing sideslip at the tail). Thus the components of the velocity vector
⎡ ⎤
1 σ
v = ⎣ −σ 1 0 ⎦ vB
− 0 1
give the angle of attack at the tail:
v3 cos φHT + v2 sin φHT v2 cos φHT − v3 sin φHT

αHT = tan−1 βHT = sin−1
v1 |v|
v2 cos φV T + v3 sin φV T −v3 cos φV T + v2 sin φV T
αV T = tan−1 βV T = sin−1
v1 |v|
The angle of attack change is calculated from:

% &
HT
= E (L/q)W B − (Lα /q)W B α̇W B
V
where E = ∂/∂(L/q) can be estimated from the airframe geometry. The second term in the expression for is
produced by a time lag τ between the angle of attack at the wing and at the tail. Thus a differential equation model
is:
∂
τ ˙ + = α
∂α
For airplane flight dynamics, it is conventional to approximate this equation as follows:

∂ ∼ ∂ ∂ ∂
= (1 + τ s)−1 α = (1 − τ s) α= (α − τ α̇) = (Lα /q) α − τ (Lα /q) α̇
∂α ∂α ∂α ∂(L/q)
Assuming that the wing wake causing is convected by the airframe velocity V , it follows that τ = HT /V , where
HT is the horizontal tail length (ref. 1). The sideslip change is calculated from:
∂σ ∂σ ∂σ ∼ hV T
σ= β+ r+ p= pW B
∂β ∂r ∂p V
where hV T is the vertical tail height (ref. 1).
The conventional aerodynamic description of the airframe uses wind axis loads, and a yaw-then-pitch definition
of the angle of attack and sideslip, as given above. With this definition, the angles for special cases are

v1 = 0 : α = 90 sign(v3 ) β = sin−1 v2 / v22 + v32

v2 = 0 : β=0 α = sin−1 v3 / v12 + v32

v3 = 0 : α=0 β = sin−1 v2 / v12 + v22
So for v1 = 0, β is not 90 degrees, and there is a jump in the value of α at v3 = 0. If β is near 90 degrees, then α
does not produce a first-order change in the velocity relative the body axes. So for sideward flight in particular, a
pitch-then-yaw definition of the angle of attack and sideslip might be useful. For this option, the rotation matrix
is C BA = Z−β Yα . Then
α = sin−1 v3B /|v B |
β = tan−1 v2B /v1B
(v12 + v22 )v̇3 − v3 (v1 v̇1 + v2 v̇2 )
α̇ =
|v|2 v12 + v22
for the wing-body, and
v3 cos φHT + v2 sin φHT v2 cos φHT − v3 sin φHT
αHT = sin−1 βHT = tan−1
|v| v1
v2 cos φV T + v3 sin φV T −v3 cos φV T + v2 sin φV T
αV T = sin−1 βV T = tan−1
|v| v1
for the tail. With a pitch-then-yaw definition, the angles for v1 = 0 are

β = 90 sign(v2 ) α = sin−1 v3 / v22 + v32
(the angles for v2 = 0 and v3 = 0 are the same as for yaw-then-pitch). So for v1 = 0, β is ±90 degrees, and α is
produced by v3 (roll). If α is near 90 degrees, then β does not produce a first-order change in the velocity relative
the body axes.
Another option is to use body axes rather than wind axes for the aerodynamic loads. With the loads in body
axes, the rotation matrix C BA is not used to calculate the loads. The wing-body and tail loads are thus
⎛ ⎞ ⎛ ⎞
−D/q −D/q
B
FNL = q ⎝ Y /q ⎠ B
FNL = qRH ⎝ 0 ⎠
−L/q −L/q
⎛ ⎞ ⎛ ⎞
Mx /q −D/q
B
MNL = q ⎝ My /q ⎠ B
FNL = qRV ⎝ 0 ⎠
Mz /q −L/q
Now −L, Y , and −D are respectively the z-axis (down), y-axis, and x-axis (forward) components of the aero-
dynamic force in the body axes (in the canted axes for the tail forces). The options of body axis loads and/or
pitch-then-yaw definition of the angles are not conventional, hence must be used with caution, but they can be
useful for operating conditions such as sideward flight. The equations or tables for the aerodynamic loads must
be consistent with these choices. The conventional description is recommended whenever the stability derivative
model is used.
There are ten loads in the model (6 WB, 2 HT, 2 VT). Each load can be obtained from one of the following
sources: equations, two-dimensional table, three-dimensional tables, or four-dimensional tables. The table options
still require equations for some terms.
19-4.1.1 Equations
The following equations can be used to calculate the velocity axis loads. The symmetric and antisymmetric
loads are assumed to be completely uncoupled. For the wing-body:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
My /q M0 /q + (Mα /q)(αW B + iW B M ) Mδe /q
⎝ D/q ⎠ = ⎝ fW B + (gW B − fW B ) sin2 (αW B + iW B D ) + DI (L/q)2 ⎠ + ⎝ Dδe /q ⎠ δe
L/q L0 /q + (Lα /q)(αW B + iW B L ) Lδe /q
⎡ ⎤
Mδf /q MδF /q
δf
+ ⎣ Dδf /q DδF /q ⎦
δF
Lδf /q LδF /q
where δF is a flap angle (a parameter rather than a control); and

⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎡ ⎤⎛ ⎞
Mx /q Nx0 /q + (Nxβ /q)βW B Nxδr /q V Nxp /q V Nxr /q Nxδa /q p/V
⎝ Mz /q ⎠ = ⎝ Nz0 /q + (Nzβ /q)βW B ⎠ + ⎝ Nzδ /q ⎠ δr + ⎣ V Nzp /q V Nzr /q Nzδa /q ⎦ ⎝ r/V ⎠
r
Y /q Y0 /q + (Yβ /q)βW B Yδr /q V Yp /q V Yr /q Yδa /q δa
The elevator and rudder terms in the wing-body loads should only be used if the tail loads are not present. The
last two terms in the wing-body drag both give an α2 dependence. The second term represents stall drag, while
the third term is the induced drag:
∂Di /q 1
DI = =
∂(L/q)2 πe 2W
where W is the wing span, and e < 1 the span efficiency factor. For the horizontal tail:

D/q fHT + (gHT − fHT ) sin2 (αHT + iHT D ) Dδe /q
= + δe
L/q L0 /q + (Lα /q)(αHT + iHT L ) Lδe /q
For the vertical tail:

D/q fV T + (gV T − fV T ) sin2 (αV T + iV T D ) Dδr /q
= + δr
L/q L0 /q + (Lα /q)(αV T + iV T L ) Lδr /q
The incidence angles i are measured relative to the B axes, such that zero load is obtained for α = −i. Lift and
moment stall are modelled by using the truncated angle of attack
αe = sign(α) min(|α|, αmax )
in place of α in the wing-body lift and pitch moment, and in the tail lift.
19-4.1.2 Two-Dimensional Table
The velocity axis loads can be obtained from a two-dimensional table, as a function of α or β, and Mach
number M . Equations are still needed for the rate, flaperon, aileron, and flap terms; and for the elevator and rudder
terms. The symmetric and antisymmetric loads are assumed to be completely uncoupled. For the wing-body:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎡ ⎤
My /q ScCM Mδe /q Mδf /q MδF /q
⎝ D/q ⎠ = ⎝ SCD ⎠ + ⎝ Dδe /q ⎠ δe + ⎣ Dδf /q DδF /q ⎦ δf
δF
L/q SCL Lδe /q Lδf /q LδF /q
where δF is a flap angle (a parameter rather than a control); and

⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎡ ⎤⎛ ⎞
Mx /q SbC Nxδr /q V Nxp /q V Nxr /q Nxδa /q p/V
⎝ Mz /q ⎠ = ⎝ SbCn ⎠ + ⎝ Nzδ /q ⎠ δr + ⎣ V Nzp /q V Nzr /q Nzδa /q ⎦ ⎝ r/V ⎠
r
Y /q SCy Yδr /q V Yp /q V Yr /q Yδa /q δa
The elevator and rudder terms in the wing-body loads should only be used if the tail loads are not present.
The symmetric coefficients (CM , CD , and CL ) are a function of angle of attack αW B and Mach number. The
antisymmetric coefficients (C , Cn , and Cy ) are a function of sideslip angle βW B and Mach number. The
coefficients are based on the wing-body area S, chord c, and span b. For the horizontal tail:

D/q SCD Dδe /q
= + δe
L/q SCL Lδe /q
The coefficients (CD and CL ) are a function of angle of attack αHT and Mach number. The coefficients are based
on the horizontal tail area S. For the vertical tail:

D/q SCD Dδr /q
= + δr
L/q SCL Lδr /q
The coefficients (CD and CL ) are a function of angle of attack αV T and Mach number. The coefficients are based
on the vertical tail area S.
19-4.1.3 Three-Dimensional or Four-Dimensional Tables
The velocity axis loads can be obtained from three-dimensional or four-dimensional tables, as a function of
angle of attack α, sideslip angle β, Mach number M , flaperon δf , elevator δe , aileron δa , rudder δr , or flap δF (any
combination of three or four independent variables). Equations are still needed for the rate terms, and perhaps
for the control and flap terms. The loads are calculated using the same equations as for a two-dimensional table.
Control effects may be in the equations or in the tables, but should not be in both. The elevator and rudder terms
in the wing-body loads should only be used if the tail loads are not present. Sideslip angle, flaperon deflection,
aileron deflection, and flap angle are ignored for the horizontal and vertical tails. Using a two-dimensional table,
the coefficients must all be functions of the same independent variables (angle and Mach number). Using three-
dimensional or four-dimensional tables, different independent variables can be defined for each coefficient. If
there is only one value in the table for the third independent variable, the three-dimensional table is effectively
two-dimensional.
19-4.2 Stability Derivatives Model
The stability derivatives model considers the loads acting on the entire airframe. The loads acting at the
wing-body collocation point are maintained at their trim values (from the nonlinear model). The perturbation
loads are calculated from stability derivatives and the perturbed motion, and applied at the stability derivative
collocation point. The stability derivative forces and moments are
⎛ ⎞
X/q
BA ⎝
FLB = qtrim Ctrim Y /q ⎠
Z/q
⎛ ⎞
L/q
BA ⎝
MLB = qtrim Ctrim M/q ⎠
N/q
with q and C BA evaluated using the reference (trim) motion. The angle of attack and sideslip angle provide the
transformation between body axes and velocity axes:
BA
Ctrim = Yα Z−β
The angle of attack, sideslip angle, Mach number, and dynamic pressure are obtained from the trim velocity at the
stability derivative collocation point:
α = tan−1 v3B /v1B
β = sin−1 v2B /|v B |
M = |v B |/cs
These quantities are calculated and saved during the trim task. The definition of the velocity at the stability
derivative collocation point must be consistent with the definition of the stability derivatives (hence the stability
derivative and wing-body points are defined separately). In particular, the stability derivatives may or may not
include the effects of aerodynamic interference in the reference quantities. The perturbation motion at the stability
derivative collocation point is obtained by subtracting the trim motion, and rotating to velocity axes:
⎛ ⎞
u
⎝ v ⎠ = C AB (vtotal
B
− vtrim
B
)
w
⎛ ⎞
p
⎝ q ⎠ = ωtotal
B
− ωtrim
B
r
⎛ ⎞
u̇
⎝ v̇ ⎠ = C AB (v̇total
B
− v̇trim
B
)
ẇ
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
δf δf δf
⎜ δe ⎟ ⎜ δe ⎟ ⎜ δe ⎟
⎝ ⎠=⎝ ⎠ −⎝ ⎠
δa δa δa
δr δr total δr trim
The trim quantities are calculated and saved during the trim task. The option to use a pitch-then-yaw definition of
angle of attack and sideslip, described for the nonlinear model, applies to the stability derivative model as well,
but is too unconventional to be recommended. The option to use loads in body axes instead of wind axes is not
applied to the stability derivative model.
The following expressions are used to evaluate the aerodynamic loads in terms of stability derivatives. The
symmetric and antisymmetric loads are assumed to be completely uncoupled.

⎞⎛
ẇ/V 2
⎛ ⎞ ⎡ ⎤ ⎜ u/V ⎟
M/q V Mα̇ /q 2M/q + V Mu /q Mα /q V Mq /q Mδe /q Mδe /q ⎜ ⎟
⎝ −X/q ⎠ = ⎣ V Dα̇ /q ⎜ w/V ⎟
2D/q + V Du /q Dα /q − L/q V Dq /q Dδe /q Dδe /q ⎦ ⎜ ⎟
⎜ q/V ⎟
−Z/q V Lα̇ /q 2L/q + V Lu /q Lα /q + D/q V Lq /q Lδe /q Lδe /q ⎝ ⎠
δe
δf
⎛ ⎞
ẇ/V 2
⎡1 ⎤
2
2 Sc CMα̇ Sc(2CM + M CMM ) ScCMα 1 2
2 Sc CMq ScCMδe ScCMδe ⎜ ⎜
u/V ⎟
⎟
⎜ w/V ⎟
=⎣ 1
2 ScCDα̇ S(2CD + M CDM ) S(CDα − CL ) 12 ScCDq SCDδe SCDδe ⎦ ⎜ ⎟
⎜ q/V ⎟
1
2 ScCLα̇ S(2CL + M CLM ) S(CLα + CD ) 12 ScCLq SCLδe SCLδe ⎝ δ ⎠
e
δf
⎛ ⎞
⎛ ⎞ ⎡ ⎤ v/V
L/q Nxβ /q V Nxp /q V Nxr /q Nxδa /q V Nxδr /q ⎜ p/V ⎟
⎝ N/q ⎠ = ⎣ Nzβ /q ⎜ ⎟
V Nzp /q V Nzr /q Nzδa /q V Nzδr /q ⎦ ⎜ r/V ⎟
⎝ ⎠
Y /q Yβ /q V Yp /q V Yr /q Yδa /q Yδr /q δa
δr
⎛ ⎞
⎡ ⎤ v/V
1 2 1 2
SbC β 2 Sb C p 2 Sb C r SbC δa SbC δr ⎜ p/V ⎟
⎜ ⎟
= ⎣ SbCnβ 1 2
2 Sb Cnp
1 2
2 Sb Cnr SbCnδa SbCnδr ⎦ ⎜ r/V ⎟
1 1 ⎝ ⎠
SCyβ 2 SbCyp 2 SbCyr SCyδa SCyδr δa
δr
The model has 33 coefficients. The required coefficients are either all constants; or all obtained from a two-
dimensional table, as a function of trim angle of attack and Mach number. The coefficients are based on the
airframe area S, chord c, and span b:
lift side force

CL = Cy =
qS qS
drag roll moment
CD = C =
qS qSb
pitch moment yaw moment
CM = Cn =
qSc qSb
The dimensionless rate derivatives are:

∂C ∂C
(C)α̇ = (C)p =
∂(α̇c/2V ) ∂(pb/2V )
∂C ∂C
(C)q = (C)r =
∂(qc/2V ) ∂(rb/2V )
The trim lift, drag, and moment coefficients are required (CL , CD , and CM ). These quantities are the total loads,
evaluated from the saved trim wing-body loads. Hence separate tail loads can not be used for trim, and the wing-
body point must be at the same location as the stability derivative point (otherwise the locations of the collocation
points would be needed in order to calculate CM ).
19–5 References
1) Etkin, B. Dynamics of Flight. John Wiley and Sons, New York, 1959.
vW B vHT vV T vSD control
FW B yes no no X yes
FHT yes yes no X yes
FV T yes no yes X yes
FSD X X X yes yes
L/q yes yes yes yes yes
sensor yes (a) yes (a) yes (a) yes (a) yes (a)
Notes:
a) yes only if collocation points of sensor and input
are same
Figure 19-3 Functionality of rigid airframe aerodynamics component.

Chapter 20
AIRFRAME FLOW FIELD COMPONENT
20–1 Description
An airframe flow field component calculates the perturbation aerodynamic interference velocity produced by
an airframe (or other object). The aerodynamic model is quasistatic, designed to be used with the mean (filtered)
values of the component input. The interference can be obtained from the following sources:
a) velocity calculated;
b) velocity from table file.
The calculated interference velocity is that produced by a simple representation of the airframe, consisting of a
set of wings and bodies. A wing is modelled as a horseshoe vortex (lift) and a doublet line (thickness). A body
is modelled by the potential flow about a nonlifting body of revolution. The table file is used to provide access to
externally calculated interference, typically obtained using a panel method with a detailed representation of the
airframe geometry. Interference velocities in a table file will have been calculated for specified locations of the
collocation points relative the airframe. The internal calculation of the velocities can use the actual locations of
the collocation points and airframe, but for efficiency only implements a simple representation of the airframe
aerodynamics.
The component implements a quasistatic model of the aerodynamics. It does not include the shed wake or
noncirculatory loads that would be required for high frequency aerodynamics. By using the appropriate frame for
the analysis, the location and velocity of the wings and bodies can usually be filtered to eliminate high frequency
oscillations. Thus the flow field will be quasistatic relative to the analysis frame. The location of the collocation
point relative the wings and bodies might still be time varying however, hence should not be filtered.
Figure 1 illustrates the functionality of the airframe flow field component.
Component Input
Velocity calculated:
a) Airframe velocity relative air.
b) Location of collocation point.
c) Location of wings (left tip, middle, and right tip) and bodies (center and nose).
d) Airframe lift/q (L/q at wing-body, horizontal tail, and vertical tail).
Velocity from table file:
a) Airframe velocity relative air.
178 AIRFRAME FLOW FIELD COMPONENT
location of collocation point perturbation aero

velocity at
location of wings, bodies collocation point
airframe velocity relative air

sensors
airframe lift/q
Figure 20-1 Functionality of airframe flow field component.

AIRFRAME FLOW FIELD COMPONENT 179
Component Output
a) Perturbation aerodynamic velocity at collocation point.
b) Sensors: circulation Γ/V of each wing (available if velocity is calculated).
The analysis is performed in axes F, which must be a frame or parent frame of the body and of all the
collocation points. It is assumed that the locations and velocity (component input) are in the F axes, and the
calculated interference velocity (component output) are thus also in the F axes. The locations required are the
position relative the origin of F, in F axes: rF . Let rAF/F be a location on the airframe (on a wing or body), and
rCF/F be the location of a collocation point (where the interference velocity is required). Then
rCA/F = rCF/F − rAF/F
is the position from the airframe point to the collocation point. The airframe velocity relative the air is in the F
axes also: V F . Then the magnitude and direction of the airframe velocity relative the air is:
V = |V F |
F = V F /|V F |
If V = 0, then zero interference velocity is returned. The interference velocity is produced at each collocation
F
point, in F axes: vA . The component first obtains the interference velocity of the air, in F axes, scaled with V :
Δv F /V . Then
F
vA = (Δv F /V ) V
is the required interference.
20-3.2 Velocity From Table File
The velocity can be obtained from a two-dimensional table. The table dependent variables are the three
components of the interference velocity Δv F /V . The frame F must be consistently identified when these velocities
are used by other components. The first independent variable of the table is a span station value. The span station
must be specified for each collocation point. Then the table data are used by interpolating the span station variable
to the collocation point value. If the collocation point span stations match the table entries exactly, then the
interpolation has no effect. A typical use of the component is with a wing, hence the term “span station.” More
generally, the span station is just a variable (possibly even discrete) that connects the collocation points with the
table entries. The second independent variable of the table is time or azimuth. The table data are interpolated
(perhaps cyclically for azimuth) to the current time value t. The following options are implemented.
a) Table function of span and time τ : τ = t − t0 , where t0 is a specified reference time.

b) Table function of span and azimuth ψ: ψ = Ωt + ψ0 , where the frequency Ω is obtained from a specified period.
c) Table function of span only, constant in time (the table class is still two-dimensional, but there is only one time
value in the table).
20-3.3 Calculated Velocity
The calculated interference velocity is that produced by a simple representation of the airframe, consisting of
a set of wings and bodies. A wing is modelled as a horseshoe vortex (lift) and a doublet line (thickness). A body
is modelled by the potential flow about a nonlifting body of revolution. The bodies of revolution considered are
spheres, ellipsoids, and airfoil-shaped bodies.
20-3.3.1 Wings
The airframe lifting surfaces are represented by a set of simplified wings. The model for each wing consists
of a constant strength horseshoe vortex for lift, and a constant strength dipole line for thickness. Such a model is
correct in the far field, but will require calibration for reasonable results in the near field. To allow for sweep and
dihedral, the wing is defined by the quarter-chord in two straight segments. Figure 2 illustrates the geometry. The
circulation and thickness lines are placed a distance xcw behind the leading edge (where cw is the wing chord), to
allow calibration of the velocity field near the wing. The wing geometry is thus described by the five points rLF ,
rLT , rM , rRT , and rRF (left downstream, left tip, middle, right tip, and right downstream). The left tip, middle,
and right tip (at the quarter chord) are input locations; the downstream points are calculated from the airframe
velocity direction. The wing is not necessarily horizontal or symmetric, so the terms “right” and “left” do not
necessarily correspond to the actual configuration. The right-hand rule for a vector from the “left” tip to the “right”
tip defines the positive direction of bound circulation (hence the wing upper surface).
The circulation-produced interference velocity is obtained by summing over the four segments of the horseshoe
vortex:
Δv F Γ
= Δv
V V
wing
where Δv is the velocity produced by a constant strength vortex line segment, of unit strength. The wing circulation
strength is:
Γ L/q cw CL
= =
V 2bw 2
where bw is the wing span. This circulation is evaluated from a combination of an input value, and the wing-body,
horizontal tail, and vertical tail lifts:
Γ Γ0 fW (L/q)W B + fH (L/q)HT + fV (L/q)V T
= +
V V 2bw
The factors fW , fH , and fV can be used to calibrate the model, and to account for the effects of wing sweep. The
thickness-produced interference velocity is obtained by summing over the two segments of the dipole line:
Δv F
= −AXS SF
V
wing
where SF is the velocity produced by a constant strength dipole line segment. Ignoring sweep, the dipole strength
has been obtained from the wing airfoil cross section area AXS (approximately 0.68τw c2w , where τw is the wing
thickness ratio).
20-3.3.2 Bodies
The airframe nonlifting surfaces are represented by a set of bodies of revolution. The interference velocities are
obtained from the potential flow solution (no separation), at arbitrary angle of attack. The following configurations
are considered.
a) An ellipsoid, created by revolving an ellipse about its major or minor axis (exact solution).
b) A sphere (exact solution).
velocity of
rRT horseshoe vortex
airframe
rRF
relative air
rM dipole line
rLF
rLT
velocity of z
airframe
relative air y
rN rC x
body of revolution
Figure 20-2 Geometry of wings and bodies.

c) An airfoil-shaped body, created by revolving the NACA 4-digit airfoil thickness distribution about the chord
line (modified slender-body-theory solution).
The body of revolution is analyzed in a coordinate system A, with the body length scaled to . The A axes have
the origin at the body center, and the x-axis directed aft along the axis of revolution. Figure 2 illustrates the
geometry. The body geometry is described by the two points rC and rN (center and nose), in the F axes. From
these two points, the rotation matrix C AF between the A and F axes can be obtained. Thus the scaled position of
the collocation point relative the origin of A, in A axes, is:
⎛ ⎞
x
) = ⎝y⎠
AF/F
rCA/A = C AF (rCF/F − rC
L
z
where L is the actual body length (along the axis of revolution). The velocity of the air relative to the body has
direction
A = −C AF F
in A axes. The interference velocity is obtained from the potential flow solution for the body of revolution:
Δv F Δv A
= CF A = C F A T A = −C F A T C AF F
V V
where T A is the velocity produced by the body, in A axes.
20-3.4 Sensors
A sensor vector is available if the velocity is calculated. The sensor vector contains the bound circulation
Γ/V of each wing.
20–4 Theory
20-4.1 Calculated Velocity, Wings
20-4.1.1 Geometry
The airframe lifting surfaces are represented by a set of simplified wings. The model for each wing consists
of a constant strength horseshoe vortex for lift, and a constant strength dipole line for thickness. The circulation
and thickness lines are placed a distance xcw behind the leading edge, where cw is the wing chord. The trailed
vortices extend from the wing tips to 10bw downstream, where bw is the wing span. Hence the wing geometry is
defined by the following points:
1 F
rLT = (rLT )QC − x − cw
4
1
rM = (rM )QC − x − cw F
4
1 F
rRT = (rRT )QC − x − cw
4
rLF = rLT − 10bw F
rRF = rRT − 10bw F
where the subscript QC refers to the quarter-chord. These points define four vortex line segments, and two dipole
line segments.
Two dimensional airfoil theory provides guidance for the placement of the circulation and thickness lines
such that the correct velocity will be obtained at the collocation point. Consider a symmetric Joukowski airfoil,
with chord, thickness ratio, and leading edge position as follows:
1
= 3 + 2 +
1 + 2
√
3 3
τ=
4
1
xL = − 1 + 2 +
1 + 2
For the airfoil with circulation but no thickness ( = 0), a point vortex at z = x + iy = −μ with
1
μ=1−
ζ
matches the airfoil velocity exactly at ζ, where z = ζ + 1/ζ. In the far field, the requirement is that the vortex be
at the quarter-chord (μ = 1). In the near field, the vortex position can be obtained for the velocity matched in the
plane of the wing: y = 0 and 4
x x 2
ζ= − −1
2 2
(x < −2 forward of the leading edge). For an airfoil with thickness but no circulation, a doublet (cylinder) at
z = −μ with
(ζ + ) 1 + /2
μ= 4 −z
ζ
+
ζ +1 2
matches the airfoil velocity exactly at ζ. In the far field, the requirement is that the doublet be at 3/8 chord
(μ = 1/2). For a point in the plane of the wing, a distance cw Δx forward of the leading edge, x = xL − Δx. The
position of the vortex or doublet required to match the velocity is then −(xL + μ) , as a fraction of the chord aft of
the leading edge. This is the parameter required in the present definition of the wing geometry. The parameter is
not very sensitive to Δx for points more than a quarter-chord ahead of the leading edge. As an example, to match
the velocity 1/2 chord ahead of the leading edge, the vortex should be placed about 3/16 chord aft of the leading
edge, and the doublet about 1/4 chord aft of the leading edge (for both, closer to the leading edge than is required
to match the far field velocity).
20-4.1.2 Vortex Line Segment
The velocity produced by a constant strength vortex line segment is required. Figure 3 shows the configuration.
The vortex has strength Γ. The line segment extends from point 1 to point 2 in space, and the velocity is required
at point P. The geometry is thus defined by the position vectors r1 and r2 , from the ends of the line segments to
P. The position vectors are in axes F, and hence the induced velocity will be in axes F. The Biot-Savart law gives
the induced velocity produced by this line segment:
$
1 Γ r × dσ
Δv = −
4π r3
where r is the vector from the element dσ on the segment, to the point P; and r = |r|. The coordinate σ is measured
along the vortex segment, from s1 to s2 :
1
s1 = (r1 · r2 − r12 )
s
1
s2 = (r22 − r1 · r2 ) = s1 + s
s
r1
1
r
r2
dσ
Γ
2
Figure 20-3 Constant strength vortex line segment

where s is the length of the segment:
s2 = |r1 − r2 |2 = r12 + r22 − 2r1 · r2
Write r = rm − σ1 , where rm is the minimum distance from the vortex line (including its extension beyond the
end points of the segment) to the point P, and 1
is the unit vector in the direction of the vortex:
1 1
rm = r1 (r22 − r1 · r2 ) + r2 (r12 − r1 · r2 ) = (r1 s2 − r2 s1 )
s2 s
1
1
= (r1 − r2 )
s
The vectors rm and 1

are perpendicular, and
s = r12 r22 − (r1 · r2 )2

2 2
rm
It follows that $
r1 × r2 s2
1
Δv = Γ 2 + σ )3/2
dσ
4πs s1 (r m 2
% &σ=s2
r1 × r2 σ/rm2
= Γ
4πs (r2 + σ )1/2
m 2 σ=s1

r1 × r2 s2 s1
= 2
Γ −
4πsrm r2 r1
The influence of the vortex core is accounted for by multiplying the induced velocity of the line segment by the
factor
2
min(rm /rc2 , 1)
for concentrated vorticity (solid body rotation). The core radius rc is the location of the maximum tangential
velocity.
20-4.1.3 Dipole Line Segment
The velocity produced by a constant strength dipole line segment, representing the wing thickness, is required.
Consider a thin, planar wing. The air has speed V relative to the wing (in the x direction; with x positive aft and
z positive up). The velocity potential of a source distribution represents a wing with thickness but no lift:
$ $$
q dS(y) V dt/dξ
φ(x) = =− dη dξ
wing 4πr 4πr
where r = x − y, x = (x, y, z), y = (ξ, η, 0), and t(x) is the wing chordwise thickness distribution. Integrating
by parts gives $$
Vt d 1
φ= dη dξ
4π dξ r
which is a distribution of dipoles in the ξ direction (the direction of V ). The doublet line approximation follows
by evaluating the integrand at the wing mean line:
$ $ $
V d 1 VA d 1
φ∼
= dη t dξ = dη
4π dξ r ξm 4π dξ r
where A = ∫ t dξ is the wing cross-section area. This approximation implies far field, large aspect-ratio assump-
tions. The corresponding two-dimensional result shows that in the far field the thin wing is equivalent to a cylinder
with area equal to one-half the wing cross-section area. The dipole line result is generalized to an arbitrary direction
of V (the velocity of the wing relative to the air now):
%$ &
A 1
φ=− ∇y dη · V
4π r
Then the velocity is:

Δv = ∇x φ = −A S · V
The dyadic (matrix) S is evaluated following the geometry and notation for the vortex line segment:
$
1 1
S= ∇x ∇y dσ
4π r
$
1 1 3rr
= − 5 dσ
4π r3 r
1
= (ii + jj + kk)I1 + rm rm I2 + (rm 1+1 rm )I3 + 11
I4
4π
where r = rm − σ1
, r2 = rm
2
+ σ 2 , and
$
dσ σ
I1 = = 2
r3 rm r
$ 2
dσ σ rm
I2 = −3 =− 4 +2
r5 rm r r 2
$
σ dσ 1
I3 = 3 =− 3
r5 r
$ 2
σ dσ σ3
I4 = −3 5
=− 2 3
r rm r
evaluated from σ = s1 to σ = s2 .
20-4.2 Calculated Velocity, Bodies
20-4.2.1 Geometry
The airframe nonlifting surfaces are represented by a set of bodies of revolution. The body of revolution
is analyzed in a coordinate system A, with the body length scaled to . The A axes have the origin at the body
center, and the x-axis directed aft along the axis of revolution. The actual length of the body is L (along the axis
of revolution). The body geometry is described by the two points rC and rN (center and nose), in the F axes. The
point rN must be on the axis of revolution, forward of the center, but need not be the actual nose point as used
here. The rotation matrix between the A and F axes is obtained from the unit vectors of the A axes, expressed in
F axes: ⎡ T ⎤
i
C AF = ⎣ j T ⎦
kT
The unit vector along the x-axis is: ⎛ ⎞
i1
Δr
i= = ⎝ i2 ⎠
|Δr|
i3
where Δr = rC − rN . The orientation of j and k is arbitrary, since the body is axisymmetric. Let in be the largest
component of the vector i. Then

jn−1 = −in+1 / i2n+1 + i2n−1
jn = 0

jn+1 = in−1 / i2n+1 + i2n−1
gives j perpendicular to i. Finally, k = i × j.
20-4.2.2 Sphere
The induced velocity of a sphere is derived following references 1 and 2. The length (diameter) of the sphere
is = 1. The interference velocity is:
⎡ ⎤
A y 2 + z 2 − 2x2 −3xy −3xz
Δv 1 ⎣ ⎦ A
= −3yx x2 + z 2 − 2y 2 −3yz
V 16s5
−3zx −3zy x + y − 2z
2 2 2
where s2 = x2 + y 2 + z 2 . Inside the sphere (s < .5), the interference velocity is zero.
20-4.2.3 Ellipsoid
The induced velocity of an ellipsoid is derived following references 1 and 2. The body is generated by
revolving an ellipse about its major or minor axis. The x-axis is the axis of revolution, and is the length of the
body along the axis of revolution. The thickness ratio τ is the ratio of the axes of the ellipse. An ovary (thin)
ellipsoid is generated by revolving an ellipse about its major axis. Then is the major axis, and τ is the ratio of
minor to major axes (τ < 1). A planetary (flat) ellipsoid is generated by revolving an ellipse about its minor axis.
Then is the minor axis, and τ is the ratio of major to minor axes (τ > 1). The limit of τ = 0 gives a line; the
limit of τ = ∞ gives a disk; and τ = 1 is a sphere.
√
For the ovary (thin) ellipsoid, the length (major axis) is = 2/ 1 − τ 2 , where the thickness ratio τ is less
than one. The ellipsoid is analyzed using coordinates ν and μ:
x = μν
r2 = y 2 + z 2 = (1 − μ2 )(ν 2 − 1)
or
5 2
1 + x2 + r2 1 + x2 + r2
ν2 = + − x2
2 2
μ = x/ν
The coordinate ν ranges from ν0 = /2 on the surface of the ellipsoid, to infinity. With
% &−1
ν0 1 ν0 + 1
A= 2 − ln
ν0 − 1 2 ν0 − 1
% &−1
1 ν0 + 1 ν02 − 2
B= ln −
2 ν0 − 1 ν0 (ν02 − 1)
the interference velocity is:

⎡ ⎤
A
A 0 0
Δv 1 ν+1 ν ⎢ ⎥
= ln − ⎣0 −B 0 ⎦ A
V 2 ν − 1 ν2 − 1
0 0 −B
⎡ 2xy 2xz ⎤
(1 − μ2 )A − B − B
⎢ ν2 ν2 ⎥
⎢ ⎥
⎢ 2y 2 ⎥ A
+ 2
ν ⎢ − yx A − 2 B
2yz
− 2 B⎥
(ν − μ )(ν − 1) ⎢
2 2
⎢ ν2 ν −1 ν −1 ⎥ ⎥
⎣ zx 2zy 2z 2 ⎦
− 2A − 2 B − 2 B
ν ν −1 ν −1
Inside the ellipsoid (ν < ν0 , or r2 + τ 2 x2 < τ 2 ν02 ), the interference velocity is zero.
√
For the planetary (flat) ellipsoid, the length (minor axis) is = 2/ τ 2 − 1, where the thickness ratio τ is
greater than one. The ellipsoid is analyzed using coordinates ν and μ:
x = μν
r2 = y 2 + z 2 = (1 − μ2 )(ν 2 + 1)
or 5 2
1 − x2 − r2 1 − x2 − r2
ν =−
2
+ + x2
2 2
μ = x/ν
The coordinate ν ranges from ν0 = /2 on the surface of the ellipsoid, to infinity. With
% &−1
ν0
A= − cot−1 ν0
ν02 + 1
% &−1
ν02 + 2
B = cot−1 ν0 −
ν0 (ν02 + 1)
the interference velocity is:

⎡ ⎤
A
A 0 0
Δv ν ⎢ ⎥
= cot−1 ν − ⎣0 −B 0 ⎦ A
V ν2 + 1
0 0 −B
⎡ 2xy 2xz ⎤
−(1 − μ2 )A B B
⎢ ν2 ν2 ⎥
⎢ ⎥
ν ⎢ yx 2y 2 2yz ⎥ A
+ 2 ⎢ A B B ⎥
(ν + μ )(ν + 1) ⎢
2 2
⎢ ν2 2
ν +1 ν +1 ⎥
2
⎥
⎣ zx 2zy 2z 2 ⎦
A B B
ν2 ν2 + 1 ν2 + 1
Inside the ellipsoid (ν < ν0 , or r2 + τ 2 x2 < τ 2 ν02 ), the interference velocity is zero.
20-4.2.4 Airfoil-Shaped Body of Revolution
The induced velocity of an airfoil-shaped body of revolution is calculated using the modified slender-body
theory developed in reference 1. The length of the body is = 1. Slender-body theory approximates the axial
flow about the body by a line of sources on the axis of revolution, and the cross flow by a line of doublets. The
interference velocity is: ⎡ ⎤
A −Cy −Cz
Δv A
= ⎣ By E − Dy 2 −Dyz ⎦ A
V
Bz −Dzy E − Dz 2
where $
1 q(x − ξ)
A= dξ
4π s3
$
1 q
B= dξ
4π s3
$
3 μ(x − ξ)
C= dξ
4π s5
$
3 μ
D= dξ
4π s5
$
1 μ
E= dξ
4π s3
with s2 = (x − ξ)2 + r2 . Slender-body theory (ref. 3) gives limits of integration from −1/2 to 1/2, and the source
and doublet strengths as
qSB = S = 2πr0 r0
μ = 2S = 2πr02
where S is the cross-section area. The modifications developed in reference 1 (for zero angle of attack) include
(a) offset of the limits of integration from the nose and tail by a distance equal to one-half the radius of curvature,
so the stagnation points occur at the actual nose and tail; (b) multiplication of the source strength by a constant
K, to match the actual maximum thickness point; and (c) a tail correction for bodies with pointed tails. Hence the
source strength used here is:
q(ξ) = KqSB (ξ ∗ ) + f (ξ ∗ ) (ξ ∗ − 1) (1 − Kξ ∗ ) qSB

(1)
where
ξ − xN e
ξ∗ =
xT e − xN e
1
ξ∗ =
xT e − xN e

qSB (1) = 2π(r0 (1))2
with xT e = 1/2(1 − RT ) and xN e = − 1/2(1 − RN ) the limits of integration. Reference 1 gives the requirements
for the tail correction, which are satisfied by f (ξ ∗ ) piecewise linear from (ξ ∗ f, ξ ∗ ) = (0, 0) to (−a0 , ξ1 ) to (1, ξ2 )
to (1, 1). The requirement that the body be closed gives a0 as a function of ξ1 and ξ2 . The multiplicative constant
K needed to match the maximum thickness is obtained from the expression for the dividing streamline:
$
2 q(xm − ξ)
2πrm = dξ
((xm − ξ)2 + rm 2 )1/2
where q = Kc(ξ) + d(ξ) is given above; and (xm , rm ) is the maximum thickness point.
The NACA 4-digit airfoil thickness distribution gives the body radius for the configuration considered here:
√
r0 = 5τ (.2969 x − .1260x − .3516x2 + .2843x3 − .1036x4 )
with x = 0 to 1, and the maximum thickness at x = 0.3 (the last constant should be .1015, but is changed here so
the radius is zero at the tail). The radius of curvature of the nose is RN = (r0 r0 )N = 1/2a2 , where r ≈ ax1/2 near
x = 0 (a = 1.4845τ ). So the limits of integration are − 1/2(1 − 1.1019τ 2 ) to 1/2. The tail correction parameters
are ξ1 = 0.1, ξ2 = 0.95, and a0 = 0.4629 (ref. 1), so:
6
−4.628880ξ ∗ 0 ≤ ξ ∗ ≤ .10
∗
ξ f= 1.7221045ξ ∗ − 0.6349925 .10 ≤ ξ ∗ ≤ .95
1 .95 ≤ ξ ∗ ≤ 1
The integrals over ξ are evaluated by numerical integration, with about 100 steps required. The cross flow strength
and integration limits are not modified.
20–5 References
1) Yamauchi, G.K., and Johnson, W. “Development and Application of an Analysis of Axisymmetric Body Effects
on Helicopter Rotor Aerodynamics Using Modified Slender Body Theory.” NASA TM 85934, July 1984.
2) Durand, W.F., Editor-in-Chief. Aerodynamic Theory, Volume 1. Dover Publications Inc., New York, 1936.
3) Karamcheti, K. Principles of Ideal-Fluid Aerodynamics. John Wiley and Sons, New York, 1966.
table
airframe velocity V
perturbation velocity vA yes
calculated
airframe wing r collocation L/q

velocity V body r point r
perturbation velocity vA yes yes yes (a) yes
sensor Γ/V no no no yes
Notes:
a) yes only if collocation points of output and input
are same
Figure 20-4 Functionality of airframe flow field component.

Chapter 21
LIFTING LINE WING COMPONENT
21–1 Description
A lifting line wing component calculates the aerodynamic forces on a wing. The aerodynamic model is
lifting-line theory, using steady two-dimensional airfoil characteristics and a vortex wake. The model includes the
following features:
a) stall models (static, empirical dynamic stall model, none);

b) unsteady lift and moment for attached flow;
c) yawed and swept flow corrections, and spanwise drag;
d) drag and moment increments, zero lift angle increment;
e) Reynolds number corrections;
f) trailing-edge flaps;
g) prescribed lift, drag, and moment increments from an external aeroacoustic analysis.
The wake model is implemented in a separate component. This wing component has as input the wake induced
interference velocity (and any other interference terms in the velocity of the air).
The wing is a surface, which is described by a set of quarter chord and three-quarter chord collocation points
along the span. Arbitrary geometry of the wing is considered, including droop and sweep of the quarter-chord line,
and twist. The component interfaces are described in terms of the velocity and force normal to the wing surface
(discretized). Note that the geometry is actually calculated on the structural dynamic side of the interfaces with the
wing component. Figure 1 shows the relation between the geometry of the wing and the geometry of the structure.
Lifting-line theory deals with the section aerodynamic environment. The required section geometry (bent-chord
axes) is obtained from the geometry of the collocation points. The component analyzes the entire wing.
Lifting-line theory assumes that the wing has a high aspect ratio, or more generally that spanwise variations
of the aerodynamic environment are small. This assumption allows the problem to be split into separate wing and
wake models, which are solved individually and combined. This component solves the wing problem of lifting-line
theory. Viscous and compressibility effects are included by using experimental data for the two-dimensional airfoil
characteristics that are the foundation for the wing solution. Corrections for yawed and swept flow are introduced,
and an estimate of the spanwise drag. For low angles of attack, thin-airfoil-theory results are used to calculate the
unsteady loading. For high angles of attack, an empirical dynamic stall model can be used. For communication with
an external aeroacoustic analysis, prescribed increments in the section aerodynamic coefficients can be included.
Other components solve the wake problem of lifting-line theory. A vortex wake model is used to calculate
nonuniform induced velocity on the wing. Alternatively, an approximate wake model based on ideal-wing theory
194 LIFTING LINE WING COMPONENT
STRUCTURE = RIGID BODY

STRUCTURE
WING forces at
collocation points connection point
(aerodynamic
interfaces)
QC
QC
3QC
3QC
STRUCTURE = FEW BEAM SEGMENTS position and

extrapolated from velocity at
collocation points connection point
le (aerodynamic
interfaces)
QC
te 3QC
structural dynamic
interfaces
STRUCTURE = MANY BEAM SEGMENTS connecting
structural dynamic
components
Figure 21-1 Relation between wing geometry and structure geometry.

LIFTING LINE WING COMPONENT 195
can be used. The solution from such an approximate model is called “uniform inflow,” although the induced
velocity might still have a simple variation over the surface of the wing or wing set. With uniform inflow, the
three-dimensional flow effects at the wing tips are lost, requiring additional corrections in the wing model.
Several models are implemented for the unsteady aerodynamic loads in attached flow: from incompressible
thin-airfoil theory, from ONERA (EDLIN), and from Leishman-Beddoes. The unsteady loads model introduces
a degree of freedom vector (with elements for each span station) if the model includes the effects of the airfoil
near shed wake. If the vortex wake includes all the shed vorticity, or if a dynamic inflow model is used, then the
wing component can not also include the shed wake effects. Several dynamic stall models are implemented: from
Johnson, Boeing, Leishman-Beddoes, and ONERA (EDLIN and BH). In state variable form, the dynamic stall
model introduces a degree of freedom vector (with elements for each span station). In implicit form, the model
uses the stored history of the motion. The implicit form should not be linearized.
Figure 2 illustrates the functionality of the lifting line wing component.
Degrees of Freedom
The component can have one degree of freedom vector for the unsteady loads model, and one vector for the
dynamic stall model.
Component Input
a) Aerodynamic interfaces at collocation points:
1) Velocity of air, at quarter chord and three-quarter chord.
2) Velocity of air at quarter chord from three-quarter chord induced velocity (for second-order
lifting-line theory).
3) Position relative origin of wing frame, at quarter chord and three-quarter chord.
F
4) Induced velocity (vind ).
F
5) Interference velocity (vint ).
b) Controls: for use by trailing-edge flaps.
c) Aerodynamic loads (from this component).
Component Output
a) Aerodynamic loads: forces, circulation, power, and flap loads (FQC , F3QC or MQC , Γ, ∂Pi /∂v, Po , Fh , Mh ),
for all span stations.
b) Aerodynamic interfaces at collocation points:
1) Aerodynamic force, at quarter chord and three-quarter chord;
or aerodynamic force and moment, at quarter chord.
2) Aerodynamic force and moment, at trailing-edge flap hinge.
c) Bound circulation (for all span stations).
d) Total wing force and moment (in wing frame axes, about origin of wing frame).
e) Bound circulation peaks: magnitude (Γmax , ΓL , ΓR ) and position (rmax , rL , rR ) for entire wing and multiple
far wake panels; and centroid data (Γ, rC , rG , f , δ, set) for far wake panel edges.
f) Total wing power (Pind , Pint , Po ).
g) Sensors.
aero interfaces at collocation points aero interfaces at collocation points

degrees of
velocity of air (QC, 3QC) freedom aerodynamic force (QC, 3QC)
position rel wing frame (QC, 3QC) force and moment (flap hinge)
induced and interference velocity
aerodynamic loads
bound circulation
total wing force and moment
control bound circulation peaks
(for trailing edge flap) total wing power
sensors
Figure 21-2 Functionality of lifting line wing component.

21-3.1 Lifting-Line Theory
21-3.1.1 Basic Theory
The assumption of high aspect-ratio splits the three-dimensional wing aerodynamic problem (unsteady, com-
pressible, and viscous) into inner and outer problems — into wing and wake models (figure 3). The outer problem is
the wake: trailed and shed vorticity behind the lifting-line (bound vortex). The inner problem is a two-dimensional
airfoil, or more correctly an infinite wing in a uniform, yawed free stream. These problems are connected through
the wake-induced velocity and the bound circulation. The outer problem calculates the induced velocity at the
wing, from a wake with strength determined by the bound circulation. Note that the induced velocity is not
needed at an arbitrary point, just at the lifting-line. The inner problem calculates the bound circulation from the
aerodynamic environment, with the wake-induced velocity included in the free stream. Note that the pressure on
the wing is not needed, just the bound circulation (and the section lift, drag, and moment in order to calculate
performance and couple with the structural dynamics).
The assumption of high aspect-ratio has the following consequences, comparing the lifting-line problems to
the full solution. The inner problem has simpler geometry (two-dimensional) but complex flow (Navier-Stokes
equations). In the outer limit the inner solution can be considered irrotational. The outer problem has complex
geometry (the vortex wake) but irrotational flow. In the inner limit the outer solution has simple geometry. In the
matching domain there is both simple geometry and irrotational flow.
Uniform inflow from ideal-wing (momentum or vortex) theory is an approximation for the solution of the
outer problem (the wake). This approximation introduces effects requiring additional treatment, particularly the
tip loss factor.
21-3.1.2 Perturbation Theory
Formal lifting-line theory is the solution of the three-dimensional wing loading problem using the method
of matched asymptotic expansions. Based on the assumption of large wing aspect-ratio, the problem is split into
separate outer (wake) and inner (wing) problems, which are solved individually and then combined through a
matching procedure. In general it is necessary to consider wings with swept and yawed planforms, in unsteady,
compressible, and viscous flow.
The lowest-order solution is Prandtl’s theory (steady and no sweep). Development of higher-order lifting-line
theory originated with Weissinger for intuitive methods, and with van Dyke for singular perturbation methods.
The lifting-line theory developments found in the literature, although including higher orders, unsteady, transonic,
and swept flow, are generally analytical methods. They obtain analytical solutions for both the inner and outer
problems, and are in quadrature rather than integral-equation form. Often the inner solution is inviscid, or even
a thin airfoil. These theories are not therefore directly applicable to the general case, but do provide a guide and
sound mathematical foundation for the development.
In general, it is necessary to include stall (high angles of attack) in the inner solution, and the distorted,
rolled-up wake geometry in the outer solution. Hence the objective is to obtain from lifting-line theory a separate
formulation of the inner and outer problems, with numerical not analytical solutions, and a matching procedure
OUTER
PROBLEM
e
g lin
liftin
wake
free stream
velocity
ASSUME
wing wake
bound
induced
circulation
wake HIGH velocity
ASPECT
RATIO MATCHING
yawed free
stream
g
win
nite
infi
INNER
PROBLEM
Figure 21-3 Lifting line theory.

that will be the basis for an iterative solution. A key consideration is the need to retain the two-dimensional airfoil
tables in the inner solution, for the viscous effects. Hence whatever approximations that are required to retain
the tables will be accepted. Furthermore, a practical method is needed, one that gives good accuracy without
convergence problems or singularities.
21-3.1.3 Second-Order Lifting-Line Theory
Higher order theory is considered in order to improve the calculation of the airloads without actually resorting
to lifting-surface or more advanced methods, which require more computation and still need development for some
applications. Several investigations have shown that second-order lifting-line theory gives nearly the same results
as lifting-surface theory, including the lift produced in close wing-vortex interactions. In addition, second-order
theory should also improve the loads calculations for swept tips, yawed flow, and low aspect-ratio wings.
In the second-order outer problem (the wake), the wing is a dipole line plus a quadrupole line, which for a thin
airfoil (with no thickness or camber) is equivalent to a dipole at the quarter chord. The dipole solution is a wake
of vortex sheets. In the second-order inner problem (the wing), the boundary condition is a wake-induced velocity
that varies linearly in space. For a thin airfoil, the same lift is obtained with a uniform induced velocity, by using
the value at the three-quarter chord. The correct moment is not obtained however, since a linear induced-velocity
variation over the chord produces a moment about the quarter chord, but a uniform induced-velocity does not. In
general, it is necessary to define a remainder solution that is the second-order solution after accounting for the
induced-velocity at the three-quarter chord. This remainder solution is ignored in practice, so it becomes an error
estimate. The lift error is found to be small, and the moment error also except for the moment about the quarter
chord produced by a linearly varying induced-velocity. In order to retain use of the airfoil tables, the only parts of
the second-order theory that can be used are placing the lifting-line at the quarter chord and the collocation point
at the three-quarter chord.
21-3.1.4 Integral Equation Formulation
The perturbation solution procedure alternates between the inner and outer problems, using only the solution up
to the previous order. Combining and solving all orders simultaneously is equivalent in terms of the perturbation
expansions. It is natural to combine the inner problems. The lowest-order inner problem is the airfoil with
just geometry boundary conditions, and the first order inner problem has the wake-induced velocity boundary
conditions. These are easily combined since the wake just gives an angle of attack change.
Combining the outer problems means evaluating the induced-velocity using the total bound circulation, from
the combined inner problem rather than just from the lowest-order inner problem. This changes the nature of the
solution, from direct quadrature to an integral equation. It is necessary to invert the integral equation, but with
airfoil tables an iterative solution is required anyway. Moreover, it is found that the solution of the integral equation
is well behaved, while the direct solution is singular at the wing tips for normal wing planforms.
21-3.1.5 Comparison of Formulations
A consideration of the simplified inner problem will show the differences between the common formulations
of lifting-line theory. Let θ be the geometric angle of attack of the wing section. The inner problem solution (static,
with no sweep) can be analytical or numerical:

Γ
analytical, thin airfoil: = Uθ − v
πc
Γ U
numerical, airfoil tables: = c (θ − v/U )
πc 2π
where Γ is the bound circulation, c the chord, U the free stream velocity, c (α) the lift-coefficient as a function
of angle of attack, and v the wake-induced velocity. The outer problem can obtain the induced velocity v at the
quarter chord or three-quarter chord, by integrating the effects of all wake vorticity (excluding the bound vortex).
For example, a thin planar wake gives: $
1 dΓ dη
vc/4 =
4π dη y − η
The following implementations of lifting-line theory are of interest.
a) First-order perturbation theory obtains v from Γ0 = πcU θ:

Γ
= U θ − vc/4 (θ)
πc
b) Prandtl’s integral equation is first-order, but obtains v from Γ:
Γ
= U θ − vc/4 (Γ)
πc
c) A common implementation for wings is the first-order theory, but using airfoil tables:
Γ U
= c (θ − vc/4 (Γ)/U )
πc 2π
d) A second-order implementation uses v at the three-quarter chord:
Γ U
= c (θ − v3c/4 (Γ)/U )
πc 2π
e) Weissinger’s L-theory includes the contribution of the bound vortex in evaluating v at the
three-quarter chord, and equates the induced angle of attack to the geometric angle of attack:

Γ
+ v3c/4 (Γ) = U θ
πc
This is equivalent to using the thin-airfoil solution of the inner problem in second-order theory.
The second-order theory (d) is implemented here.
21-3.1.6 Sweep and Yawed Flow
Large sweep angles can be included in the second-order theory, but it is necessary to assume small curvature
so the wake-induced velocity effects in the inner problem remain two-dimensional. For many wings the curvature
is small, except at kinks in the quarter-chord line. The analysis relies on the integral-equation form and spanwise
discretization to keep the loading well behaved at such kinks.
21-3.1.7 Unsteady Loading
With unsteady motion and loading, the inner problem is an unsteady, two-dimensional airfoil with a shed
wake, and the outer problem excludes both the bound vortex and the inner shed wake when calculating the induced
velocity. It is natural to retain the shed wake in the outer rather than the inner problem, so the shed wake and
trailed wake can be treated identically, especially since subtracting the inner shed wake from the outer problem is
difficult with complex wake geometry. Then the induced velocity from all vorticity (except the bound vortex still)
is evaluated at the three-quarter chord, and treated as a uniform flow for the inner problem. The shed wake is thus
a boundary condition of the inner solution, not a part of the inner model.
The assumption that the shed wake-induced velocity is constant over the chord is a major approximation.
With the induced velocity evaluated at a single point, the shed wake model must be modified to obtain the unsteady
loads correctly. It is found that the Theodorsen and Sears functions (the shed wake effects in two-dimensional
airfoil theory) are well approximated for low frequency (and reasonably well up to reduced frequencies of about
1.0) if the shed wake in the outer problem is created a quarter chord aft of the collocation point, not at the bound
vortex as for the trailed wake.
21-3.2 Practical Implementation of Lifting-Line Theory
Guided by the results of perturbation theory, the following is a practical implementation of lifting-line theory.
The outer problem is an incompressible vortex wake behind a lifting-line, with distorted geometry and rollup.
The lifting-line (bound vortex) is at the quarter chord, as an approximation for the quadrupole line introduced by
second-order loading. The trailed wake begins at the bound vortex. The shed wake is created a quarter chord aft
of the collocation point on the wing (the lifting-line approximation for unsteady loading). The three components
of wake-induced velocity are evaluated at the collocation points, excluding the contributions of the bound vortex.
The collocation points are at the three-quarter chord (in the direction of the local flow), as an approximation
for a linearly varying induced velocity introduced by the second-order wake. The induced velocity calculated at
the three-quarter chord is used only to calculate the angle of attack for the loading solution. The local section
aerodynamic environment, including the orientation of the lift and drag and hence the magnitude of the induced
power, is still obtained from the induced velocity at the quarter chord.
The inner problem consists of unsteady, compressible, viscous flow about an infinite aspect-ratio wing,
in a uniform flow consisting of the yawed free stream and three components of wake-induced velocity. This
problem is split into parts: two-dimensional, steady, compressible, viscous flow (airfoil tables), plus corrections.
The corrections account for unsteady flow (small angle-of-attack noncirculatory loads, without any shed wake);
dynamic stall (an empirical model); swept and yawed flow (the equivalence assumption for a swept wing); tip
flow; and perhaps Reynolds number.
This formulation is generally second-order (in the inverse of the aspect-ratio) accurate for lift, including
the effects of sweep and yaw, but less accurate for section moments (basically still first-order). In particular,
with typical wing-vortex separations, second-order lifting-line theory is as accurate as lifting-surface theory for
vortex-induced lift calculations.
Tip flow corrections include a tip loss factor, compressible tip relief (a small reduction in effective Mach
number), and the effects of a swept tip (discussed below). The tip loss factor is needed when the induced velocity
is obtained from a uniform inflow model (an approximation for the wake solution). For nonuniform inflow a tip
loss factor is not appropriate, but it is necessary to consider the span station of the rolled-up tip vortex when it
reaches the trailing edge of the wing; the implementation is similar to a tip loss factor.
Yawed and swept flow require significant aerodynamic corrections. The inner problem is an infinite wing
with yaw and sweep (the planform is defined relative to a wing reference line, so spanwise velocity produces yawed
flow, while sweep is obtained from the locus of the quarter chord relative to the wing reference line). The section
loads are obtained from airfoil tables that give the solution for a wing with no yaw or sweep, using the equivalence
assumption for swept wings. The objective is to derive equivalent angle of attack and Mach number for evaluating
the coefficients, and corrections for the coefficients from the tables (and correctly account for the chord and wing
area when multiplying the coefficients by chord, dynamic pressure, and panel width in order to obtain the section
loads). The principal assumptions are:
a) The lift-curve slope of normal section is not affected by spanwise flow.

b) The Mach number normal to the (swept) quarter chord defines compressibility effects.
c) The angle of attack and chord in the local flow direction define the drag and stall.
d) The total viscous drag force (vector addition of spanwise and chordwise components) has
the same direction as the local yawed flow, so the spanwise drag component can be obtained
from the section drag coefficient.
Corrections for the effective shape and thickness of the airfoil in yawed flow are seldom available, and none are
implemented here. Probably the airfoil tables should correspond to the shape of the cross-section in the local flow
direction (or the mean direction for time-varying flow). Hence frequently the shape perpendicular to the wing
reference line is appropriate. If the yaw angle is not too large, the absence of any correction for effective shape
and thickness does not appear to produce any worse problem than the other approximations of the model. Note
that if the yaw angles are large, then first-order lifting-line theory is not applicable.
Unsteady motion of an airfoil produces a shed wake because of the variation of the bound circulation with
time. This wake vorticity produces an induced velocity at the airfoil that tends to cancel the lift, so the shed wake
produces a feedback reducing the circulatory lift. In addition there are noncirculatory lift and moment terms,
independent of the wake effects, including rate (damping) and acceleration (virtual mass) effects. The present
model requires from unsteady theory the noncirculatory loads, since the shed wake effects are usually accounted
for through the wake-induced velocity, and the static loads are obtained from airfoil tables. The noncirculatory
terms are essential for the aerodynamic pitch damping, and sometimes the lift and even the virtual mass terms are
important. The noncirculatory lift and moment are obtained from thin-airfoil theory, including the effects of a
time-varying free stream and reverse flow.
Dynamic stall is characterized by a delay in the occurrence of separated flow produced by the wing motion,
and high transient loads induced by a vortex shed from the leading edge when stall does occur. As implemented,
the dynamic stall models still use the airfoil table for steady characteristics, evaluated at an angle of attack that
includes the dynamic stall delay.
The analysis is performed in the axes of the wing frame F. The wing frame should be the frame or parent
frame of all structural dynamic components composing the wing. A wing reference line is used to define the wing
geometry. This line can have arbitrary curvature and torsion, but a straight wing reference line is simplest and
frequently sufficient. Section properties are measured in planes perpendicular to the wing reference line. Figure
4 shows the geometry of the wing model, for the case of a straight wing reference line. The wing consists of
a set of aerodynamic panels, with a quarter chord and three-quarter chord collocation point at each panel. It is
edges of
wing section axes: aerodynamic
chord vector c panels
tangent vector t
normal vector n = c x t panel area =
c(R Δr)
straight wing reference axis

r
QC and 3QC collocation

points defined in plane
perpendicular to reference
axis
position of collocation points equivalent to locus of QC,

from origin of wing frame chord, and pitch
(calculated by structural (measured in that plane)
dynamic components)
Figure 21-4 Geometry for definition and analysis of wing.

assumed that the wing position and velocity relative the air (component input) are in the F axes, and the calculated
forces (component output) are thus also in the F axes. The required motion relative the air includes the following
quantities at each collocation point:
a) velocity v F ;
b) dynamic pressure q;
c) rate of change of velocity v̇ F .
This velocity includes interference terms. The motion relative the air is required at the quarter chord and three-
quarter chord collocation points; and at the quarter chord using the three-quarter chord induced velocity, for
second-order lifting-line theory. The quarter chord interference velocities are also input to this component (and
separated according to the definitions of “induced” and “interference” as described below), for use in calculating
the wing power. The required location of the wing includes the following quantities at each collocation point:
a) position rF ;
b) rate of change of position ṙF .
This position is measured from the origin of the wing frame F. The input induced and interference velocities
must also be in the wing frame axes. The required aerodynamic forces are F F at each collocation point. These
discretized section loads are summed to obtain the total force and moment of the wing: FwF and MwF (about the
origin of F, using the position vectors rF ).
The section properties, including chord and quarter-chord position, are measured in planes perpendicular to
the wing reference line. The wing can have arbitrary sweep and droop, defined in terms of offsets of the quarter
chord from the wing reference line. Sweep is the angle between the quarter-chord line and the wing reference line.
This angle is implemented as an input parameter, rather than being calculated from the quarter-chord position. The
wing can have arbitrary twist, defined by the positions of the three-quarter chord and quarter chord points in the
section plane. Note that the definition of the geometry in terms of properties measured perpendicular to the wing
reference line has nothing to do with the airfoil geometry used in the lifting-line analysis. The latter information
is contained in the airfoil tables assigned to the section.
The wing geometry is described in terms of a span station r, which is a variable (often dimensionless) that
identifies positions on the wing reference line. This variable must be used consistently by the component, including
its definition in the airfoil table. Positive r can run from left tip to right tip, or from right tip to left tip. There
is a span scale factor R such that Rr is a true measure of distance along the wing reference line; and (R Δr c)
is the panel area (scale factor times panel width times chord). In general R varies along the span. It is assumed
that this scale factor defined in terms of the wing area can be used to perform all spanwise integrations of section
properties.
The wing consists of K aerodynamic panels, aligned spanwise. The panels are defined by the span stations
of the edges: rEk , for k = 1 to K + 1. The wing tips are thus at span stations rE1 and rE(K+1) . The collocation
points (quarter chord and three-quarter chord) are at the panel midpoints:
1
rAk = (rE(k+1) + rEk )
2
for k = 1 to K. The corresponding panel width (dimensional) is
ΔrAk = (rE(k+1) − rEk ) R

If rAk were specified directly, it could not be assured that there is a set of panel edges with rAk as the midpoints.
The panel edges are also used to define the wing wake. The panel width is used to discretize the loads:
(panel load)k = (section load)k ΔrAk
and to integrate over the span: $

f (r)R dr = f (rAk ) ΔrAk
span
(assuming that Rr is a true measure of distance along the span). The aerodynamic properties of the wing are
defined parametrically, in terms of values at an arbitrary set of span stations. The properties are evaluated at the
panel midpoints (and wherever else they are required) assuming linear variation between the specified points. The
wing chord ck is required such that ΔrAk ck is the true area of the panel. Hence the chord value used should be
the mean over the panel: $ rE(k+1)
1
ck = c(r)R dr
ΔrAk rEk
which can be evaluated analytically from the input data. If the rate of change of the chord does not vary inside the
panel, then linear interpolation gives the same result.
21-3.4 Trailing-Edge Flaps
The wing can have one or more trailing-edge flaps, defined by the span stations of the left and right flap edges.
Each flap extends over one or more aerodynamic panels. It is assumed that if a flap covers the midpoint of a panel,
then the flap extends over the entire panel. So panel edges should be placed at the flap edges. For each flap, the
deflection angle φ and its derivatives are identified as elements of some control vector (component input). The flap
chord is c f , where c is the wing section chord. The flap hinge is a distance cxh aft of the section leading edge
and czh above the chord line. An aerodynamically balanced flap has the flap leading edge forward of the hinge. It
is assumed that the flap leading edge is not aft of the hinge (xh ≥ 1 − f ).
If a wing section has a trailing-edge flap, the quarter chord and three-quarter chord collocation points define
the motion of the wing forward of the flap. Chordwise flexibility of the wing section is generally neglected when
using lifting-line theory, so these collocation points can be attached to a rigid extension of the chord line, even if
the quarter chord or three-quarter chord is aft of the flap leading edge. The motion of the flap relative the chord
line is defined by the flap angle φ (positive for increased section lift).
21-3.5 Airfoil Table
The component obtains the static, two-dimensional airfoil data from a table, of class = AIRFOIL and type
= CAMRAD. This is a three-dimensional table, containing lift, drag, and moment coefficient data as a function
of angle of attack, Mach number, and span station. The table data are interpolated in α and M . The table can
be searched or interpolated spanwise (as specified in the table definition). Wings are often built with spanwise
transition regions, in which the geometry is linearly interpolated between specified airfoils at two span stations. It
is easy to accommodate such cases by using spanwise interpolation (linear) of the airfoil characteristics, although
there is no theoretical justification of such an approach in the presence of nonlinear aerodynamic phenomena. In
any case, spanwise interpolation of the table can produce anomalous results, particularly in the stall region, so the
table data should be examined at the interpolated stations before they are used. Best results will be obtained if the
two-dimensional tables all have the same zero lift angle (offsets can be moved to the definition of aerodynamic
twist).
If the wing has trailing-edge flaps, then the airfoil table (class = AIRFOIL and type = CAMRAD) includes
the effects of the flaps. So the coefficients are functions of flap angle φ as well as angle of attack, Mach number,
and span station; and the table includes coefficients of lift, drag, and moment acting on the flap (c f , cdf , and
cmf ). The coefficients c , cd , and cm are the total loads on the airfoil, including the flap loads. Thus this is a
four-dimensional table; the analysis first interpolates over α and M , then φ, and finally span station. In order to
use spanwise interpolation on a wing with trailing-edge flaps, there should be tables at the span stations of the
flap edges. Note that in the context of a trailing-edge flap, “total” refers to the loads acting on the entire section,
including the flap loads. Otherwise, the “total coefficient” is the sum of all terms that are used in the model, such
as static, dynamic stall, and unsteady contributions.
The airfoil table data are defined in terms of some reference chord in the section, which defines how angle
of attack is measured; and some reference axis, which defines how the moment is measured. With a trailing-edge
flap, there is also some reference axis that defines how the flap moment is measured. These reference axes are on
the chord line. The wing section geometry is defined by the positions of the quarter chord and three-quarter chord
points. These points define a chord line, hence a pitch or twist angle measured in the wing frame. The orientation
of the airfoil relative this geometry is specified by the following parameters:
a) θAF : the angle that the reference chord of the airfoil coefficients is above the chord line
defined by the structural dynamic interfaces.
b) xAF : the distance that the reference axis of the airfoil coefficients (total loads) is aft of the
leading edge (fraction of chord).
c) xAF f : the distance that the reference axis of the flap coefficients is aft of the leading edge
(fraction of chord).
Usually the reference axis for the total airfoil loads is the quarter chord (xAF = .25). Often the aerodynamic twist
is defined in terms of the airfoil table reference chord (θAF = 0). Typically the reference axis for the flap loads
is the flap hinge (xAF f = xh ). This reference axis is on the reference chord line, although the flap hinge may
not be. The position of the flap (xh and zh ) is specified relative the chord line defined by the structural dynamic
interfaces. The definition of the flap angle must also be consistent.
21-3.6 Induced and Interference Division
The power and the interference velocity are divided into induced and interference terms, with the following
definition intended:
a) Induced: from wings in this wing set.

b) Interference: from all other sources.
This convention must be used when constructing the system, for proper interpretation of the output quantities.
The induced and interference velocities (component input) must be the same as are used by the structural dynamic
components to calculate the velocity relative the air at the collocation points.
21-3.7 Controls
to a system input piece or to an input/output interface. The component can use the controls for the trailing-edge
flaps. A particular control vector and element can be used once, more than once, or not at all.
21-3.8 Sensors
component input; component output; section environment (section axes); section environment (wing plane axes);
unsteady motion; airfoil table parameters; section coefficients; wing axis forces; forces and flap forces; power.
The sensor vector consists of the quantity values at all span stations, plus the total. The definition of the total
depends on the quantity y. For the discretized forces at the collocation points, the total is the sum over the span:
ytotal = yk
(discretization multiplies the section load by ΔrAk ). For force, moment, or power per unit length, the total is the
integral over the span: $
ytotal = yR dr = yk ΔrAk = y b
where b is the wing span. For all other quantities, the total is the average over the span:

yk ΔrAk
ytotal = y =
ΔrAk
(weighted by the aerodynamic panel width). The value of any sensor can be multiplied by a scale factor. Automatic
scaling produces as appropriate a value in g’s, degrees, Mach number, or horsepower. Automatic scaling can also
be applied to produce values in coefficient form: section coefficients, section coefficients times a power of Mach
number, wing coefficients, or rotor coefficients divided by solidity. The sensor coefficient scales are based on the
air density ρ; a velocity scale v; and a length scale :
coefficient form velocity scale v length scale

section coefficient section velocity U section chord c
M n × section coefficient speed of sound cs section chord c
wing coefficient reference velocity Vref reference chord cref
rotor coefficient/solidity tip speed Ωref Rref reference radius Rref
The time scale is τ = /v. Here U and c are the calculated values for the section; Vref , cref , and Rref are input
parameters; and the rotor rotation speed Ωref is obtained from a specified period. Then the sensor scales used are
as follows:
quantity rotor coefficient/solidity other coefficients
angle deg deg
angular rate deg/τ deg/τ
angular acceleration deg/τ 2 deg/τ 2
displacement
velocity /τ /τ
acceleration /τ 2 /τ 2
time τ τ
circulation v v
force ρv 2 cref Rref 1/2ρv 2 b
moment ρv 2 cref Rref 1/2ρv 2 2 b
force/length ρv 2 cref 1/2ρv 2
moment/length ρv 2 cref 1/2ρv 2 2
power/length ρv 3 cref 1/2ρv 3

Note that the force, moment, and power scales for the rotor coefficient/solidity form are obtained by multiplying
the standard scales by (2cref / ), where the input reference chord provides the solidity (σ = N cref /πRref ). The
span integrals (of force, moment, or power per unit length) are further scaled with the wing span b (Rref for the
rotor coefficient/solidity form).
21-3.9 External Aeroacoustic Analysis
CAMRAD II can be used with an external aeroacoustic analysis. The external analysis typically employs the
methods of computational fluid dynamics, possibly with a limited computational domain that does not encompass
the entire wing wake. The external analysis may also not calculate the effects of the structural dynamic motion
on the aerodynamic boundary conditions. Figure 5 illustrates the communication between CAMRAD II and an
external aeroacoustic analysis. The basic procedure, known as loose coupling, has the following steps:
a) Run CAMRAD II to obtain the converged trim solution; save the wing loads (lift, drag,
and moment, including any prescribed increments) as ctotal ; and calculate the wing motion
using the appropriate structural dynamic sensors. Depending on the requirements of the
external analysis, may also calculate a partial angle-of-attack αP at the wing. The wake-
induced velocity used to evaluate αP is calculated excluding the wake vorticity that is in the
computational domain of the external analysis; and αP includes the effects of the structural
motion.
b) Run the external aeroacoustic analysis, using the wing motion (or αP ) to calculate the wing
loads cext . Calculate the loading increments (Δc)n+1 = (Δc)n + (cext − ctotal )n .
c) Run CAMRAD II to obtain the converged trim solution, using the prescribed loading
increments Δc.
d) Repeat the procedure as required to obtain a consistent solution from CAMRAD II and
the external analysis. The procedure has converged when the wing loading calculated by
CAMRAD II equals the wing loading calculated by the external analysis (so Δc would be the
same in the next cycle).
Often the external analysis requires a higher resolution in span, time, and wake age than CAMRAD II needs
to obtain a converged trim solution for the coupled aerodynamic and structural dynamic behavior. Hence it is
appropriate to calculate high resolution quantities in a separate loop in the trim task, using fixed wake geometry
and structural motion (and fixed circulation if the partial angle-of-attack is needed), after the converged solution
for the structural dynamic response has been obtained.
If the external aeroacoustic analysis accounts for the entire flow field, then it requires just the wing motion
from CAMRAD II. Otherwise, the communication between CAMRAD II and the external aeroacoustic analysis
is based on the partial angle-of-attack, so that the wake in the computational domain of the external analysis is not
used twice. The external analysis may account for just the wake vorticity directly behind the wing; for tip vortices
from other wings as well; or for all vorticity inside the computational domain. A wing component can then use
this induced velocity to calculate αP .
Simply setting the loads in the wing component to the values calculated by the external analysis would
not account for changes in the angle-of-attack as the wing motion and wake effects are updated. Therefore the
CAMRAD II
blade total loads

cl total cdtotal cmtotal prescribed
motion load
(including
increments
(perhaps increments) Δcl Δcd Δcm
partial
angle of
attack) Δcn+1 = Δcn + (cext – ctotal )
external loads cl ext cd ext cm ext
external aeroacoustic analysis
Figure 21-5 Communication with an external aeroacoustic analysis (loose coupling).

component uses the following expression for the lift coefficient:

c total (α) =c ext (αold ) + c α (α − αold )
=c + c int (α) − c old (αold )
ext (αold )

= c int (α) + c ext (αold ) − c old (αold )
=c int + Δc
where c ext is the lift obtained from the external analysis; c int is the lift calculated by this component; and c old
is the lift calculated by this component in the previous cycle (when αP was evaluated). Similar expressions are
used for the drag and moment coefficients. Hence communication with an external aeroacoustic analysis can
be implemented by introducing prescribed increments in the section aerodynamic coefficients: Δc , Δcd , Δcm .
These increments can be written
(Δc)n+1 = (cext )n − (cold )n = (cext )n − (ctotal − Δc)n = (Δc)n + (cext − ctotal )n
So for each iteration the increment is updated using the difference between the results from the external aeroacoustic
analysis and from CAMRAD II. The increment can be specified as Δ(M 2 c) instead of Δc, perhaps to avoid
difficulties at sections with small Mach number. Alternatively, the increment can be specified as the discretized
forces on the section.
It is essential to be consistent in the definition of internally and externally calculated loads. In particular, c old
must include or exclude the angle-of-attack rate terms if c ext from the external analysis does. Let c US (α̇) be the
rate term in c int (α, α̇). Then if the external analysis includes the angle-of-attack rate in its boundary conditions:
c (α, α̇) = c ext (αold , α̇old ) +c int (α, α̇) −c old (αold , α̇old )

=c int (α, α̇) + c ext (αold , α̇old ) −c old (αold , α̇old )
=c int + Δc
while if the external analysis does not include the angle-of-attack rate in its boundary conditions:
c (α, α̇) = [c ext (αold , 0) + c int (α, 0) − c old (αold , 0)] + c US (α̇)

= c int (α, 0) + c US (α̇) + c ext (αold , 0) − c old (αold , 0)
=c int + Δc
The above equations are representative of the approach. In general, α refers to the boundary condition defined by
the uniform term in the velocity distribution over the chord; this boundary condition might actually be obtained
from the velocities at the wing surface. In general, α̇ refers to the linear term in the velocity distribution over the
chord (the effective pitch rate). The aerodynamic sensors of the component provide the quantities required for
the boundary conditions in the external analysis; as well as the wing coefficients for c old (total with or without
unsteady terms, but excluding the prescribed increments).
21–4 Theory
21-4.1 Wing Geometry
21-4.1.1 Wing Reference Line
A wing reference line is used to define the wing geometry. This line can have arbitrary curvature and torsion.
Section properties are measured in planes perpendicular to the wing reference line: the chord, quarter chord locus,
and three-quarter chord locus. Positions along the reference line are identified by a span station variable r. A span
scale factor R is defined such that Rr is a true measure of distance along the wing reference line; and (R Δr c) is
the panel area. The scale factor R is used to perform all spanwise integrations of section properties. This section
provides background information to assist in selecting a reference line and determining the span scale factor.
Consider an arbitrary reference line, quarter chord locus, and three-quarter chord locus, defined by position
vectors as a function of span station r: y, yQ , and y3Q respectively. Let s be the arc length along the reference
line, so ds/dr = |y |. Then the tangent, normal, and binormal vectors of the reference line are:
dy y
u= =
ds |y |
du u
n=ρ =ρ
ds |y |
y × y
b=u×n=
|y × y |
With ρ the radius of curvature and τ the radius of torsion of the reference line, these vectors satisfy:
du 1
= n
ds ρ
dn 1 1
= b− u
ds τ ρ
db 1
=− n
ds τ
The wing geometry is specified in terms of the reference line:
yQ (r) = y + zQ
y3Q (r) = y + z3Q = yQ + zC
where c = 2|zC | is the section chord. The analysis section is perpendicular to the reference line, so:
zQ = ξQ n + ηQ b
zC = ξC n + ηC b
The wing surface then is given by the position vector x as a function of r and a chordwise coordinate σ = − 1/2
to 3/2:
x = yQ + σzC = y + zQ + σzC
The wing area is calculated from the vectors ur and uσ on the wing surface:
$$ 3/2 $
A= |ur × uσ | dσ dr = cR dr
−1/2
Hence the span scale factor R is obtained from:
$ 3/2
cR = |ur × uσ | dσ
−1/2
For the general wing geometry as described here, the surface vectors are as follows.
∂x
ur = = y + zQ + σzC
∂r
∂x
uσ = = zC
∂σ
y = |y |u = Y u
zQ = (ξQ

n + ηQ b) + (ξQ n + ηQ b ) = Qu u + Qn n + Qb b
zC = (ξC

n + ηC b) + (ξC n + ηC b ) = Cu u + Cn n + Cb b
|ur |2 = |y + zQ + σzC |2 = (Y + Qu + σCu )2 + (Qn + σCn )2 + (Qb + σCb )2

|uσ |2 = |zC |2 = (c/2)2
2 2
(ur · uσ )2 = (y + zQ + σzC ) · zC = (Qn + σCn )ξC + (Qb + σCb )ηC

|ur × uσ | = |ur |2 |uσ |2 − (ur · uσ )2 = a + bσ + cσ 2
Consider the geometry for the special case of the wing reference line being a curve lying in a plane. Such a curve
has no torsion, so b = 0 and n ∼ u; and
y = Y u
zQ = ξQ

n + ηQ b + ξQ n
zC = ξC

n + ηC b + ξC n
If in addition the reference line follows the wing droop, then ξQ = 0; and
zQ = ηQ

b = Qb b
With no droop, the reference line is then straight. If also the wing pitch is zero, then ξC = 0 and ηC = c/2; and
zC = ηC

b = Cb b
So the span scale factor is:

|ur |2 = Y 2 + (Qb + σCb )2
(ur · uσ )2 = (Qb + σCb )2 (c/2)2

|ur × uσ | = Y 2 (c/2)2 = |Y |c/2

ds
R = |Y | =
dr
Alternatively, consider the geometry for the special case of a straight wing reference line. A straight line has no
torsion or curvature, so normal and binormal axes can follow the wing twist θ:
⎛ ⎞ ⎛ ⎞
cos θ sin θ
b=⎝ 0 ⎠ n=⎝ 0 ⎠
− sin θ cos θ
b = −θ n n = θ b
Thus:
y = Y u
zQ = (ξQ
− θ ηQ )n + (ηQ
+ θ ξQ )b
c
zC = ηC b = b
2
c c
zC = b − θ n
2 2
4 2
c − θ η − σθ c
|ur × uσ | = Y 2 + ξC Q
2 2
Integrating over the chord then gives R. Note that twist increases the wing area. Neglecting the effect of twist, the
span scale factor is:
R= 2
Y 2 + ξQ

where ξQ gives the effect of droop on the span scale factor, when a straight reference line is used.
21-4.1.2 Mean Chord
The wing chord is defined parametrically, in terms of values at an arbitrary set of span stations: ci at ri , for
i = 1 to I. The chord is evaluated assuming linear variation between the specified points. The mean chord over
an aerodynamic panel or over the entire wing can be evaluated analytically from these input data. Consider the
mean chord over a segment of the span extending from rL to rR :
$ rR
1
c= c dr
rR − r L rL
assuming a constant span scale factor. Find j and k such that rj−1 ≤ rL < rj and rk < rR ≤ rk+1 . Then
6 3 rR
rL
c dr if rj > rk
(rR − rL ) c = 3 rj k−1 3 ri+1 3 rR
rL
c dr + i=j ri
c dr + rk
c dr if rj ≤ rk
Each of these integrals is over one segment of the input data, hence has linear variation of the chord. Hence:
$ rb
Δr
c dr = (cb + ca )
ra 2
with Δr = rb − ra ; and the chord is evaluated at the limits of integration by linear interpolation as required. With
a scale factor R that varies over the span, the mean chord is:
3 rR
cR dr
c = 3rLrR
rL
R dr
The integrals over each segment of the input data are then:
$ rb
Δr
cR dr = cb (2Rb + Ra ) + ca (2Ra + Rb )
r 6
$a rb
Δr
R dr = R b + Ra
ra 2
since R is specified at the same span stations as the chord.

21-4.1.3 Section Axes
The wing geometry is defined by the position vectors xk at the quarter chord and three-quarter chord collocation
points, for aerodynamics panels at span stations rk , k = 1 to K. The position is measured from the origin of the
wing frame F, in F axes. Figure 6 illustrates the geometry. The order of the collocation points from the left tip
to the right tip defines the upper side of the wing surface (if the collocation points are specified from right to left
in the input, their order is reversed for the purposes of this section). The position includes the effects of all wing
motion. Any change in wing area or chord because of elastic deflection is ignored. The aerodynamic analysis is
performed in section axes X, defined by the chord, normal, and tangent vectors shown in figure 6 (c, n, and t). The
section plane is thus c–n, and the wing plane is c–t. The chord vector is
ck = x3QCk − xQCk
The vectors in the wing plane are

ak = xQC(k+1) − xQC(k−1)
bk = x3QC(k+1) − x3QC(k−1)
(the position at k is used for panels at the wing tips, where k − 1 or k + 1 does not exist). The normal vectors are
then
nQCk = ck ak
n3QCk =
ck bk
The section axes are defined using the quarter-chord normal, nk = nQCk , and finally the tangent vector is
k ck . Thus the rotation matrix from the wing frame F to the section axes X is:
tk = n
⎡ ⎤
Tc
⎢ ⎥
C XF = ⎣ Tt ⎦
Tn
where is the appropriate unit vector. With this definition, the X axes have x chordwise (positive aft), y spanwise,
and z normal (positive up). The time derivatives of the vectors can be evaluated as required, using the time
derivatives of the wing position, ẋk . Thus
d c ċ 1
˙c = = I − c Tc = ċ − c (Tc ċ)
dt |c| |c| |c|
d n ṅ 1
˙n = = I − n Tn = ṅ − n (Tn ṅ)
dt |n| |n| |n|
are the time derivatives of the chord and normal unit vectors.
21-4.1.4 Wing Plane Axes
It is more conventional to use wing plane axes W, which have the chordwise axis in the x–y plane of the wing
frame F. Such axes can not always be found, so the chord-normal section axes described above are more general.
The wing aerodynamics are analyzed using the section axes X. The wing plane axes W are available for sensor
quantities. Wing plane axes are obtained by rotating the section axes by pitch angle −θ about the spanwise axis
(t) until the chordwise axis (c) is in the x–y plane of the wing frame F. Thus:
C W F = Y−θ C XF
n
t
a
quarter
c chord
b
k-1 k k+1 three-quarter
chord
Figure 21-6 Wing geometry.

The criterion
(i )F F F
3 = (c )3 cos θ + (n )3 sin θ = 0
(where i is the unit vector along the x-axis of W) gives the pitch angle of the wing section:

θ = tan−1 −(c )F F
3 /(n )3

sin θ = −(c )F
3/ (c )F
3
2 + ( )F 2
n 3

cos θ = (n )F F 2 + ( )F 2
3 / (c )3 n 3
Note that (c )F

3
2
+ (n )F
3
2
= 1 − (t )F
3 , so the pitch angle can not be found by this method when (t )3 = ±1,
2 F
hence when the tangent vector is perpendicular to the x–y plane of the wing frame.
21-4.2 Section Motion and Loads
21-4.2.1 Section Axes
Figure 7 shows the aerodynamic environment in section axes. The velocity of the wing relative the air, at the
F
quarter chord, is vQC . So the velocity of the air relative the section, in X axes, is:
⎛ ⎞
ux
u X
= ⎝ uy ⎠ = −C XF vQC
F
uz
Then the angle of attack, resultant velocity, Mach number, and yaw angle are:
uz
α = tan−1 + θAF
ux

U = u2x + u2z
M = U/cs
uy
Λ = tan−1
U
These quantities, from the quarter-chord aerodynamic environment, are used to obtain the quasistatic load using
the airfoil tables. The aerodynamic analysis obtains the lift and drag forces acting on the section. The lift L and
drag D act in the x–z plane of the X axes, respectively normal and parallel to the resultant velocity U . There is
also a spanwise drag force R, in the y-axis direction, and a nose up pitching moment M . The forces act at a point
cxAF aft of the airfoil leading edge. The total force acting on the section, in X axes, is:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
D D cos(α − θAF ) − L sin(α − θAF ) (Dux − Luz )/U
F X = Yα−θAF ⎝R⎠ = ⎝ R ⎠=⎝ R ⎠
L L cos(α − θAF ) + D sin(α − θAF ) (Lux + Duz )/U
⎛ ⎞
0
MX = ⎝M ⎠
0
The section forces in the wing frame axes are then
F F = CF X F X
section axes X
z
α -θ AF
D
wing plane axes W

z
θ α -θ AF x
D
φ U
Figure 21-7 Section aerodynamic environment (nose up pitch moment and

spanwise drag not shown; lift and drag act a distance cxAF aft of the leading
edge).
Note that the components of F X are the chord and normal forces (cX and cN ).
Reverse flow (from trailing edge to leading edge) occurs when ux is negative, or when |α| is greater than
90 deg (not quite the same when θAF = 0). The sign conventions in reverse flow must be consistent with the
definition of angle of attack in the airfoil tables. Positive pitch or negative uz gives an angle of attack near −180
degrees. The airfoil table convention is that the lift coefficient is then positive (as for testing in a wind tunnel). The
drag coefficient is always positive. The transformation to X axes above then gives a negative z-axis force from the
lift, and a negative x-axis force from the drag.
If the section has a trailing-edge flap, then the aerodynamic analysis also provides the loads acting on the flap:
lift Lf , drag Df , spanwise drag Rf , and pitch moment Mf . The forces act at the flap hinge, cxh aft of the airfoil
leading edge and czh above the airfoil chord line. The loads F X and M X given above are the total loads acting
on the section. The flap loads are evaluated in a similar fashion:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
Df Df cos(α − θAF ) − Lf sin(α − θAF ) (Df ux − Lf uz )/U
FfX = Yα−θAF ⎝ Rf ⎠ = ⎝ Rf ⎠=⎝ Rf ⎠
Lf Lf cos(α − θAF ) + Df sin(α − θAF ) (Lf ux + Df uz )/U
⎛ ⎞
0
MfX = ⎝ Mf ⎠
0
in the X axes; and FfF = C F X FfX in the wing frame axes.
21-4.2.2 Wing Plane Axes
Figure 7 shows the aerodynamic environment in wing plane axes as well. The section pitch angle is θ,
measured from the wing plane. The velocity of the air relative the section, in W axes, is:
⎛ ⎞
uT
uW = ⎝ uR ⎠ = Y−θ uX = −C W F vQC
F
−uP
Then the inflow angle and angle of attack are:

uP
φ = tan−1
uT
α = θ − φ + θAF
The W axes are obtained from the X axes by a rotation about the y-axis, hence the values of the resultant velocity
U , Mach number, and yaw angle Λ are not changed; and uR = uy . The total force acting on the section, in W
axes, is: ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
D D cos φ + L sin φ (DuT + LuP )/U
FW = Y−φ ⎝ R ⎠ = ⎝ R ⎠=⎝ R ⎠
L L cos φ − D sin φ (LuT − DuP )/U
⎛ ⎞
D
= Yα−θAF −θ ⎝ R ⎠ = Y−θ F X
L
The moments are not affected by the rotation about the y-axis. The flap loads in wing plane axes are evaluated in
a similar fashion.
21-4.2.3 Unsteady Motion
The unsteady aerodynamic loads are calculated using thin-airfoil theory, which requires the upwash w and
velocity V in the section axes X:
w = −(eTn v F )QC = uz
V = −(eTc v F )QC = ux
Here w is the upwash normal velocity of the wing relative to the air. The upwash gradient along the chord (typically
from a pitch rate) is approximated using the upwash velocities at the quarter chord and three-quarter chord points:
∂w ∼ w3QC − wQC
w = =
∂x c/2
The time derivatives are then calculated using the time derivatives of the wing velocity, v̇ F .
ẇ = −(eTn v̇ F + ėTn v F )QC
V̇ = −(eTc v̇ F + ėTc v F )QC
ẇ3QC − ẇQC
ẇ =
c/2
The dynamic stall model requires the rate of change of angle of attack. The rate can be calculated using a two-step
finite-difference expression:
1
α̇ = 3α(t) − 4α(t − Δt) + α(t − 2Δt)
2Δt
1
α̈ = α(t) − 2α(t − Δt) + α(t − 2Δt)
Δt2
The values of α at previous times are stored for use in this calculation. The dimensionless rate α̇c/U is calculated
using the section velocity U . Alternatively, from the definition of angle of attack in section axes X:
u̇z ux − u̇x uz
α̇ = Kα̇
u2x + u2z
u̇x = −(eTc v̇ F + ėTc v F )QC

u̇z = −(eTn v̇ F + ėTn v F )QC
or the dynamic stall model can be formulated in terms of ẇ:
ẇ sign ux
α̇ = Kα̇
U
With these analytical expressions, the upwash gradient is used for the acceleration: α̈ = ẇ Kα̇ . The finite-
difference calculation is preferred, since the time derivatives of the wing velocity (v̇ F ), as obtained from the
standard aerodynamic interface of a structural dynamic component, do not include the effects of a time-varying
induced or interference velocity. However, the finite-difference form can not be linearized. The factor Kα̇ is
introduced to compensate for the absence of the rate of change of induced velocity in the calculation of α̇.
21-4.2.4 Section Power
The section induced, interference, and profile power losses can be evaluated from the loads and velocities.
The section induced power is identified as the product of the induced velocity and the wing lift. Thus:
dPind
= −(vind
F T F
) L
dr
dPint
= −(vint
F T F
) L
dr
where the section lift gives

⎛ ⎞ ⎛ ⎞
0 −uz /U
LF = C F X Yα−θAF ⎝ 0 ⎠ = CF X ⎝ 0 ⎠ L
L ux /U
The section power inherits the definition of induced and interference used for the velocities. The section profile
power is defined as:
dPo
= U D + uR R
dr
which is the vector product of the velocity and drag forces on the wing. The total wing power is obtained by
integrating the section power losses over the wing span. Thus
dPind
Pind = ΔrA
dr
dPint
Pint = ΔrA
dr
dPo
Po = ΔrA
dr
where ΔrA is the panel width. The summation is over all spanwise panels.
21-4.2.5 Panel Loads
The aerodynamic analysis provides the section force vector F X and pitch moment M (load per unit length),
acting at a point cxAF aft of the airfoil leading edge. The component must provide discretized loads acting at the
collocation points, which are the section quarter chord and three-quarter chord. The point of application of the
aerodynamic loads is cΔx aft of the quarter chord, and c( 1/2 − Δx) forward of the three-quarter chord, where
1
Δx = xAF −
4
Figure 8 shows the geometry. Forces at the quarter chord and three-quarter chord that are equivalent to F X and
M must be found. The solution ⎛ ⎞
0
F3QC = ⎝
X 0
2
⎠
2ΔxFzX − M
c
X
FQC = F X − F3QC
X
satisfies the criteria

X X
FQC + F3QC = FX
X
(FQC )z cΔx − (F3QC
X
)z c( 1/2 − Δx) = M
Note that the chord and spanwise forces act entirely at the quarter chord. Only the normal force is used at the
three-quarter chord. The discretized loads, in wing frame axes, are thus:
⎛ ⎞
0
F
F3QC = CF X ⎝ 0
2
⎠ ΔrA
2ΔxFzX − M
c
F
FQC = C F X F X ΔrA − F3QC
F
F
section
loads
(X axes)
M
c xAF
discretized FQC
panel
loads
(F axes) F3QC
or c Δx c ( 1
2
)
− Δx
FQC
M QC
Figure 21-8a Panel loads.

F
section
loads Ff
(X axes)
M
Mf
czh
c xAF
c xh
discretized
panel FQC
loads Fhinge
(F axes) F3QC
M hinge
.
or c Δx c ( 1
2
)
− Δx
FQC
Fhinge
M QC
M hinge
.
Figure 21-8b Panel loads, with trailing-edge flap.

where ΔrA is the panel width. Alternatively, the component provides discretized force and moment at the quarter
chord
X
FQC = FX
⎛ ⎞
0
MQCX
= M X + ⎝ −cΔxFzX ⎠
cΔxFyX
or
F
FQC = C F X F X ΔrA
⎛ ⎞
0
F
MQC =C FX ⎝ M − cΔxFzX ⎠ ΔrA
cΔxFyX
F
in wing frame axes. To these discrete loads can be added prescribed increments from an external analysis: ΔFQC
F F F
and ΔF3QC , or ΔFQC and ΔMQC . The total wing force and moment are obtained by summing the panel loads
over the wing surface. The loads about the origin of the wing frame F, in F axes, are thus:
FwF = FF
F
MwF = M + rF F F
where rF is the position of the collocation point relative the origin of the wing frame. The summation is over all
collocation points: quarter chord and three-quarter chord points of all spanwise panels.
If the wing section has a trailing-edge flap, then the discretized loads at the quarter chord and three-quarter
chord collocation points act on the portion of the wing forward of the flap, and are calculated from the difference
between the total loads and the flap loads:
X
Ffwd = F X − FfX
Mfwd = M − Mf + c(xh − xAF )Ff z − czh Ff x
Chordwise flexibility of the wing section is generally neglected when using lifting-line theory, so these collocation
points can be attached to a rigid extension of the chord line, even if the quarter chord or three-quarter chord is aft
of the flap leading edge. The discretized loads on the flap, in wing frame axes, are:
F
Fhinge = C F X FfX ΔrA
⎛ ⎞
0
F
Mhinge = C F X ⎝ Mf ⎠ ΔrA
0
acting at the flap hinge (cxh aft of the airfoil leading edge and czh above the chord line). Optionally the flap loads
can be ignored even though the section has a flap. Then the loads at the quarter chord and three-quarter chord
collocation points are again the total loads on the airfoil section. Without the flap loads, the aeroelastic affects of
the flap motion can not be modelled. The total wing force and moment are obtained by summing the panel loads
over the wing surface. The loads about the origin of the wing frame F, in F axes, are thus:
FwF = FF + F
Fhinge
F
MwF = M + rF F F + F
Mhinge + rhinge
F F
Fhinge
F
where rhinge is the position of the flap hinge relative the origin of the wing frame:
1
rhinge = rQC + (4xh − 1)(r3QC − rQC ) + czh n
2
The summation is over all collocation points of all spanwise panels, including points on the flap hinges.
21-4.2.6 Bound Circulation
The aerodynamic analysis provides the bound circulation Γ of each panel. For the wake components, a vector
of the bound circulation at all span stations of the wing is provided (component output). Wake components may
also require the peak bound circulation values along the span, and the corresponding span station of the peaks.
The following algorithm is used to find the bound circulation peaks.
The set of bound circulation values for all panels is searched for the maximum positive value Γ+ and the
minimum negative value Γ− . The span stations at which these peaks are located are also saved (r+ and r− ). The
span stations searched may be limited to the range rLmin to rRmax (input parameters). For the uniform inflow
models, the circulation is suppressed in a tip loss region, so the span stations are only searched in the range BL to
BR . For the nonuniform inflow models, the tip vortex can roll up inboard of the tip, so the span stations are only
searched in the range rT V L to rT V R . As implemented, the search ignores panels where the lift factor defined by
B or rT V is zero. The positive and negatives peaks are designated the left or right peak according to the values
of their span stations. The maximum magnitude (positive or negative) is identified. Thus the vector of bound
circulation peaks consists of the following quantities:
a) Maximum, left, and right circulation peak values: Γmax , ΓL , and ΓR .

b) Corresponding span stations: rmax , rL , and rR .
Note that Γmax equals ΓL or ΓR (not the absolute value). If the circulation has only one sign over the entire span,
then only one peak will be found. In such a case, the peak found (positive or negative) is designated the right peak
and the maximum, the left peak circulation has a value of zero, and the span station of the left peak is set to a value
beyond the wing tips.
The wake model may divide the far wake trailed vorticity into several spanwise panels, in order to provide
more detailed structure of the inboard vorticity. For example, a trailing-edge flap or a rapid change in the wing
chord can produce a step change in the bound circulation, which might produce significant rollup of the inboard
trailed vorticity. Such inboard rollup can be introduced by using more than one spanwise panel for the far wake.
The far wake panels are defined by specifying the panel boundaries at span stations rBi , i = 0 to M + 1 for M
panels (where rB0 and rB(M +1) are always the wing tips). A wing panel edge rEj should be placed at rBi . In
any case, the analysis replaces the input rBi with the nearest rEj . The bound circulation peaks are found using
the above algorithm for each far wake panel, as well as for the entire wing.
The wake and wake geometry models may use the centroid and moment of the trailed vorticity. This
information is calculated and stored by the wing component with the circulation peak values. The wing bound
circulation Γk is calculated at the collocation points rAk , k = 1 to K; and the bound circulation is zero at the wing
tips (k = 0 and k = K + 1). This bound circulation distribution produces trailed vorticity sheets with strength
−∂Γ/∂r = −(Γk − Γk−1 )/(rAk − rA(k−1) ) for rA(k−1) < r < rAk ; discretized as vortex lines with strength
δk = Γk−1 − Γk at the aerodynamic panel edges rEk , k = 1 to K + 1. The trailed vorticity is partitioned into sets
of adjacent lines that have the same sign: δi , i = i1 to i2 . The total strength Γ, centroid rC , and moment (radius
of gyration) rG of the trailed vorticity in the set are:
$ 2 i
∂Γ
Γ= − dr = (Γi−1 − Γi )
∂r i=i 1
$ i2 $ i
∂Γ ri
Γi − Γi−1 2
1
Γ rC = − r dr = − r dr = (Γi−1 − Γi ) (ri−1 + ri )
∂r i=i ri−1 ri − ri−1 i=i
2
1 1
$ i2 $
∂Γ ri
Γi − Γi−1
2
Γ rG = − (r − rC )2 dr = − (r − rC )2 dr
∂r i=i ri−1 ri − ri−1
1
i2
1
= (Γi−1 − Γi ) (ri−1 − rC )2 + (ri − rC )2 + (ri−1 − rC )(ri − rC )
i=i1
3
A set can be ignored (combined with an adjacent set) if its strength is too small. A set is small if its strength is
less than a specified fraction of the total wake strength: |Γ| < fG Γsum , where Γsum is the sum of all positive Γ.
The sets with largest positive and negative strengths are never considered small. One or more adjacent small sets
are combined with the previous or next set (for an even number of adjacent small sets, half are combined with the
previous set and half with the next set). For the k-th trailed vorticity line, at rEk , the fraction of the vorticity in the

set that is within the distance |rEk − rC | from the centroid is fk = δi /Γ, where the sum is over all i such that
|rEi − rC | ≤ |rEk − rC |. This fraction is used in the wake geometry consolidation model. The vector of bound
circulation peaks thus includes the centroid data for the trailed vortex lines (at the panel edges):
a) Total strength, centroid, and moment of the trailed vorticity: Γ, rC , and rG (identical values
for all lines within a set).
b) Trailed vorticity fraction and strength, and the number of the set: fk , δk , and set.
This centroid data is for the edges of the far wake trailed vorticity panels, which do not necessarily include all the
wing aerodynamic panel edges. However, the moment rG is still calculated considering all trailed vortex lines.
While rC and rG are stored as span station values, the integration does account for the span scale factor R.
21-4.3 Section Aerodynamic Model
21-4.3.1 Aerodynamic Coefficients
The aerodynamic analysis obtains the lift and drag forces acting on the section. The lift L and drag D act in
the x–z plane of the X axes, respectively normal and parallel to the resultant velocity U . There is also a spanwise
drag force R, in the y-axis direction, and a nose up pitching moment M . The forces act at a point cxAF aft of the
airfoil leading edge. The bound circulation Γ is required for the wake analysis. These loads are evaluated from
the corresponding coefficients, defined as follows:
L = 1/2ρU 2 cc
D = 1/2ρU 2 ccd
M = 1/2ρU 2 c2 cm
R = 1/2ρU 2 ccr
Γ = 1/2U ccg
where ρ is the air density, c the wing chord, and U the resultant velocity at the section (always positive). The bound
circulation coefficient is the same as the lift coefficient, cg = c , except for the unsteady and dynamic stall terms.
The spanwise drag coefficient is evaluated based on the assumption that the viscous drag force on the section has
the same sweep angle as the local section velocity. Optionally cr can be evaluated from either the chordwise drag
coefficient cd , or from cd at zero lift: 0
cd (α) tan Λ
cr =
cdz tan Λ
It is the skin friction drag that has a spanwise component, not the pressure drag. The chordwise drag at zero lift
is intended to be an estimate of the skin friction drag, but the actual spanwise drag is probably between these two
options. Also, the skin friction is in the direction of the velocity at the bottom of the boundary layer, which does
not necessarily have the same yaw angle as the outer flow.
If the section has a trailing-edge flap, the above coefficients are the total loads, and the aerodynamic analysis
also provides the loads acting on the flap:
Lf = 1/2ρU 2 cc f
2
Df = 1/2ρU ccdf
Mf = 1/2ρU 2 c2 cmf
Rf = 1/2ρU 2 ccrf
The spanwise drag on the flap is evaluated from the chordwise drag as for the total loads. The airfoil tables give
the flap moment about the reference axis cxAF f aft of the section leading edge, on the reference chord line. Then
⎛ ⎞
cf
cmf = cmf table + [ (xh − xAF f ) 0 −zh ] Y−(α−θAF ) ⎝ 0 ⎠
cdf
⎛ ⎞
(c f uz + cdf ux )/U
= cmf table + [ (xh − xAF f ) 0 −zh ] ⎝ 0 ⎠
(cdf uz − c f ux )/U
= cmf table + Δcmf hinge
is the moment about the flap hinge, cxh aft of the section leading edge and czh above the reference chord line.
Several adjustments can be made to the airfoil table data. Input lift, drag, and moment coefficient increments
can be specified, as a function of span station, for the total loads and for the flap loads. These increments provide a
means to modify the aerodynamic characteristics without requiring that the airfoil table be changed. For example,
the drag increment might be used to represent a Reynolds number effect. An input aerodynamic center increment
can be specified as a function of span station, producing a moment coefficient increment. The quantity (c ΔxAC )
is the distance the aerodynamic center is aft of the airfoil table aerodynamic center. Such a shift might be used
to represent three-dimensional effects at the wing tip. If the wing has trailing-edge flaps, then the airfoil table
includes the effects of the flaps. In addition, input derivatives can be specified, so the flap deflection φ generates
increments in the coefficients. These increments are set to zero in reverse flow. If the airfoil table contains data
only for zero flap angle, then the static loads produced by the flap come only from these increments. Note also
that the parameter θAF can be given a prescribed value, to generate a shift in the zero lift angle of attack of the
section. The two-dimensional airfoil coefficients are thus:
c 2D = c + Δc + c φ φ c f2D =c f + Δc f +c fφ φ
cd2D = cd + Δcd + cdφ φ cdf2D = cdf + Δcdf + cdfφ φ
cm2D = cm + Δcm + cmφ φ − ΔxAC c cmf2D = cmf + Δcmf + cmfφ φ + Δcmf hinge
and then
cg = c cr = cd tan Λ crf = cdf tan Λ
The coefficients on the right-hand side come from the airfoil table. These adjustments produce effective airfoil
table data, to which the corrections described below are applied.
The section aerodynamic model is based on two-dimensional airfoil characteristics. Effective angle of attack
and Mach number are evaluated from the actual α and M of the section. For second-order lifting-line theory, this
angle of attack is obtained using the induced velocity calculated at the three-quarter chord. The lift, drag, and
moment coefficients are evaluated from the airfoil tables at the effective αe and Me . Then these coefficients are
corrected, to obtain values representing the quasistatic loads on the section, plus the effects of any dynamic stall
delay. Finally various increments are added to the coefficients. The coefficients thus may consist of the following
terms:
a) corrected airfoil table;

b) dynamic stall leading-edge vortex loads;
c) unsteady loads in attached flow;
d) prescribed coefficient increments from an external analysis.
The tip flow corrections are applied to the lift and circulation. The yawed flow, Reynolds number, and tip flow
corrections are applied to the coefficients of the flap loads, as for the coefficients of the loads on the entire section.
The “total” coefficient is the sum of all of the terms (with tip flow corrections) that are used by the model. The
“total static” coefficient excludes the dynamic stall and unsteady motion terms. The “old” coefficient excludes any
prescribed increment from an external aeroacoustic analysis.
21-4.3.2 Yawed Flow and Reynolds Number Corrections
Yawed flow over the wing section is accounted for using the equivalence assumption for swept wings: that
the yawed section drag coefficient is given by two-dimensional airfoil characteristics, and the normal section lift
coefficient is not influenced by yawed flow below stall. Since the wing viewed in a frame moving spanwise at
a velocity V sin Λ (where V is the wing velocity, yawed at angle Λ) is equivalent to an unyawed wing with free
stream velocity V cos Λ, except for changes in the boundary layer, there should be no effect of spanwise flow on
the loading below stall. It is also assumed that the total wing drag force is in the direction of the yawed flow.
Accounting for the effective dynamic pressure and angle-of-attack of the yawed section relative to the normal
section leads to
c (α) = c 2D (α cos2 Λ)/ cos2 Λ
cd (α) = cd2D (α cos Λ)/ cos Λ
cm (α) = cm2D (α cos2 Λ)/ cos2 Λ
for the section aerodynamic coefficients in terms of two-dimensional airfoil characteristics. These results are
largely verified by the experimental data for yawed wings. The definition of the section yaw angle gives
5
u2x + u2z
cos Λ =
u2x + u2y + u2z
(|Λ| is less than 90 degrees, so cos Λ is always positive). In the above expressions, the factors of cos Λ that multiply
the coefficients arise from geometric scaling, not aerodynamics. The aerodynamic assumptions define the effective
angle of attack αe . For lift and moment, assuming that in the linear range the normal section coefficients are not
affected by yaw results in αe = α cos2 Λ. It is assumed that the yawed-section angle of attack determines the drag,
giving αe = α cos Λ.
Next consider a wing with both yaw Λ of the free stream and sweep of the wing quarter chord relative to
the analysis section (figure 9). The analysis section is in a plane perpendicular to the wing reference line. Hence
the wing sweep angle is the angle between the tangent to the quarter chord locus, and the tangent to the wing
reference line (positive for aft sweep of the right tip). The total sweep angle of the free stream relative to the quarter
chord is ΛT = Λ + (in reverse flow ΛT = Λ − , since Λ has the same sign as uy ). The swept wing equivalence
assumption relates the aerodynamics of the wing normal and yawed sections. Here the aerodynamic analysis
is conducted for sections perpendicular to the wing reference line. For this analysis section, the aerodynamic
coefficients are:
c (α) = K2 c 2D
(αK−1 cos2 ΛT )/ cos2 ΛT = c 2D
(αK cos2 Λ)/ cos2 Λ
cd (α) = K cd2D (αK−1 cos ΛT )/ cos ΛT = cd2D (α cos Λ)/ cos Λ
cm (α) = K2 cm2D (αK−1 cos2 ΛT )/ cos2 ΛT = cm2D (αK cos2 Λ)/ cos2 Λ
The sweep of the wing by introduces the factor K :

0
cos ΛT cos − sin tan Λ normal flow
K = =
cos Λ cos + sin tan Λ reverse flow
(K is always positive). Without yawed flow, K = cos , so sweep of the wing reduces the lift-curve slope of the
analysis section. The compressibility effects on the swept wing depend on
cos(Λ + )
Mn = M = K M
cos Λ
which is the normal section Mach number. Lifting-line theory is applicable for sweep that varies slowly along the
span (unless a special deriviation and implementation is used), so kinks in the quarter-chord locus can result in
incorrect prediction of the induced velocity, hence of the induced power. An empirical correction is introduced to
compensate: a factor on the near-wake induced velocity, proportional to the change in sweep. The sweep change
corresponding to the panel edge is Δ(rEj ) = (rAj ) − (rA(j−1) ), for j = 2 to K (wing tips are not considered).
Then the inflow correction factor at the collocation point rAk is
K
Fsk = 1 + fs Δ(rEj ) max(0, 1 − 2|rAk − rEj |/Δrs )
j=2
where Δrs is the spanwise extent of the effect, and fs is the correction scale (empirically determined).
If the Reynolds number Re of the rotor section does not equal the Reynolds number Ret of the airfoil table,
the drag and lift coefficients may be corrected as follows (ref. 1):
1
cd = cd
KD 2D
c = KL c 2D (α/KL )
where K = K0 (Re/Ret )n . Experimental data for turbulent flow suggests n = 0.125 to 0.2 for the exponent;
n = 0.2 is the 1/5-th power law for a turbulent flat plate boundary layer. If the Reynolds number is larger for
analysis section
in plane perpendicular
to wing reference line
yaw
Λ
tangent to
reference line
wing normal to analysis
section plane
ε
sweep
normal
chord
analysis
chord
yawed
chord
Figure 21-9 Swept and yawed wing aerodynamics.

the airfoil than for the table, then K > 1, and the above corrections reduce the drag coefficient and increase the
maximum lift coefficient. The table Reynolds number is defined in terms of the Reynolds number corresponding
to Mach number M = 1:
Ret = Me Re1
where Me is the effective Mach number. The actual Reynolds number, for the yawed section, is
Vy cy Uc
Re = =
ν ν cos2 Λ
The analysis also includes lift, drag, and moment coefficient increments, which are an alternative means for
introducing a Reynolds number correction. The input factors KL0 and KD0 provide a direct means to change the
maximum lift and the drag, independent of the Reynolds number.
The analysis must combine the yawed flow and Reynolds number corrections, consider the case of large angle
of attack and reverse flow, and allow for nonzero lift at zero angle of attack. The correction in reverse flow (for
angles of attack near ±180) is analogous to that in normal flow (for angles of attack near 0). The airfoil table
(with adjustments) is searched for the angle of attack αz at which the lift is zero. Then the drag and moment
are evaluated at zero lift, giving cdz = cd (αz ) and cmz = cm (αz ). These quantities are evaluated and stored at
initialization, as a function of Mach number for normal and reverse flow, at each span station. The unsteady loads
models also require the lift and moment slopes, c α and cmα , which are calculated using a one-degree increment
from the zero-lift angle of attack. Then the following effective angle of attack is used:
0
(cos2 ΛT /KL )(α/K − αz ) lift and moment
αe − αz = C(α/K − αz ) =
(cos ΛT )(α/K − αz ) drag
where α/K is the normal section angle of attack; and the corrected coefficients are:
KL
c = K2 c (αe )
cos2 ΛT 2D
1
cd = K cd (αe )
KD cos ΛT 2D
% &
KL
cm 2
= K cm2D (αe ) − cmz + cmz
cos2 ΛT
The form of the lift and moment corrections ensures that the coefficients below stall are unchanged, except for the
K factors associated with sweep. The corrections are washed out for angles of attack near ±90 (so αe = α at
±90), using linear interpolation over a range of 30 deg. So if
|(|α| − 90)|
w= <1
30
then the corrections are washed out by using (1 + (C − 1)w) in place of C (and also for the inverse of the
corresponding coefficient factor). In addition, the corrections are limited to a maximum amplitude of 20.
21-4.3.3 Static Stall Delay
There is evidence that rotational effects on the boundary layer produce a delay of separation, particularly for
the inboard sections of rotating wings (refs. 24 and 25). Such a stall delay can be modelled using input factors
Ksd to modify the airfoil coefficients obtained from the tables:
c =c table + KsdL (c L
−c table ) c L
= c α (αe − αz )
cd = cdtable + KsdD (cdL − cdtable ) cdL = cdz
cm = cmtable + KsdM (cmL − cmtable ) cmL = cmα (αe − αz ) + cmz
Generally the factors Ksd are between 0 and 1. This modification is only used in normal flow, not in reverse flow.
The modification is washed out from 30 to 60 deg angle of attack by using wKsd in place of Ksd , where
60 − |α|
w=
30
For a rotor in axial flow, Selig (ref. 25) gives the following for the stall delay factor:
% &
1 1.6 c/r a − (c/r)D
Ksd = −1
2π .1267 b + (c/r)D

with D = d/(λr/R) for lift and D = d/(2λr/R) for drag; λ = Ωr/ V 2 + (Ωr)2 . The constants a, b, and d are
approximately 1. Here c is the rotor blade chord, R the blade radius, r the distance from the center of rotation, V
the rotor axial velocity, and Ω the rotational speed. So c/r is a measure of the rotation of the streamlines on the
blade surface. However, this expression is not valid for c/r > 1.
Corrigan and Schillings (ref. 24) implement a stall delay as c = c 2D

(α − Δα) + c α Δα, where Δα =
f (cmax /c α ). The factor is given as f = (Kθ TE /.136)n
− 1, with θ ∼
TE = c/r and c/r = .1517K −1.084 . This
correction is equivalent to using here
n
c/r .1517 1/1.084 n
KL = f + 1 = = 1.291(c/r).0775
.136 c/r
The exponent n = 0.8 to 1.8, with the larger values typically giving better correlation.
21-4.3.4 Tip Flow Corrections
Tip flow corrections include a tip loss factor, and compressible tip relief (a small reduction in effective Mach
number). The tip loss factor is needed when the induced velocity is obtained from a uniform inflow model. With
such an approximation for the wake solution, the lift does not go to zero at the wing tip. Let rL be the span station
of the left wing tip, and rR the span station of the right wing tip. The conventional tip loss correction assumes that
the wing has drag but no lift for span stations to the left of BL or to the right of BR (r < BL or r > BR ). For an
aerodynamic panel extending from rEk to rE(k+1) , the uncorrected portion is
max(rEk , BL ) ≤ r ≤ min(rE(k+1) , BR )
Thus the tip loss is implemented by multiplying the lift and moment coefficients of the panel by the factor
min(rE(k+1) , BR ) − max(rEk , BL )
f=
rE(k+1) − rEk
(0 ≤ f ≤ 1). Typically the tip loss factor B represents about 3% of the wing span. Separate tip loss factors are
used for the steady, unsteady, and dynamic stall lift terms, and the steady, unsteady, and dynamic stall moment
terms.
For nonuniform inflow a tip loss factor is not appropriate, but it is necessary to consider the span station of
the rolled up tip vortex when it reaches the trailing edge of the wing. The implementation is similar to that for the
tip loss factor. Let rT V L and rT V R be the span stations of the rolled up tip vortex, at the left and right wing tips
respectively. The wing has no bound circulation to the left of rT V L or to the right of rT V R . Figure 10 illustrates
RECTANGULAR PLANFORM
r TV
collocation
point edge of
aerodynamic
rolled up tip vortex
panel
(at wing trailing edge)
HIGHLY TAPERED TIP

r TV
Figure 21-10 Formation of vortex on wing tip.

the formation of a vortex on the wing tip. Thus the effect of the tip vortex rollup is implemented by multiplying
the lift and moment coefficients of the panel by the factor
min(rE(k+1) , rT V R ) − max(rEk , rT V L )
f=
rE(k+1) − rEk
(0 ≤ f ≤ 1). Typically the tip vortex forms at the trailing edge at a span station rT V 1–2% inboard of the wing tip.
Such small displacements of the tip vortex rollup have little influence on the wing loading, so to avoid convergence
problems rT V should be set to the wing tip value (suppressing the effect entirely). For a highly tapered wing tip,
the tip vortex can form at a span station 5–6% inboard of the tip. In such cases the effect of rT V is significant.
Note that if all or most of an aerodynamic panel is outboard of rT V , the wake induces a large upwash at it, perhaps
stalling the wing section. For a highly tapered tip this is a representation of a physical effect; for a rectangular
tip it is just a source of convergence problems, and inaccurate drag. The lifting line wing component must know
whether the wake model being used is uniform or nonuniform inflow, in order to implement the correct tip loss
factor. The wake model might depend on the status of a loop in the solution procedure.
The three-dimensional flow at the wing tip increases the critical Mach number of the tip sections, compared
to the two-dimensional flow characteristics. This compressible tip relief may be accounted for by reducing the
effective section Mach number by the factor KM , which must be specified at each span station, for the lift, drag,
and moment coefficients. Thus the effective Mach number, used in the airfoil tables, is:
Me = K K M M
The sweep factor K (defined above) makes this the Mach number normal to the swept quarter-chord, while M is
the Mach number of the section in X axes.
21-4.3.5 Prescribed Coefficient Increments from an External Analysis
For communication with an external aeroacoustic analysis, prescribed increments in the section aerodynamic
coefficients can also be obtained from a two-dimensional table. The table dependent variables are the coefficient
increments:
lift Δc or Δ(M 2 c ) flap lift Δc f or Δ(M 2 c f )

drag Δcd or Δ(M 2 cd ) flap drag Δcdf or Δ(M 2 cdf )
normal force Δcn or Δ(M 2 cn ) flap normal force Δcnf or Δ(M 2 cnf )
chord force Δcx or Δ(M 2 cx ) flap chord force Δcxf or Δ(M 2 cxf )
moment Δcm or Δ(M 2 cm ) flap moment Δcmf or Δ(M 2 cmf )
spanwise drag Δcr or Δ(M 2 cr ) flap spanwise drag Δcrf or Δ(M 2 crf )
circulation Δcg or Δ(M 2 cg )
The flap coefficient increments are about the hinge. Lift and drag increments are obtained from the normal and
chord force coefficients by transforming from section axes to wind axes:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
Δcd Δcx (Δcx ux + Δcn uz )/U
⎝ 0 ⎠ = Y−α ⎝ 0 ⎠ = Y−θAF ⎝ 0 ⎠
Δc Δcn (Δcn ux − Δcx uz )/U
If Δ(M 2 c) is present in the table, then the corresponding Δc is not used. If the lift or drag coefficient is present
in the table, then the normal and chord force coefficients are not used. If the spanwise drag coefficient is not
present in the table, then it is calculated from Δcr = Δcd tan Λ. If the circulation coefficient is not present in the
table, then it is calculated from Δcg = Δc (although the unsteady terms in lift and circulation are not the same).
Otherwise, a zero value is used for the coefficient increment if it is absent from the table.
Alternatively, the load increments can be the discretized force and moment on the section, in wing frame
axes. Then the table dependent variables are the load components:
F F F
quarter chord force ΔFQCx ΔFQCx ΔFQCz
F F F
three-quarter chord force ΔF3QCx ΔF3QCy ΔF3QCz
F F F
quarter chord moment ΔMQCx ΔMQCy ΔMQCz
F F F
flap force ΔFhinge x ΔFhinge y ΔFhinge z
F F F
flap moment ΔMhinge x ΔMhinge y ΔMhinge z
If a discrete load is present in the table, then coefficient increments are not used. Either the three-quarter chord
force or the quarter chord moment is used, depending on the discretization option.
The first independent variable of the table is the span station. The table data are interpolated to the span
stations of the panel midpoints. The second independent variable of the table is time or azimuth. The table data are
interpolated (perhaps cyclically for azimuth) to the current time value t. The following options are implemented.
a) Table function of span and time τ : τ = t − t0 , where t0 is a specified reference time.

b) Table function of span and azimuth ψ: ψ = Ωt + ψ0 , where the frequency Ω is obtained
from a specified period.
c) Table function of span only, constant in time (the table class is still two-dimensional, but
there is only one time value in the table).
These prescribed coefficient increments are used without any corrections. It is assumed that the external analysis
already accounts for three-dimensional flow, yawed flow, Reynolds number, and other effects.
21-4.4 Unsteady Airfoil Motion
Unsteady lift and moment in attached flow are calculated based on thin-airfoil theory. Models from in-
compressible thin-airfoil theory, ONERA (EDLIN), and Leishman-Beddoes are implemented. Only the unsteady,
noncirculatory terms are required, since steady loads are obtained from the airfoil tables and lift deficiency function
effects are usually accounted for in the wake-induced velocity. The unsteady loads depend on the upwash w (the
normal velocity of the wing relative to the air), and the upwash gradient along the chord w = ∂w/∂x (typically
from a pitch rate). Optionally the unsteady loads can be set to zero for stalled flow, defined as fd < .7 (where
fd = f (αd ) is the trailing-edge separation point of the Leishman-Beddoes dynamic stall model).
21-4.4.1 Incompressible Thin-Airfoil Theory
The incompressible unsteady loads are derived following reference 2. The following modifications of the
thin-airfoil-theory results are introduced.
a) The steady loads are obtained from the angle of attack based on the quarter-chord velocity
of the section. Thus the terms in the thin-airfoil-theory result that are produced by the steady,
quarter-chord upwash are excluded.
b) The theoretical aerodynamic center is at the quarter chord. Thus the moment about the
quarter chord is evaluated, and identified as the moment about the actual aerodynamic center,
located a distance cxAC aft of the leading edge. Then the moment about the airfoil table
reference axis can be obtained.
c) The theoretical lift-curve slope is 2π. The loads are multiplied by the factor a/(2π) to get
the real lift-curve slope a.
Finally, the loads are expressed in coefficient form. The results are:
% &
a c 1 1
cgU S = w ±
U 2 2 2
% &
a c 1 1 c c
c US
= 2 |V |w ± ± ẇ + ẇ
U 2 2 2 4 4
% &
a c c 3c
cmU S = (xAF − xAC )c U S + 2 −|V |w − ẇ + ẇ
U 16 16 8
The upper sign in ± or ∓ is used for normal flow, and the lower sign for reverse flow. Note that U is always
positive, while V is negative in reverse flow. The following options are implemented for the lift-curve slope a:
1) Constant: a = K Ka 2π, where Ka is an input factor; typically Ka = .9 or so.

2) Prandtl-Glauert: a = K Ka 2π/β, with β = 1 − Me2 (an upper limit of .95 is placed on
the Mach number in β).
3) Prandtl-Glauert with lift divergence (typically Mdiv = .75 or so):

⎧ % &
⎪ 1
⎪ K
⎪ aK 2π √ M < Mdiv
⎪
⎪ 1 − M2
⎪
⎪ " #
⎪
⎪
⎨ 1−M
a = K Ka 2π Mdiv < M < Mdiv + .1
⎪
⎪ (1 − Mdiv ) 1 − Mdiv
2
⎪
⎪ " #
⎪
⎪
⎪
⎪ 1 − M M − M − .1
⎪
⎩ K Ka 2π +
div
M > Mdiv + .1
(1 − Mdiv ) 1 − Mdiv
2 1 − Mdiv − .1
4) From the airfoil table at zero lift: a = K c α at (Me , αz ).
5) From the airfoil table at the local angle of attack: a = K c α at (Me , αd ).
6) Secant slope from the airfoil table: a = c (αd )/(α − K αz ) = K c α (c (αd )/c L ).
All cases include the sweep factor K . Different choices can be made for the circulatory lift, other lift, and moment;
and a separate Ka value can be used for the flap terms. According to steady thin-airfoil theory, both the lift and
moment scale with the Prandtl-Glauert factor. The following options are implemented for the aerodynamic center
shift δx = xAF − xAC :
1) Constant: δx = xAF − xAC , where xAC is the input aerodynamic center (fraction of chord
aft of the leading edge).
2) From the airfoil table at zero lift: δx = cmα /c α at (Me , αz ); less 1/2 in reverse flow.
3) From the airfoil table at the local angle of attack: δx = (cm − K2 cmz )/c , using cm and
c at (Me , αd ), and cmz is the moment at zero lift; less 1/2 in reverse flow.
The lift and moment slopes at zero lift can be evaluated and stored at initialization. The secant slope for a, and
δx at the local angle of attack, require the lift and moment at αd , which are already available for the section. The
lift-curve slope at the local angle of attack is calculated using a one-degree increment from αd , hence requires an
additional evaluation of the lift from the airfoil table.
For an airfoil with a trailing-edge flap, the unsteady loads are derived following references 2 to 4. The flap
chord is f (fraction of section chord c). Then
θf = cos−1 [±(1 − 2 f )]
ξf = 1 − 2 f
ξh = −1 + 2xh
ξa = ξh − ξf = 2(−1 + xh + f)
An aerodynamically balanced flap (flap leading edge forward of the hinge) has ξa positive. The upwash along the
section is c c c
wtotal = wQC + x + w + V φ + x − ξh φ̇ − V ξa δ(xf )φ
4 2 2
(here x is measured aft from the midchord, so θ = cos−1 (±2x/c)). The aerodynamic balance term (proportional
to ξa ) is present with a sealed gap at the flap leading edge, and absent with an open gap. The upwash is expressed
as a cosine Fourier series in θ, with harmonics as follows.
% & ' 7
c 1 c 1 c
w0 = w + w + |V |θf φ + (S1 ∓ ξh θf ) φ̇ − |V |ξa φ + −|V |φ − ξh φ̇
4 π 2 S1 2
% & '
c 2 c 1 S2 C1 c 7
w1 = ± w + |V |S1 φ + θf + ∓ ξh S1 φ̇ − |V |ξa φ + − φ̇
2 π 2 2 2 S1 2
% &
2 Sn c 1 Sn−1 Sn+1 Sn Cn
wn = |V | φ + + ∓ ξh φ̇ − |V |ξa φ
π n 2 2 n−1 n+1 n S1
where Sk = sin kθf and Ck = cos kθf . In evaluating the derivatives of these harmonics, V̇ terms are included.
The solution for the pressure is then expressed as a Glauert series, with the following coefficients.
p0 = 2ρ|V |w0
ρc
p1 = 2ρ|V |w1 + (2ẇ0 − ẇ2 )
2
ρc
pn = 2ρ|V |wn + (ẇn−1 − ẇn+1 )
2n
The unsteady loads are corrected as described above, and written in coefficient form. The results are:
% &
a 1
cgU S = ± w0 + w1
U 2
% &
a 1 1
c US = 2 ± p 0 + p1
U 2ρ 2
% &
a 1
cmU S = (xAF − xAC )c U S + 2 − p0 (2 ∓ 2) + p1 ± p2
U 16ρ
" !#
a 1 1 S2 1 Sn−1 Sn+1
c fU S = 2 p0 (θf − S1 ) + p1 θf − + pn −
U 2πρ 2 2 n=2
2 n−1 n+1
0 % &8
a 1 1
+ − p 0 + p 1
U2 2ρ 2
%
a 1
cmfU S = 2 ∓ p0 (∓2 − 4ξh )θf + S1 (±4 + 4ξh ) ∓ S2
U 16πρ

S3 S2
+ p1 ± S 1 − − 2ξh θf −
3 2

S4 S3
+ p2 ± θ f − − 2ξh S1 −
4 3
!#
Sn−2 Sn+2 Sn−1 Sn+1
+ pn ± − − 2ξh −
n=3
n−2 n+2 n−1 n+1
0 % &8
a 1
+ − p 0 (2 − 4ξh ) − p 2ξ
1 h − p 2
U2 16ρ
The steady terms (w and φ contributions) are excluded from these unsteady loads. The upper sign in ± or ∓ is
used for normal flow, and the lower sign for reverse flow; and the terms in braces {. . .} are only present for reverse
flow. The series converge with ten pressure terms (except that the aerodynamic balance terms in the flap lift and
flap moment produced by φ do not converge at all; but these are static loads, here contained in the airfoil tables
instead).
21-4.4.2 ONERA EDLIN Theory
The ONERA EDLIN (Equations Differentielles Lineaires) theory for the unsteady loads is presented in
reference 5. The extended model includes the effects of heave as well as pitch, and time-varying free stream. In
the absence of stall, it was found that thin-airfoil theory results compared well with measured behavior. To include
the effects of compressibility, Küssner’s coefficients are used, as tabulated by van der Vooren (ref. 6) and curve-fit
by Petot (ref. 5). The following modifications are introduced here:
a) The steady loads are excluded.
b) The moment about the quarter chord is used for the moment about the actual aerodynamic
center. Then the above options for δx = xAF − xAC are available.
c) The loads are multiplied by the factor a/(2π/β) to get the real lift-curve slope. Then the
above options for a are available.
d) A constant is changed so the circulatory loads match the lift deficiency function value of
C = .5 at high frequency (instead of C = .53).
e) The expressions are extended to reverse flow.
In coefficient form, the results are:

% &
a c 1 1 a
cgU S = w ± + L1
U 2 2 2 U
% & a
a c 1 1 c c
c U S = 2 |V |w ± ± ẇfL0 + ẇ fL1 + L1
U 2 2 2 4 4 U
% &
a c c 3c
cmU S = (xAF − xAC )c U S + 2 −|V |w fM 0 − ẇfM 0 + ẇ fM 1
U 16 16 8
% &
c 1 1
L̇1 + λL1 = μ ±ẇ + ẇ ±
2 2 2
where
fL0 = β 1 + 5(β .57 − 1)
fL1 = β [1 + 3.92(β − 1)]

fM 0 = β 1 + 1.4M 2

fM 1 = β −1.2625 + 1.5330 tan−1 (10.5 − 15M )

with β = 1 − Me2 (an upper limit of .95 is placed on the Mach number in β). These factors give a good
representation of Küssner’s coefficients, except that the moment produced by heave (cmh = π4 k 2 fM 0 ) is always
real, when it should exhibit a phase shift for M > 0. The L1 term accounts for the airfoil shed wake effects (lift
deficiency function), with
2U
λ= λ0 (1. − .76M )
c
1
μ = − (3 − β)
4
Petot (ref. 5) gives a value of 0.17 for λ0 . For an airfoil with a trailing-edge flap, the effects of compressibility are
approximated by introducing the lift and moment factors in the incompressible results as follows.
term lift factor moment factor

w0 , w1 , wn 1 fM 0
ẇn → ẇ, φ̇ fL0 fM 0
ẇn → ẇ , φ̈ fL1 fM 1
No shed wake terms (such as L1 ) are included in the unsteady loads for flap lift or flap moment. The ONERA
EDLIN theory as presented by Petot (ref. 5) uses a = 2π/β. The ONERA BH theory (ref. 7) implies using the
lift-curve slope at the local angle of attack.
21-4.4.3 Leishman-Beddoes Theory
The Leishman-Beddoes theory for unsteady loads in attached flow is presented in references 8 to 10. The
theory is based on the indicial response of a thin airfoil in compressible flow to heave, pitch, and flap motions. The
indicial response is a combination of impulsive (small time) and circulatory (long time) terms, each approximated
by exponential functions of time. In this form, the equations for the loads can be transformed from indicial response
to Laplace domain, and thence to state equations (ordinary differential equations in time). The impulsive indicial
response is derived using piston theory, which is valid for nonzero Mach number and small enough time. Thus
while giving nonsingular results at zero Mach number, this theory does not include the incompressible limit exactly.
The amplitude of the circulatory indicial response is obtained from the quasistatic incompressible response, scaled

with β = 1 − Me2 . The following modifications are introduced here:
a) The steady loads are excluded.
b) The moment about the quarter chord is used for the moment about the actual aerodynamic
center. Then the above options for δx = xAF − xAC are available.
c) The loads are multiplied by the factor a/(2π/β) to get the real lift-curve slope. Then the
above options for a are available.
d) For an airfoil with a trailing-edge flap, the expressions are extended to include the effects
of aerodynamic balance, and to include the flap hinge moment produced by pitch and heave.
e) The expressions are extended to reverse flow.
In coefficient form, the results are:

a
cgU S = (LC1 + LC2 )
U
a aβ
c US
= (LC1 + LC2 ) + 2 c c0α kα Lα + c2 c0q kq Lq + c c0φ kφ Lφ + c2 c0φ̇ kφ̇ Lφ̇
U U

L̇C1 + λ1 LC1 = λ1 A1 c∞q cw + c∞φ̇ cφ̇ − A1 c∞α ẇ + c∞φ (|V |φ)˙

L̇C2 + λ2 LC2 = λ2 A2 c∞q cw + c∞φ̇ cφ̇ − A2 c∞α ẇ + c∞φ (|V |φ)˙
λ1 = (2U/c)b1 β 2 λ2 = (2U/c)b2 β 2
L̇α + λα Lα = λα ẇ
L̇q + λq Lq = λq ẇ
L̇φ + λφ Lφ = λφ (|V |φ)˙
L̇φ̇ + λφ̇ Lφ̇ = λφ̇ φ̈
λα = U/(cM kα ) λq = U/(cM kq )
λφ = U/(cM kφ ) λφ̇ = U/(cM kφ̇ )
a aβ
c fU S (FC1 + FC2 ) + 2 c c0α kα Fα + c2 c0q kq Fq + c c0φ kφ Fφ + c2 c0φ̇ kφ̇ Fφ̇
=
U U
a aβ
cmfU S = HC + 2 c c0α kα Hα + c2 c0q kq Hq + c c0φ kφ Hφ + c2 c0φ̇ kφ̇ Hφ̇
U U

ḢC + λ6 HC = λ6 c∞q cw + c∞φ̇ cφ̇ − c∞α ẇ + c∞φ (|V |φ)˙
λ6 = (2U/c)b6 β 2
a
cmU S = (xAF − xAC )c U S + (MC1 + MC2 )
U
aβ
+ 2 c c0α kα (Mα3 + Mα4 ) + c2 c0q kq Mq + c c0φ kφ Mφ + c2 c0φ̇ kφ̇ Mφ̇
U

ṀC1 + λ5 MC1 = λ5 c∞q cw

ṀC2 + λ6 MC2 = λ6 c∞φ̇ cφ̇ − c∞φ (|V |φ)˙
λ5 = (2U/c)b5 β 2 λ6 = (2U/c)b6 β 2
Ṁα3 + λα3 Mα3 = λα3 b3 A3 ẇ
Ṁα4 + λα4 Mα4 = λα4 b4 A4 ẇ
Ṁq + λq Mq = λq ẇ
Ṁφ + λφ Mφ = λφ (|V |φ)˙
Ṁφ̇ + λφ̇ Mφ̇ = λφ̇ φ̈
λα3 = U/(cM kα b3 ) λα4 = U/(cM kα b4 )
λq = U/(cM kq ) λφ = U/(cM kφ ) λφ̇ = U/(cM kφ̇ )
1−η a
cdU S = (c + c US
) c + (L C1 + LC2 ) − c2
cα U
1 a
− (c + c U S ) (LC1 + LC2 )
cα U
cdfU S = 0
where the pattern of the definition of the F and H states follows that of the L states, and
A1 = .3 b1 = .14
A2 = .7 b2 = .53 b5 = .5
A3 = 1.5 b3 = .25 b6 = 1.5
A4 = −.5 b4 = .1
The time constants are:

c0 κ
T = 2M k = 2M
c1 (1 − M ) + c∞ 2M 2 β (Ab)

where (Ab) is A1 b1 + A2 b2 = .413 for the lift and flap lift; b5 for the moment kq and zero for the moment kα ;
and b6 for the moment kφ and kφ̇ and for the flap moment. The factors κ are introduced to improve correlation
with measured loads. Good results are obtained with κL = .75 for lift and κM = .80 for moment. The constants
c0 , c1 , and c∞ are given in the following table.
derivative load 2πc0 2πc1 2πc∞ reverse flow

α c 4 4 ±2π or ±c α −1
cm −1 −1 −π( 2 ∓ 2 ) or ±cmα
1 1
cf 2x 2 ITAT or ±c fα −1
cmf − 12 xy −x ITAT or ±cmf α
q = α̇c/U c 1 1 π( 12 ± 12 ) −1
cm − 12
7
− 54 − π8
cf 1
2 x(2 + ξf ) 2 ITAT −1

cmf − 24
1
x y(7 + 2ξh ) + 4z 2 − 34 x ITAT
φ c 2x − 2ξa 2 ITAT or c φ
cm − 12 x(2 + ξf ) + 12 ξa (1 + 2ξf ) − 32 ITAT or cmφ −1
cf 2x − 2ξa 2 ITAT or c fφ
cmf − 12 xy − ξa z −x ITAT or cmf φ −1
φ̇c/U c 1
2 xy x ITAT −1

cm − 241
x y(7 + 2ξh ) + 4z 2 − 34 x ITAT
c f
1
2 xy x ITAT −1

cmf − 16 x y(1 − ξh ) + z 2 − 12 x2 ITAT
where x = (1 − ξf ), y = (1 + ξf − 2ξh ), z = (ξh − ξf ). The aerodynamic balance term (proportional to ξa )

is present with a sealed gap at the flap leading edge, and absent with an open gap. The last column indicates the
sign changes in reverse flow for c0 and c1 . The expressions for c1 are obtained from reference 9, neglecting any
effect of flap aerodynamic balance (c1 is required to have the same sign as c0 , and the maximum allowed value of
k is 100). The quasistatic terms of incompressible thin-airfoil theory (ITAT) give c∞ . However, the flap lift and
moment derivatives with flap angle must be obtained from the airfoil tables if the flap has aerodynamic balance
with a sealed gap (the thin-airfoil theory expressions do not then converge). Optionally all the α and φ derivatives
for c∞ can be evaluated from the airfoil tables (then a = 2π/β should be used). This theory includes the effects
of the airfoil shed wake, but not entirely in the “circulatory” terms. So that the wake effects are not duplicated, the
near shed wake elements should not be included in the vortex wake model of nonuniform inflow.
In cdU S , c is the lift coefficient without the unsteady load or dynamic stall terms. The unsteady drag is based
on approximating the steady drag in attached flow as follows (ref. 11):
1−η 2
cd = cdz + c
cα
where η < 1 is the chord force recovery factor (typically η = .95), and cdz is the drag at zero lift. The following
options are implemented for the drag recovery factor η:
1) Constant: input value of η.
2) From the airfoil table at small angle of attack: (1 − η)/c α = (cd − cdz )/c2 .
3) From the airfoil table at the local angle of attack: (1 − η)/c α = (cd − cdz )/c2 at (Me , αd ).
Evaluating this quantity at the local angle of attack requires the lift and drag at αd , which are already available
for the section. At small angle of attack, it is calculated and stored at initialization, using a four-degree increment
from the zero-lift angle of attack. The Leishman-Beddoes theory as presented in ref. 8 uses the secant slope for a
in the LC terms, and a = 2π/β in the remaining terms. It uses the aerodynamic center δx from the airfoil tables
at the local angle of attack. The Leishman-Beddoes theory uses (1 − η)/c α at the local angle of attack.
21-4.5 Dynamic Stall
Dynamic stall is characterized by a delay in the occurrence of separated flow produced by the wing motion,
and high transient loads induced by a vortex shed from the leading edge when stall does occur. Dynamic stall
models from Johnson, Boeing, Leishman-Beddoes, and ONERA (EDLIN and BH) are included. As implemented,
the dynamic stall models still use the airfoil table for steady characteristics, evaluated at an angle of attack that
includes the dynamic stall delay. Let αd be the delayed angle of attack, calculated from the effective angle of
attack αe that incorporates the corrections for yawed flow and Reynolds number. Then the corrected coefficients
are:
KL αe − αz
c = K2 c 2D (αd ) + Δc DS
cos2 ΛT αd − αz
" 2 #
1 αe − αz
cd = K cd2D (αd ) − cdz + cdz + ΔcdDS
KD cos ΛT αd − αz
% &
KL αe − αz
cm = K 2
cm2D (αd ) − cmz + cmz + ΔcmDS
The form of the lift and moment corrections ensures that the coefficients below stall are unchanged, except for the
K factors associated with sweep. A maximum value of the angle of attack delay can be specified, in particular to
avoid difficulties at small values of the section velocity: max |αd − αe | = Δαmax delay . A limit is also placed on
the magnitude of the ratio (αe − αz )/(αd − αz ) in the drag correction, to avoid problems near (αd − αz ) equal
to zero. The ΔcDS ’s are increments defined by the dynamic stall model, generally attributed to the leading-edge
vortex. The change in lift produced by the leading-edge vortex occurs very fast over time and span, particularly
for the typical resolutions used with lifting-line theory. Optionally the corresponding circulation increment can be
ignored: ΔcgDS = 0 instead of ΔcgDS = Δc DS . The wake strength and resulting induced velocity do not then
include the effect of Δc DS . The dynamic stall effects are washed out for angles of attack near ±90, by multiplying
(αd − αe ) and ΔcDS by the factor w.
Stall of the airfoil can be suppressed by using for the delayed angle of attack the smaller of the actual angle
of attack α, and a specified maximum angle αmax in the linear range (say 4 degrees):
αd − αz = min(|αe − αz |, αmax ) sign (αe − αz )
The coefficient increments produced by the leading-edge vortex are set to zero as well.
21-4.5.1 Johnson Model
The Johnson dynamic stall model (adapted from ref. 12) uses an angle of attack delay proportional to α̇, plus
impulsive lift and moment increments from the leading-edge vortex. The angle of attack is evaluated with a time
delay Δt = τd c/2U that accounts for the hysteresis effects around stall:
αd = αe (t − Δt) ∼
= αe − Δt α̇e = αe − τd α̇e c/2U
with α̇e = C α̇/K . Alternatively, αd is the solution of a state equation:

α̇d + λd (αd − αz ) = λd (αe − αz )
where λd = 2U/cτd . McCroskey (ref. 13) and Beddoes (ref. 14) found that the dynamic stall delay correlates
fairly well in terms of the normalized time constant τd . Their results for lift and moment stall are:
lift moment
McCroskey τL = 8.6 ± 1.0 τM = 5.6 ± 0.4
Beddoes τL = 10.8 ± 1.2 τM = 4.9 ± 1.0
The values τL = 9.2 and τM = 5.4 are typical. A time constant is also required for the drag stall delay. The
equations for the loads include the increments Δc DS , ΔcdDS , and ΔcmDS , which are produced by the leading-
edge vortex. When the blade section angle of attack reaches the dynamic stall angle αDS , a leading-edge vortex is
shed. As this vortex passes aft over the airfoil upper surface it induces large transient loads. The experimental data
of reference 15 show that the peak incremental aerodynamic coefficients depend on the pitch rate at the instant of
stall, α̇c/U , approximately as follows:

Δc peak = Δc min |α̇e c/U |/.05, 1
LEV

Δcmpeak = ΔcmLEV max 0, min (|α̇e c/U | − .02)/.03, 1

Δcdpeak = ΔcdLEV min |α̇e c/U |/.05, 1
with α̇e = C α̇/K . The appropriate sign changes for negative angle of attack and reverse flow are included, and in
reverse flow 1/2Δc DS is added to the moment. It is assumed that the incremental coefficients caused by the shed
vortex (ΔcDS ) rise linearly to the above peak values in the time increment ΔtDS = τv c/2U , and then fall linearly
to zero in the time ΔtDS again. Hence the model involves impulsive lift and nose down moment changes when
dynamic stall occurs. After these transient loads decay, the wing section is assumed to be in deep stall, and dynamic
stall is not allowed to occur again until the flow has reattached. Flow reattachment takes place when the angle
of attack drops below the angle αRE . The dynamic stall angle αDS and reattachment angle αRE correspond to
fd = f (αd ) = .7, where f is the trailing-edge separation point of the Leishman-Beddoes model. The experimental
data of reference 15 give Δc LEV = 2.0 and ΔcmLEV = −0.65. Typically τv = 3.6 to 5.6. Note that if the total
rise and fall time 2ΔtDS is interpreted as the time the leading-edge vortex takes to traverse the chord (distance c),
then the speed of the vortex is vvortex = U/τv . So τv = 4 implies the leading-edge vortex travels at one-fourth
the free stream speed. The vortex-induced loads are calculated using the following information, for lift, drag, and
moment: a state variable, STATE; the peak vortex-induced load, Δcpeak ; and the time since the occurrence of
dynamic stall, tDS . The state variable tracks the flow condition:
60 attached flow (α < αRE )
STATE = 1 dynamic stall (α > αDS )
2 crossed reverse flow boundary
After crossing a reverse flow boundary, dynamic stall can not occur until the flow first reattaches. If STATE defines
the flow as attached or having crossed the reverse flow boundary, there are no vortex loads: Δc DS
= ΔcdDS =
ΔcmDS = 0. If STATE defines the flow as in dynamic stall, then the vortex-induced loads exist:
KL
Δc DS = K2 Δc peak F
cos2 ΛT
1
ΔcdDS = K Δcdpeak F
KD cos ΛT
KL
ΔcmDS = K2 Δcmpeak F
cos2 ΛT
where
tDS tDS
F = max 0, min ,2 −
ΔtDS ΔtDS
with tDS the time since dynamic stall occurred. Figure 11 describes the dynamic stall algorithm used to calculate
the leading-edge vortex load.
21-4.5.2 Boeing Model
The Boeing dynamic stall model (developed in refs. 16 to 18) uses an angle of attack delay proportional to
the square-root of α̇, which produces the basic hysteresis effects. The coefficient increments produced by the
leading-edge vortex are not used in this model. The delayed angle of attack is

αd = αe − τd |α̇e c/2U | sign α̇e
with α̇e = C α̇/K . Alternatively, this can be considered a time lag dependent on the pitch rate, giving a state
equation for αd :
α̇d + λd (αd − αz ) = λd (αe − αz )

where λd = (2U/cτd ) |α̇e c/2U |. The time constant τd is a function of Mach number and the airfoil section,
obtained from oscillating airfoil tests. For an NACA 0012 airfoil, ref. 18 gives
τL = max(0, min(1.71, 2.16 − 2.81M ))

τM = max(0, min(1.12, 1.71 − 2.98M ))
τD = 0
which can be used to define the input values.
21-4.5.3 Leishman-Beddoes Model
The Leishman-Beddoes dynamic stall model (refs. 11, 19, and 20) uses a delayed angle of attack, plus lift
and moment increments from the leading-edge vortex. This model characterizes the airfoil static stall behavior by
the trailing-edge separation point f (fraction of chord from leading edge), and a critical lift coefficient c CR at the
separation onset boundary (leading-edge separation at low Mach number, shock reversal at high Mach number).
The airfoil data for lift are used to identify constants s1 , s2 , and αs that generate f (α) as follows:
6
1. − .3 exp((|α − αz | − αs )/s1 ) |α − αz | ≤ αs
f=
.04 + .66 exp((αs − |α − αz |)/s2 ) |α − αz | > αs
Then |α − αz | = αs or f = .7 is taken as the definition of stall. The parameters c CR , s1 , s2 , and αs are required
as a function of Mach number, for positive and negative angle of attack, normal and reverse flow, at each span
station. The Leishman-Beddoes model for unsteady flow is based on fd = f (αd ) at the delayed angle of attack.
Here the model is modified to use the static loads directly from the airfoil tables, instead of fitting the static loads
to analytical functions. No change to the model is implied for lift and moment, since these analytical functions are
intended to be equivalent to the airfoil table data. With this modification, fewer parameters are required, perhaps
only those less sensitive to the airfoil shape. The model is also extended to reverse flow. In order to handle
oscillations through αz , a continuous monotonic function of α is needed:
6
fz − f α − αz ≥ 0
f1 =
f − fz α − αz < 0
if crossed reverse flow boundary

STATE = 2
else if last state is attached flow
if dynamic stall has occurred, |αd − αz | ≥ αDS
STATE = 1
calculate Δcpeak from α̇e c/U
in reverse flow, add 1/2Δc peak to Δcmpeak
if αd − αz < 0, change sign of Δc peak and Δcmpeak
tDS = Δt (|αd − αz | − αDS )/(|αd − αz | − |αd − αz |last )
else
STATE = 0
else if last state not attached flow
if still not attached, |αd − αz | ≥ αRE
STATE = last state
if STATE = 1 then
Δcpeak = last Δcpeak
tDS = last tDS + Δt
else
STATE = 0
if STATE = 1, calculate ΔcDS
Figure 21-11 Dynamic stall algorithm for leading-edge vortex load (Johnson model).
where fz = f (αz ) = 1. − .3 exp(−αs /s1 ); note that fz may not be the same for positive and negative angle of
attack. In order to handle large angle of attack, the function f is modified:
⎧
⎪ 1. − .3 exp((|α − αz | − αs )/s1 ) |α − αz | ≤ αs
⎨
f = .04 + .66 exp((αs − |α − αz |)/s2 ) |α − αz | > αs
⎪
⎩
fH (90 − |α − αz |)/(90 − αH ) |α − αz | > αH
The corresponding inverse α(f ) is
⎧
⎪ αs + s1 ln((1. − f )/.3) f ≥ .7
⎨
|α − αz | = αs − s2 ln((f − .04)/.66) f < .7
⎪
⎩
90 − (90 − αH )f /fH f < fH
Here αH = αs + Hs2 and fH = .04 + .66e−H , and H = 6. is used. This modification to the definition of f does
not affect the model for attached flow or around stall, but with it the delayed angle of attack behaves reasonably at
very large angles.
The delayed angle of attack αd is calculated as follows. Static hysteresis around stall is modelled by using a
smaller αs when the angle of attack is decreasing:
αs = αs input − Δαs (1 − fd )1/4
(with fd from the last time step). There is a lag in the leading-edge pressure relative c , so a lagged lift is used in
the stall criterion:
L̇p + λp Lp = λp (αe − αz + LC /A)
αp = Lp + LI /A + αz
where A = ∂c /∂αe = K2 (KL / cos2 ΛT )c α . The dimensionless time constant Tp gives λp = 2U/cTp . The
unsteady lift c U S is split into LC (the LC1 + LC2 term in Leishman-Beddoes theory) and LI (the remaining
terms), calculated using a = K c α instead of the secant slope. The stall criterion is based on
c p = A (αp − αz )
If |c p | ≥ c CR , the critical condition for leading-edge or shock-induced separation has been reached; while if
|c p | < c CR , reattachment is allowed. This angle of attack gives a trailing-edge separation point f1p = f1(αp ).
There is an additional lag in the boundary layer response, modelled as a lag in f :
f1̇d + λf f1d = λf f1p
where λf = 2U σf /cTf . There are separate fd equations for lift and moment, to allow different behavior during
reattachment (implemented using different values of σf ). If fd is decreasing (Δfd < 0), the flow is separating;
if fd is increasing, the flow is reattaching. The difference Δfd is calculated at the end of the procedure, for use
during the next time step. Finally, the delayed angle of attack αd is calculated from f1d . Vortex lift accumulation
begins at the onset of stall (indicated by |c p | = c CR ), driven by the difference between the linear and nonlinear
lifts:
cv = c L
−c
c L
= K2 c α (α/K − αz ) = A (αe − αz )

KL αe − αz
c = K2 c 2D (αd )
The leading-edge vortex reaches the trailing edge at time τDS = Tvl , where tDS = τDS c/2U is the time since
the onset of stall. The speed of the vortex implied is vvortex = 2U/Tvl , or one-fourth the free stream velocity for
Tvl = 8. The vortex loads are obtained from cv with a time lag:
L̇v + λv Lv = Dċv
Δc DS
= Lv
ΔcdDS = Lv tan(αe − αz )

πτDS
ΔcmDS = −xs 1 − cos Δc DS
Tvl
where xs = .20 or .25 typically. The moment exists only until τDS = 2Tvl . In reverse flow, 1/2Δc DS is added to
the moment. The time constant gives λv = 2U σv /cTv . The switch D is calculated by starting with the value at
the last time step, and then
a) D = 0 if τDS ≥ Tvl .
b) D = 1 if |c p | < c CR and Δfd < 0; or 0 ≤ τDS ≤ Tvl ; or α is increasing and Δfd > 0.
Interactions between the mechanisms are accounted for by modifying the time constants, using the parameters σf
(for lift and moment) and σv . Figure 12 describes the algorithm. With the switch D present, it is not possible
to formulate the equation for Lv in terms of cv instead of ċv , such that the analysis can deal with the resulting
equation of motion. So an implicit solution is always used for Lv .
An alternative form of the delayed angle of attack calculation uses a lagged lift coefficient c p to obtain the
angle of attack αp :
c C
= A (αe − αz ) + LC
L̇p + λp Lp = λp c C
c p = Lp + LI
αp = c p /A + αz
To match the implicit solution of the Leishman-Beddoes model in references 11 and 19, the following form is
required:
c C = A (αe − αz ) + LC
L̇p + λp Lp = ċ C
c p =c C
− Lp + LI
αp = c p /A + αz
L̇f + λf Lf = f1̇p
f1d = f1p − Lf
The time derivatives on the right hand side of these state equations are not available analytically, so in this form
an implicit solution must be used.
The trailing-edge separation point f is related to the airfoil lift using the Kirchhoff expression (ref. 11), as
follows: √ 2

1+ f
c =cα (α − αz )
2

STATE = 2
if dynamic stall has occurred, |c p | ≥ c CR
STATE = 1
tDS = Δt (|c p | − c CR )/(|c p | − |c p |last )
else
STATE = 0
if still not attached, |c p | ≥ c CR
STATE = last state
if STATE = 1, then tDS = last tDS + Δt
else
STATE = 0
if flow separating, Δfd ≤ 0

if not near onset of leading-edge separation, |c p | ≤ c CR σf L = σf M = 1
else reached critical condition for separation, |c p | > c CR σf L = σf M = 1.75
but if separation forward of 70% chord, last fd ≤ .7 σf L = σf M = 2
but if α decreasing σf L = σf M = 2
else if flow reattaching, Δfd > 0: σf M = 5 and
if reattaching, |c p | ≤ c CR σf L = .5
else if separating, |c p | > c CR σf L = 1.
but if shedding leading-edge vortex, 0 ≤ τDS ≤ Tvl σf L = .25
but if α increasing σf L = .75
for vortex lift: σv = 1, then

if post stall, Tvl ≤ τDS ≤ 2Tvl σv = 3
if reattaching, Δfd > 0 σv = 4
if α decreasing and shedding vortex, 0 ≤ τDS ≤ Tvl σv = 2
if α decreasing and still σv = 1 σv = 4
if α decreasing and flow reattaching, Δfd > 0 σv = 1
Figure 21-12 Dynamic stall algorithm for modified time constants (Leishman-Beddoes model).
Hence the airfoil table data for lift define f as a function of α:

2
f = 2 c /c L
−1
with c L = c α (α − αz ). The parameter αs is given by f = .7 (c /c L = .8433); and then s1 and s2 are identified
by fitting the table data for |α − αz | < αs and |α − αz | > αs respectively. The critical lift c CR should be identified
from pressure data (ref. 11), but can also be determined based on the break in chord force (loss of leading-edge
suction) or the sudden increase in drag. The time constants are determined by correlation with unsteady airfoil
data. The following table gives the parameters as a function of Mach number for an NACA 0012 airfoil (ref. 19).
M .30 .40 .50 .60 .70 .75 .80

αs 15.25 12.5 10.5 8.5 5.6 3.5 0.7
Δαs 2.1 2.0 1.45 1.0 0.8 0.2 0.1
s1 3.0 3.25 3.5 4.0 4.5 3.5 0.7
s2 2.3 1.6 1.2 0.7 0.5 0.8 0.18
c CR
1.45 1.20 1.05 0.92 0.68 0.50 0.18
Tp 1.7 1.8 2.0 2.5 3.0 3.3 4.3
Tf 3.0 2.5 2.2 2.0 2.0 2.0 2.0
Tv 6.0 6.0 6.0 6.0 6.0 6.0 4.0
Tvl 7.0 9.0 9.0 9.0 9.0 9.0 9.0
The parameters are required as a function of Mach number, for positive and negative angle of attack, normal and
reverse flow, at each span station. Positive or negative angle of attack is defined relative the zero lift angle, so
α = 90 to 180 degrees is negative angle of attack in reverse flow. The convention is that αs and c CR are always
positive, so if necessary the values for normal flow and positive angle of attack can be used in all four flow regimes.
The INPUT program can identify values of the parameters αs , s1 , s2 , and c CR
from the airfoil tables.
The trailing-edge separation point value fd = f (αd ) = .7 can be used as a criterion for stall. The correspond-
ing value of αs can be identified from the airfoil table as the angle of attack where f = .7, instead of requiring αs
be input just for this purpose.
21-4.5.4 ONERA EDLIN Model
The ONERA EDLIN (Equations Differentielles Lineaires) dynamic stall model (ref. 5) uses a stall delay,
plus lift, drag, and moment increments calculated from second-order differential equations. The extended model
includes the effects of heave as well as pitch, and time-varying free stream. The load is divided into two parts.
The first part is the load in the absence of stall, which here gives the unsteady load for attached flow. The second
part of the load is driven by the difference between the linear load extrapolated to the unstalled domain, and the
real nonlinear static load. Tests show that dynamic stall occurs at a higher angle of attack than does static stall.
The absence of stall is preserved in the model by forcing the difference between the linear and nonlinear loads to
be zero for a time τd after exceeding the static stall angle. Here the model is modified in several ways. The static
and unsteady terms are separated from the dynamic stall effects. A pitch rate term in the lift (with coefficient d
below), that reference 5 associates with attached flow unsteady loads, is here contained in the dynamic stall load.
The loads are written in terms of α̇ instead of ẇ. Reference 5 uses ẇ in order to include the effects of time-varying
free stream, but for a three-dimensional wing α̇ also includes the wake-induced velocity. For all the dynamic stall
models, the option is available to evaluate α̇ from ẇ. The model is also extended to reverse flow.
The dynamic stall loads of the ONERA EDLIN model are calculated from the following second-order differ-
ential equations:
L̈2 + aL̇2 + bL2 = −bU Δc − eU α̇e
M̈2 + aṀ2 + bM2 = −bU Δcm − eU α̇e
D̈2 + aḊ2 + bD2 = −bU Δcd − eU α̇e sign(αe − αz )
with α̇e = C α̇/K , and αe in degrees here. These equations are driven by the difference between the linear and
nonlinear loads:
Δc = c L
−c c L
= K2 c α (α/K − αz ) = K2 Bc α (αe − αz )

Δcm = cmL − cm cmL = K2 cmα (α/K − αz ) + cmz = K2 Bcmα (αe − αz ) + cmz
Δcd = cdL − cd cdL = K cdz /(KD cos ΛT )
where B = KL / cos2 ΛT ; and here c , cm , and cd are the corrected coefficients at αe , without the ΔcDS and
ΔcU S terms. Then the load increments are
1
Δc DS
= (L2 + U Δc + dU α̇e )
U
1
ΔcmDS = (M2 + U Δcm + dU α̇e ) + M3
U
1
ΔcdDS = (D2 + U Δcd + dU α̇e sign(αe − αz ))
U
In reverse flow, 1/2Δc DS is added to the moment. The stall delay is accounted for by setting the right-hand side
of the differential equation to zero if τSS < τd , where tSS = τSS c/2U is the time since the static stall angle
was exceeded. The static stall angle corresponds to f (α) = .7, where f is the trailing-edge separation point of
the Leishman-Beddoes model. Figure 13 describes the algorithm used to calculate tSS . Petot (ref. 21) described
a refined transition model, intended to accommodate airfoils that exhibit larger nose-down pitching moments at
dynamic stall. The refined transition model assumes that the extra lift from dynamic stall is convected aft from
the quarter chord after moment stall occurs, producing the extra moment term M3 in ΔcmDS . Here this refined
transition model is implemented by
M3 = −(τSS − τdM ) μ Δc DS
for (τSS − τdM )μ = 0 to 1.5 (convection from quarter chord to trailing edge). This extra moment is turned off
after lift stall, by multiplying M3 by the factor (2 − τSS /τdL ) when τdL < τSS < 2τdL . The coefficients in these
equations depend on the lift difference Δc :

a = (2U/c) a0 + a2 (Δc )2
2
b = (2U/c)2 b0 + b2 (Δc )2
e = (2U/c) e2 (Δc )2
0
(c/2U ) d1 |Δc | lift and moment
d=
(c/2U ) (d0 |αe − αz | + d1 |Δc |) drag
The following table gives the range and typical values of the parameters (ref. 5). The notation has been changed,
so the table also shows the original ONERA notation.
ONERA
equation parameter range typical notation
all b0 .10 to .40 .2 r0
b2 0 to .50 .2 r2
lift τd 8 8 τd
a0 .10 to .40 .3 a0
a2 0 to .60 .2 a2
e2 0 to −.20 −.05 E2
d1 0 to −.75 −.2 σ1 /λ
moment τd 2 2 τd
a0 .10 to .40 .25 a0
a2 0 to .60 .1 a2
e2 0 to .06 .01 E2
d1 0 to .15 0 σ1
μ .02 to .08 .02
drag τd 0 0 τd
a0 0 to .50 .25 a0
a2 0 to .60 0 a2
e2 0 to −.05 −.015 E2
d0 .003 .003 σ0
d1 0 to −.05 −.04 σ1
The lift, moment, and drag use the same b0 and b2 . Values of the parameters for various airfoils are given in
references 22 and 23. The parameters are required as a function of Mach number, for positive and negative angle
of attack, normal and reverse flow, at each span station.
21-4.5.5 ONERA BH Model
The ONERA BH (Bifurcation de Hopf) dynamic stall model (ref. 7) uses a delayed angle of attack, plus lift
and moment increments calculated from second-order differential equations. The Hopf bifurcation model replaces
the time-invariant equilibrium state of flow by a time-varying equilibrium state, as the angle of attack exceeds
a critical value. The load is divided into two parts, a “steady” part (static plus attached flow unsteady) and an
“unsteady” part (dynamic stall). The ONERA EDLIN theory can be used for the unsteady load in attached flow.
For time-varying airfoil motion, the loads are evaluated at a delayed angle of attack that is calculated as in the
Leishman-Beddoes dynamic stall model. There are also dynamic stall load increments, driven by the pitch rate
and pitch acceleration. Here the model is modified in several ways. The static and unsteady terms are separated
from the dynamic stall effects. The model is also extended to reverse flow.
The dynamic stall loads of the ONERA BH model are calculated from the following first-order and second-
order differential equations (corresponding to the two parts of the load):
L̇1 + λL1 = −λU Δc L̈2 + aL̇2 + bL2 = eU α̇e + dU α̈e
Ṁ1 + λM1 = −λU Δcm M̈2 + aṀ2 + bM2 = eU α̇e + dU α̈e


STATE = 2
if static stall has occurred, |αe − αz | ≥ αSS
STATE = 1
tSS = Δt (|αe − αz | − αSS )/(|αe − αz | − |αe − αz |last )
else
STATE = 0
if still not attached, |αe − αz | ≥ αSS
STATE = last state
if STATE = 1, then tSS = last tSS + Δt
else
STATE = 0
tSS = 0
Figure 21-13 Dynamic stall algorithm for stall delay (ONERA EDLIN model).
with α̇e = C α̇/K , and αe in degrees here. The equations for L1 and M1 are driven by the difference between
the linear and nonlinear loads: Δc = c L − c and Δcm = cmL − cm (as for the EDLIN model). Then the load
increments are
1
Δc DS
=
(L1 + U Δc + L2 )
U
1
ΔcmDS = (M1 + U Δcm + M2 )
U
In reverse flow, 1/2Δc DS is added to the moment. The delayed angle of attack is calculated using a simplified
version of the Leishman-Beddoes model:
α̇p + λp (αp − αz ) = λp (αe − αz )
f1p = f1(αp )
f1̇d + λf f1d = λf f1p
αd = α(f1d )
where λp = 2U/cTp and λf = 2U/cTf . Optionally, here the full Leishman-Beddoes model for αd can be used.
The coefficients in these equations depend on the load increments:

a = (2U/c) ωs −a0 + a2 Δc2

b = (2U/c)2 ωs2 1 − b1 Δc sign(αe − αz ) − b2 Δc2
e = (2U/c) ωs e0
d = ωs d0
λ = (2U/c) λ0
using Δc = L2 /U and M2 /U in the lift and moment equations respectively (from the last time step if an implicit
solution is used). The parameters have separate values for separating and reattaching flow: separating flow values
are used if fp < fCR ; reattaching flow values are used if fp ≥ fCR (note the use of fp rather than fd ). For lift,
fCR = .7 is used. The critical angle of attack is Δαm larger for moment than for lift. So
6
.04 + .66 exp(−Δαm /s2 ) if α increasing
fCR =
1. − .3 exp(−Δαm /s1 ) if α decreasing
is used for the moment. The parameters must be evaluated from data on airfoils oscillating in the stalled flow
regime. The following table gives the parameters for an NACA 0012 airfoil (ref. 7). The notation has been
changed, so the tables also show the original ONERA notation.
ONERA
equation parameter separating reattaching notation
all ωs 2π(.124) 2π(.124) ωs
lift a0 .015 −3. β
a2 .75 γ
b1 0 a2
b2 −.6 η
e0 .186 E
d0 −.89 D
λ0 b
ONERA
equation parameter separating reattaching notation
moment a0 .015 −3. β
a2 7.5 γ
b1 −.75 a2
b2 0 η
e0 −.062 E
d0 .455 D
λ0 b
fCR Δαm 1 deg 1 deg
For reattaching flow, only a0 has a nonzero value, so the dynamic stall loads decay. The parameters are required
as a function of Mach number, for positive and negative angle of attack, normal and reverse flow, at each span
station.
21-4.6 Solving the State Equations
Some models for the unsteady loads and dynamic stall introduce ordinary differential equations for aerody-
namic state variables. These differential equations can be solved along with the structural dynamic equations, by
a trim, transient, or flutter part. There are a large number of states however, and the equations can be nonlinear as
well as time-varying. So an implicit solution, implemented within the lifting line wing component, is useful for
the trim and transient tasks. The equations must be formulated as differential equations in order to be linearized
in the flutter task, although a linearized solution of the highly nonlinear aerodynamics involved in dynamic stall is
not entirely consistent.
Two implicit methods are implemented to solve the differential equations: a finite-difference solution, based
on trapezoidal integration; and a sampled-data solution, based on the convolution integral or Duhamel’s integral
(ref. 11). Consider the first-order differential equation ẋ + λx = l. The solution xn at time tn is obtained from
the solution at time tn−1 = tn − Δt:

ln ln−1 ln ln−1
xn = + D xn−1 − −E −
λn λn−1 λn λn−1
For sufficiently large λ, this reduces to the static solution xn = ln /λn . Consider the first-order differential equation
˙ where s is an optional switch (needed in the equation for Lv ). The solution is:
ẋ + λx = ls,
xn = Dxn−1 + E (ln − ln−1 ) sn
For both equations, the coefficients are
6
(1 − λn Δt/4)/(1 + λn Δt/4) finite difference (E = 0 for λ Δt > 4)
E = En =
exp(−λn Δt/2) sampled data
and D = En En−1 . Consider the second-order differential equation ẍ+aẋ+bx = l. The finite-difference solution
is:
!
xn+1 − ln+1 /bn+1
Δt Δt2
( 1 + 2 an+1 + 4 bn+1 ) =
ẋn+1
" Δt2 Δt2
# ! !
1+ Δt
2 an+1 − 4 bn Δt + 4 (an+1 − an ) xn − ln /bn 1+ Δt
2 an+1 ln ln+1
+ −
Δt2
− Δt
2 (bn+1 + bn ) 1− Δt
2 an − 4 bn+1 ẋn − Δt
2 bn+1
bn bn+1
The static result is used if a Δt > 2. A sampled-data solution is not implemented for second-order equations.
The implicit methods use the stored history of the motion. The information required from the previous time
step is x, l/λ or l, and E. In addition, the angle of attack at previous time steps is needed, in order to calculate α̇
using the finite-difference expression. At the end of the process, these quantities are saved, along with the current
time. At the beginning of the process, if the current time is greater than the saved time, then the saved quantities
are identified as the quantities at the previous time step. In the trim task, a time-varying load is only appropriate if
the equilibrium solution is periodic. So the static solutions of the state equations are used if the trim part is static
or not periodic. For the transient task to be started properly, the trim solution should be static; or the time at which
the transient begins should be the end of a period; or degrees of freedom should be used instead of an implicit
solution. Otherwise, there will be an initial disturbance during the transient solution, produced by the incorrect
initial conditions implied by the stored motion. The implicit form should not be linearized. So when matrices are
being calculated, for the flutter task or a part solution, the static solutions of the state equations are used, and the
stored motion is not changed.
21–5 References
1) Yamauchi, G.K., and Johnson, W. “Trends of Reynolds Number Effects on Two-Dimensional Airfoil Charac-
teristics for Helicopter Rotor Analyses.” NASA TM 84363, April 1983.
2) Johnson, W. Helicopter Theory. Princeton University Press, Princeton, New Jersey, 1980.
3) Küssner, H.G., and Schwarz, L. “The Oscillating Wing with Aerodynamically Balanced Elevator.” NACA TM
991, October 1941.
4) Theodorsen, T., and Garrick, I.E. “Nonstationary Flow About a Wing-Aileron-Tab Combination Including
Aerodynamic Balance.” NACA Report 736, 1942.
5) Petot, D. “Differential Equation Modeling of Dynamic Stall.” La Recherche Aerospatiale, Number 1989-5
(corrections dated October 1990).
6) van der Vooren, A.I. “The Theodorsen Circulation Function and Aerodynamic Coefficients.” AGARD Manual
on Aeroelasticity, Volume VI, January 1964.
7) Truong, V.K. “A 2-D Dynamic Stall Model Based on a Hopf Bifurcation.” European Rotorcraft Forum, Italy,
September 1993.
8) Leishman, J.G. “Validation of Approximate Indicial Aerodynamic Functions for Two-Dimensional Subsonic
Flow.” Journal of Aircraft, Volume 25, Number 10, October 1988.
9) Hariharan, N., and Leishman, J.G. “Unsteady Aerodynamics of a Flapped Airfoil in Subsonic Flow by Indicial
Concepts.” Journal of Aircraft, Volume 33, Number 5, September-October 1996.
10) Leishman, J.G., and Nguyen, K.Q. “State-Space Representation of Unsteady Airfoil Behavior.” AIAA Journal,
Volume 28, Number 5, May 1990.
11) Leishman, J.G., and Beddoes, T.S. “A Semi-Empirical Model for Dynamic Stall.” Journal of the American
Helicopter Society, Volume 24, Number 3, July 1989.
12) Johnson, W. “The response and Airloading of Helicopter Rotor Blades Due to Dynamic Stall.” Massachusetts
Institute of Technology, ASRL TR 130-1, May 1970.
13) McCroskey, W.J. “Recent Developments in Dynamic Stall.” Symposium on Unsteady Aerodynamics, Tucson,
Arizona, March 1975.
14) Beddoes, T.S. “A Synthesis of Unsteady Aerodynamic Effects Including Stall Hysteresis.” Vertica, Volume 1,
Number 2, 1976.
15) Ham, N.D., and Garelick, M.S. “Dynamic Stall Considerations in Helicopter Rotors.” Journal of the American
Helicopter Society, Volume 13, Number 2, April 1968.
16) Harris, F.D.; Tarzanin, F.J., Jr., and Fisher, R.K., Jr. “Rotor High Speed Performance, Theory vs. Test.”
Journal of the American Helicopter Society, Volume 15, Number 3, April 1970.
17) Tarzanin, F.J., Jr. “Prediction of Control Loads Due to Blade Stall.” Journal of the American Helicopter
Society, Volume 17, Number 2, April 1972.
18) Gormont, R.E. “A Mathematical Model of Unsteady Aerodynamics and Radial Flow for Application to
Helicopter Rotors.” USAAVLABS TR 72-67, May 1973.
19) Leishman, J.G., and Beddoes, T.S. “A Generalized Model for Airfoil Unsteady Aerodynamic Behavior and
Dynamic Stall Using the Indicial Method.” American Helicopter Society Forum, June 1986.
20) Leishman, J.G., and Crouse, G.L., Jr. “State-Space Model for Unsteady Airfoil Behavior and Dynamic Stall.”
AIAA Paper Number 89-1319, April 1989.
21) Petot, D. “An Investigation of Stall on a 4.2m Diameter Experimental Rotor.” Seventh International Workshop
on Dynamics and Aeroelastic Modeling of Rotorcraft, St. Louis, October 1997.
22) Petot, D. “Dynamic Stall Modeling of the NACA 0012 Profile.” La Recherche Aerospatiale, Number 1984-6.
23) Petot, D. “Progress in the Semi-Empirical Prediction of the Aerodynamic Forces Due to Large Amplitude
Oscillations of an Airfoil in Attached or Separated Flow.” European Rotorcraft Forum, Italy, September 1983.
24) Corrigan, J.J., and Schillings, J.J. “Empirical Model for Stall Delay Due to Rotation.” American Helcopter
Society Aeromechanics Specialists Conference, San Francisco, January 1994.
25) Du, Z., and Selig, M.S. “A 3-D Stall-Delay Model for Horizontal Axis Wind Turbine Performance Prediction.”
AIAA Paper 98-0021, January 1998.
ξ panel panel panel control aero

vQC rQC vind loads
v3QC r3QC vint
motion 1 (a) yes yes no yes no
aero loads yes yes yes no yes X
panel X X X X X yes
FQC , F3QC
panel X X X X X yes
Fh
sensor yes yes yes yes yes X
Γ X X X X X yes
Γpeak X X X X X yes
force X X yes X X yes
power X X X yes X yes
Notes:
a) second order for some dynamic stall models
Figure 21-13 Functionality of lifting line wing component.

Chapter 22
WING INFLOW COMPONENT
22–1 Description
A wing inflow component calculates the wake induced velocity of a wing set. It is called a “uniform inflow”
component. There may be a corresponding nonuniform inflow component; which component gives a nonzero
velocity result then depends on the specified model, and perhaps on the wake loop level.
The induced velocity is calculated from a global, quasistatic model for a set of wings. This is an empirical
model based on an ideal-wing theory. The induced velocity is calculated at collocation points on this wing set,
and the interference velocity at other points can be calculated. The induced velocity is constant over the wing set.
The aerodynamic model is quasistatic, designed to be used with the mean (filtered) values of the component input.
The induced velocity in transient or flutter tasks can be obtained from dynamic inflow, which is a global,
low frequency model. A mean induced velocity term is related, by a first order differential equation, to the net
aerodynamic lift on the wing set. The differential equation (linearized inflow model, with time lags) is obtained
from a perturbation of the uniform inflow model. Alternatively, differential equations can be constructed by simply
introducing time lags in the uniform inflow model.
In summary, the following wake models can be used: zero, uniform inflow, nonuniform inflow, dynamic
inflow, lagged inflow and trim. The dynamic inflow or lagged inflow model calculates perturbations from trim
(which must be a uniform or nonuniform model). For the nonuniform inflow model, it is possible to still use
uniform inflow for collocation points off the wing set (a more general classification of collocation points can be
handled by using separate components). The models available depend on the solution task:
a) Trim: zero, uniform inflow, nonuniform inflow.

b) Transient: zero, uniform inflow, nonuniform inflow, dynamic inflow, lagged inflow, trim.
c) Flutter: zero, uniform inflow, dynamic inflow, lagged inflow, trim.
For the trim task, the model can be specified for each level of a designated wake loop. For the “trim” option in the
transient task, the inflow component calculates the velocity from saved trim parameters.
Figure 1 illustrates the functionality of the wing inflow component.
Degrees of Freedom
The component can have one degree of freedom vector: the dynamic inflow or lagged inflow variable.
260 WING INFLOW COMPONENT
aero interfaces at collocation points

degrees of
wing set force and moment interference velocity
freedom
wing set velocity
wing plane normal
sensors
Figure 22-1 Functionality of wing inflow component.

WING INFLOW COMPONENT 261
Component Input
a) Wing set force and moment (S or T axes).
b) Wing set velocity (S or T axes).
c) Wing set plane normal, k I .
Component Output
I
a) Aerodynamic interfaces at collocation points: interference velocity of air, vA .
b) Sensor: uniform inflow parameters.
22-3.1 Geometry and Frames
The component uses velocity and forces in wing set axes S, or the wing tip axes T, depending on the definition
of the input variables. The general convention is that the origin of the axes is at the center of the wing set, with
the x–y plane the wing set “plane”. The z-axis is up (positive lift direction), the x-axis is aft (positive velocity
direction), and the y-axis is to the right. These conventions are used to define the force and velocity components
of the wing set. It is assumed that the lift and induced velocity are in the z-axis direction of the axes used (S or T).
Dimensionless quantities are used in the analysis, based on the air density ρ, wing area S, and wing speed
Vx . The reference velocity Vx is the x-axis component of the wing set velocity. The wing area S is only used for
normalization. The wing set span b however is a parameter of the aerodynamic analysis, determining the induced
velocity.
22-3.2 Component Input
a) Wing set force and moment: the total aerodynamic load (force F and moment M ) acting on the entire wing set,
in S or T axes. The moments are about the origin of the wing set frame, and the loads are summed over all wings.
The summation and transformation can be performed by a reference frame component. The moment is not used
by this component.
b) Wing set velocity: the wing set velocity relative the air v, in S or T axes. Typically this is the velocity of the
origin of the wing set frame, from a reference frame component.
c) Wing set plane normal: k I , from a reference frame component.
A filter component can be used to get the mean input for the uniform inflow model. The input for the dynamic
inflow or lagged inflow model should not be filtered (the filter model should be direct). A filtered k I will not be a
unit vector.
22-3.3 Uniform Inflow Model
The uniform inflow model is based on ideal-wing theory, with a correction to account for nonideal induced
losses. It is assumed that the induced velocity is in the direction of the z-axis of the wing set axes being used (S
or T axes). For collocation points off the wing set, the induced velocity is obtained from the product of the mean
inflow at the wing set, and an empirical factor. The inflow model is not valid at very low speed.
22-3.4 Dynamic Inflow Model
The basic assumption of the dynamic inflow model is that the total wing forces vary slowly enough (compared
to the wake response) that the classical ideal-wing results are applicable to the perturbation as well as the trim
velocities. The dynamic inflow model is a low frequency approximation, although it can handle faster loading
variations than would be appropriate with the uniform inflow model. As implemented, the dynamic inflow model
calculates perturbation velocities, to be added to the trim solution. The dynamic inflow theory represents the
wake-induced velocity perturbation by the degree of freedom
δλ = δλu
(constant over the wing set); and represents the unsteady loading by the perturbation aerodynamic lift
δL = δCL
Then δλ is obtained from δL by a first order differential equation:

∂λ
τ δ λ̇ + δλ = δL
∂L
It is also necessary to consider the induced velocity perturbations produced by wing set velocity changes. This is
a low order, global model of the wing set unsteady aerodynamics, representing low frequency effects. Note that
a differential equation implies a lift deficiency function that is a ratio of polynomials, which is never the correct
form for a finite-order model.
The static derivative (∂λ/∂L) can be obtained from differential ideal-wing theory. The time lag is written
τ = κ(∂λ/∂L), where κ is a constant that can be obtained from the apparent mass of an impermeable surface
subject to linear motion. Omitting the time lag produces a quasistatic model, the effects of which are given by a
constant lift-deficiency function. However, frequently the time lag is needed to properly represent the effects of
the unsteady aerodynamics.
22–4 Theory
22-4.1 Force and Velocity
From the velocity of the wing set relative the air v, the velocity components are:
⎛ ⎞
−Vx
v = ⎝ Vy ⎠
Vz
So Vx is the velocity of the air seen by the wing. From the total aerodynamic force F :
⎛⎞
X
F =⎝Y ⎠
L
the dimensionless lift coefficient is:

L
CL = 1/2ρSV 2
x
The magnitude of the wing velocity, |Vx |, is used to normalize the wake-induced velocity.
For the uniform inflow model, the wake-induced velocity is obtained from the ideal-wing result, with an
empirical correction factor:
vi CL
λi = =
|Vx | πeAR
where AR = b2 /S is the wing set aspect-ratio. This equation is obtained from solution for the minimum induced-
drag CDi , and the interpretation λi = CDi /CL . The empirical factor e < 1 is included to account for the non-ideal
losses. The dimensional form of the equation is:
L
vi =
πe 1/2ρb2 |Vx |
where b is the wing set span. It is assumed that the velocity is in the direction of the z-axis of the wing axes being
used (S or T axes). Thus for collocation points on the wing set, the wake-induced velocity is:
I
vA = vi (−I )
where I is the unit vector obtained from the wing set plane normal k I . For collocation points off the wing set,
the wake-induced velocity is:
I
vA = Kvi (−I )
Each collocation point off the wing set has an empirical factor K, which is the ratio of the velocity at the collocation
point to the mean inflow at the wing set. Thus K accounts for effects such as the faction of the aerodynamic
component at the collocation point that is affected by the wake (less than 1, zero for no interference); and the
fraction of fully developed wake velocity that is achieved at the collocation point (maximum about 2).
Dynamic inflow can be used in the transient and flutter tasks. The uniform inflow parameters from the trim
task are saved. Thus the perturbation component input can be calculated, by subtracting the saved trim values from
the current values:
δVx = Vx − (Vx )trim
δCL = CL − (CL )trim
The direction of the total induced velocity (trim plus perturbation) is given by the instantaneous wing plane normal
vector I (obtained from k I ). The dynamic inflow model uses the trim wing velocity Vx . These perturbation loads
I
and motion are used to calculate the perturbation wake-induced velocity, δvA , which must be added to the trim
value. The wake-induced velocity perturbation is represented by a uniform change over the wing set:
δλ = δλu
so the degree of freedom vector is

ξ = ( δλu )
The aerodynamic model relates this inflow component to the perturbation lift and velocity of the wing set. Including
a time lag in the inflow response to loading changes, the dynamic inflow equation is:
Mb ˙ L
ξ + ξ = L δC
4|Vx |trim
where the effects of velocity perturbations are included through
L = δCL − CLtrim δVx

δC
(Vx )trim
Except for time, the dynamic inflow equation is dimensionless. The time lag is given by
8
M=
3π
The constant L relating the quasistatic inflow changes to quasistatic loading changes is obtained from a perturbation
form of the ideal-wing model (the uniform inflow model used for trim):
∂λ 1
L= =
∂L πeAR
The equation of motion is solved for the degrees of freedom ξ, then the induced velocity can be evaluated.
Dimensional velocities are obtained by multiplying the inflow ratio by the reference velocity |Vx |. The direction
of the total induced velocity (trim plus perturbation) is given by the instantaneous wing plane normal vector I
(obtained from k I ). The dynamic inflow model uses the trim wing velocity Vx . Thus for collocation points on the
wing set, the total wake-induced velocity is:
I
vA = [(vi )trim + |Vx |trim (δλu )] (−I )
For collocation points off the wing set, the total wake-induced velocity is:
I
vA = [(Kvi )trim + |Vx |trim (K δλu )] (−I )
using the empirical factor K of the uniform inflow model.
22-4.4 Lagged Inflow Model
For transient conditions there will be a lag in the change of the inflow in response to loading and velocity
changes. This time lag is modelled by using a first order differential equation to calculate the inflow. As for the
dynamic inflow model, the wake-induced velocity perturbation is δλ = δλu , so the degree of freedom vector is
ξ = (δλu ). The induced velocity vector λQS = (λu )QS is calculated from the instantaneous wing loading and
velocity, using the uniform inflow model. The corresponding trim value is λtrim . Then the equation of motion is
τ ξ˙ + ξ = λQS − λtrim
The time constant is specified in seconds; or in terms of wing spans (τ = τspan (b/|Vx |)trim ). The wake-induced
I
velocity vA , on and off the wing set, is evaluated from the trim and perturbation inflow as for the dynamic inflow
model.
ξ force velocity normal
motion 1 yes yes no
interference velocity 0 yes (a) yes
sensor no yes
Notes:
a) no if dynamic inflow or lagged inflow model for transient and flutter
Figure 22-2 Functionality of wing inflow component.

Chapter 23
ROTOR INFLOW COMPONENT
23–1 Description
A rotor inflow component calculates the wake induced velocity of a wing set. The wing set consists of all
blades of a rotor. It is called a “uniform inflow” component, even though the velocity can vary over the rotor disk.
There may be a corresponding nonuniform inflow component; which component gives a nonzero velocity result
then depends on the specified model, and perhaps on the wake loop level.
The induced velocity is calculated from a global, quasistatic model for a set of wings. This is an empirical
model based on an ideal-wing theory. The induced velocity is calculated at collocation points on this wing set,
and the interference velocity at other points can be calculated. The induced velocity at the wing set is described
by a mean term, plus a linear variation over the rotor disk (produced by edgewise flow or hub moments). The
influence of the ground is included. The aerodynamic model is quasistatic, designed to be used with the mean
(filtered) values of the component input. Optionally the induced velocity can be calculated for a ducted fan; or
using differential momentum theory (intended for axial flow).
The induced velocity in transient or flutter tasks can be obtained from dynamic inflow, which is a global, low
frequency model. Mean and linear induced velocity components are related, by first order differential equations,
to the net aerodynamic thrust and hub moments on the rotor. The differential equation (linearized inflow model,
with time lags) can be obtained from either a perturbation of the uniform inflow model, or from an actuator disk
model. Alternatively, differential equations can be constructed by simply introducing time lags in the uniform
inflow model.
inflow, lagged inflow, and trim. The dynamic inflow or lagged inflow model calculates perturbations from trim
(which must be a uniform or nonuniform model). For the nonuniform inflow model, it is possible to still use
uniform inflow for collocation points off the wing set (a more general classification of collocation points can be
handled by using separate components). The models available depend on the solution task:

b) Transient: zero, uniform inflow, nonuniform inflow, dynamic inflow, lagged inflow, trim.
c) Flutter: zero, uniform inflow, dynamic inflow, lagged inflow trim.
transient task, the inflow component calculates the velocity from saved trim parameters.
268 ROTOR INFLOW COMPONENT
Figure 1 illustrates the functionality of the rotor inflow component.
Degrees of Freedom
The component can have one degree of freedom vector: the dynamic inflow or lagged inflow variables.
Component Input
a) Aerodynamic interfaces at collocation points on wing set: location relative wing set frame (S or T axes).
b) Wing set force and moment (S or T axes).
c) Duct force and moment (S or T axes).
d) Wing set velocity and angular velocity (S or T axes).
e) Wing set plane normal, k I .
f) Normal to ground, kg (S or T axes).
g) Height origin of wing set above ground level.
h) Tip plane rotation rates.
i) Bound circulation: Γ at all span stations, for each wing.
Component Output
I
b) Sensor: uniform inflow parameters.
The component uses velocity and forces in wing set (shaft) axes S, or the wing tip (tip-plane) axes T, depending
on the definition of the input variables. The general convention is that the origin of the axes is at the center of the
wing set, with the x–y plane the wing set “plane”. The z-axis is up (positive lift direction), the x-axis is aft (positive
velocity direction), and the y-axis is to the right. These conventions are used to define the force, moment, and
velocity components of the wing set. It is assumed that the thrust and induced velocity are in the z-axis direction
of the axes used (S or T). The orientation of the axes is the same for both clockwise and counter-clockwise rotation
of the blades. The direction of rotation must be specified however, since it affects the sign of the lateral inflow
gradient.
Dimensionless quantities are used in the analysis, based on the air density ρ, rotor radius R, and rotor rotational
speed Ω (from a specified period; the absolute value is used if the rotational speed is negative). The tip speed is ΩR,
and the disk area is A = πR2 . The rotational speed Ω is only used for normalization, but R and ρ are parameters
of the aerodynamic analysis. The radius R defines the wing set span and the disk loading, which determine the
induced velocity in forward flight and axial flight respectively.
a) Location of collocation points on wing set: position r relative origin of wing set frame (S or T axes).
b) Wing set force and moment: the total aerodynamic load (force F and moment M ) acting on the entire wing set,
ROTOR INFLOW COMPONENT 269

degrees of
location rel wing set frame interference velocity
freedom
wing set force and moment
duct force and moment
wing set velocity
wing set angular velocity
wing plane normal
normal to ground
tip plane rotation rates
height origin above ground sensors
bound circulation
Figure 23-1 Functionality of rotor inflow component.

The summation and transformation can be performed by a reference frame component.
c) Duct force and moment: the aerodynamic load (force F and moment M ) acting on the duct, with the same
axes and moment origin as the wing set force and moment. The duct loads can be calculated by an aerodynamic
component, and the transformation to the required axes can be performed by a reference frame component.
d) Wing set velocity and angular velocity: the wing set velocity relative the air v and angular velocity ω, in S or T
axes. Typically this is the motion of the origin of the wing set frame, from a reference frame component.
e) Wing set plane normal: k I , from a reference frame component.
f) Normal to ground: kg , in S or T axes, for the ground effect calculation; from a reference frame component.
g) Height origin of wing set above ground level: zg , for the ground effect calculation; from a reference frame
component.
h) Tip plane rotation rates: (β̇x , β̇y ) relative wing set, from a reference plane component.
i) Bound circulation: Γ at all span stations, for each wing.
A filter component can be used to get the mean input for the uniform inflow model. The location of the
collocation points should not be filtered if they have significant motion relative the wing set origin. The input for
the dynamic inflow or lagged inflow model should not be filtered (the filter model should be direct). A filtered k I
or kg will not be a unit vector.
The uniform inflow model is based on momentum theory for a rotor, with corrections to account for the
following effects:
a) nonideal induced losses in hover and forward flight;

b) singularity of momentum theory at ideal autorotation;
c) vortex ring state;
d) circulation loop divergence for a hovering rotor near zero thrust;
e) ground effect;
f) linear inflow variation over the rotor disk, from edgewise flow and hub moments and angular
velocity.
Optionally the induced velocity can be calculated for a ducted fan; or using differential momentum theory (intended
for axial flow). It is assumed that the induced velocity is in the direction of the z-axis of the wing set axes being
used (S or T axes). For collocation points off the wing set, the induced velocity is obtained from the product of
the mean inflow at the wing set, and an empirical factor.
The basic assumption of the dynamic inflow model is that the total wing forces vary slowly enough (compared
to the wake response) that the classical ideal-wing results are applicable to the perturbation as well as the trim
velocities. The dynamic inflow model is a low frequency approximation, although it can handle faster loading
variations than would be appropriate with the uniform inflow model. As implemented, the dynamic inflow model
calculates perturbation velocities, to be added to the trim solution. The dynamic inflow theory represents the
wake-induced velocity perturbation by the degrees of freedom
⎛ ⎞
δλu
δλ = ⎝ δλx ⎠
δλy
(the terms giving uniform and linear variation over the rotor disk); and represents the unsteady loading by the
perturbation aerodynamic thrust and hub moments:
⎛ ⎞
δCT
δL = ⎝ δCM y ⎠
δCM x
Then δλ is obtained from δL by a first order differential equation:

∂λ
τ δ λ̇ + δλ = δL
∂L
It is also necessary to consider the induced velocity perturbations produced by wing set velocity changes. This is
a low order, global model of the wing set unsteady aerodynamics, representing low frequency effects. Note that
a differential equation implies a lift deficiency function that is a ratio of polynomials, which is never the correct
form for a finite-order model.
The static derivative matrix (∂λ/∂L) can be obtained from differential momentum theory (which gives good
results for hover), or from unsteady actuator disk theory (which is needed for good results in forward flight). The
time lag is written τ = κ(∂λ/∂L), where κ is a constant, diagonal matrix. The terms in κ can be obtained from the
apparent mass of an impermeable disk subject to linear or angular motion; the resulting values are supported by
experimental data. Omitting the time lag produces a quasistatic model, the effects of which are given by a constant
lift-deficiency function C . If the dominant aerodynamic forces are lift perturbations caused by angle-of-attack
changes, then the aerodynamic influence is described by a blade Lock number γ (which contains the lift-curve
slope). In this case the effects of the quasistatic dynamic inflow model are largely represented by an effective Lock
number: γ ∗ = γC . However, frequently the time lag is needed to properly represent the effects of the unsteady
aerodynamics.
23–4 Theory
23-4.1 Force and Velocity
From the velocity of the wing set relative the air v and the wing set angular velocity ω, the dimensionless
components are calculated as follows:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
−μx β̇ ω
v ω 1 ⎝ x⎠ ⎝ x⎠
= ⎝ μy ⎠ + β̇y = ωy
ΩR Ω Ω
μz 0 ωz
The tip plane rates are included, so ωx and ωy are the angular velocities of the tip plane. The advance ratio μ and
inflow ratio λ are defined as
μ = μ2x + μ2y
λ = λi + μz
where λi is the wake-induced inflow ratio. These are the inplane and normal components of the total velocity of
the air seen by the rotor disk. From the total aerodynamic force F and moment M :
⎛ ⎞ ⎛ ⎞
H Mx
F =⎝Y ⎠ M = ⎝ My ⎠
T Mz
the dimensionless thrust, roll moment, and pitch moment coefficients are:
T
CT =
ρA(ΩR)2
Mx
CM x =
ρA(ΩR)2 R
My
CM y =
ρA(ΩR)2 R
For ground effect, the dimensionless height above the ground:
zg
z=
R
is required. The position r of the collocation point relative the origin of the wing set (the rotor hub) gives
⎛ ⎞
xR
r = ⎝ yR ⎠
zR
where x and y are the dimensionless coordinates of the collocation point on the rotor disk.
23-4.2.1 Mean Inflow
For the uniform inflow model, the wake-induced velocity is obtained from the momentum theory result of
Glauert:
CT sλ2
λi = = h
2 λ2 + μ2 λ2 + μ2
where λ = λi + μz , λ2h = |CT |/2 (λh is always positive), and s = sign CT . This expression is generalized to
λi = κ λh s F (μ/λh , sμz /λh )
including the empirical factor κ to account for the effects of nonuniform inflow, tip losses, swirl, blockage, and
other phenomenon that increase the induced power losses (κ > 1). This factor is evaluated from κh for axial flow
and κf for edgewise flow:
μ2 κf + 4λ2 κh
κ=
μ2 + 4λ2
If μ is zero, the equation for λi can be solved analytically. Otherwise, for non-axial flow, the equation is written
as follows:
sλ2
λ= h + μz
λ2 + μ2
Using λ instead of λi as the independent variable simplifies implementation of the ducted fan model. A Newton-
Raphson solution for λ gives:
2
1in = sλh
λ
λ2n + μ2
λn − μz − λ1in
λn+1 = λn − f
1in λn /(λ2 + μ2 )
1+λ n
A relaxation factor of f = 0.5 is used to improve convergence. Three or four iterations are usually sufficient,
using
sλ2h
λ∼ = + μz
(sλh + μz )2 + μ2
to start the solution. To eliminate the singularity of the momentum theory result at ideal autorotation, the expression
% &
.373μ2z + .598μ2
λ = μz − .991
λ2h
is used when
1.5μ2 + (2sμz + 3λh )2 < λ2h
The equation λ = μz (aμ2z − bλ2h + cμ2 )/λ2h is an approximation for the induced power measured in the turbulent-
wake and vortex-ring states. Matching this equation to the axial-flow momentum theory result at μz = −2λh and
√ √
μz = −λh gives a = 5/6 = .3726780 and b = (4 5 − 3)/6 = .9907120. Then matching to the forward-flight
momentum theory result at (μ = λh , μz = −1.5λh ) gives c = .5980197. For axial flow (μ = 0) the solution is:
⎧ 4
⎪
⎪ μz μz 2
⎪
⎪ + s + λ2h −λh < sμz
⎪
⎪ 2 2
⎪
⎪ % &
⎨ .373μ2z
λ = μz − .991 −2λh < sμz < −λh
⎪
⎪ λ2h
⎪
⎪ 4
⎪
⎪
⎪
⎪ μ μz 2
⎩ z −s − λ2h sμz < −2λh
2 2
The inflow evaluated by the above algorithm is multiplied by κ to obtain λi .
23-4.2.2 Hovering Rotor Near Zero Thrust
For a hovering rotor near zero thrust, a modified inflow calculation is required to achieve convergence of a
circulation iteration. Uniform inflow gives for hover
1/2
CT = c1 θ − c2 λ = c1 θ − c3 CT = A(CT )
c3 −1/2
A (CT ) = − CT
2
so A = ∞ at zero thrust, and the iteration does not converge. This problem can be avoided by fixing the wake
geometry in the calculation of the hovering rotor induced-velocity at low thrust. Thus for uniform inflow, the
induced velocity can be calculated instead from the following equation:
κh CT
λi =
2 (μz + λ0 )2 + μ2

where λ0 = CT 0 /2 and CT 0 is a fixed, nominal value of the rotor thrust. For hover the inflow is
κh CT
λi =
2λ0
So fixing the denominator in this fashion is equivalent to fixing the wake geometry with the vertical convection rate
λ0 . Then A = −c4 /λ0 is finite (and independent of CT ), and the circulation iteration will converge for a small
enough nonzero relaxation factor. The resulting induced velocity is not accurate, but it is acceptable for cases of
small thrust, as long as a reasonable value of CT 0 is used. For nonuniform inflow, the equivalent approach is to
fix the wake geometry directly by using input values of the constants K1 , K2 , K3 , and K4 (rather than calculating
the wake geometry based on the rotor thrust).
23-4.2.3 Ground Effect
The wake-induced velocity is reduced when the rotor disk is in the proximity of the ground plane. The effect
of the ground is accounted for using the following approximate expression (from ref. 1) for the ratio of the induced
velocities in and out of ground effect:
v∞ T 1
= =
v T∞ 1 − (cos /4z)2
Here z is the height of the origin of the wing set (the rotor hub) above ground level, normalized by the semispan
(rotor radius); and is the angle between the ground and the rotor wake:
zg
z=
R
kgT v
cos =
|kg | |v|
⎛ ⎞
μx
v = ⎝ −μy ⎠
−λ
( = 0 for hover, and = 90 deg in forward flight). The cos factor accounts for the effect of forward speed;
no ground effect correction is applied if the wake is directed upward (cos ≤ 0). Note that ground effect is

essentially negligible for altitudes greater than the rotor diameter (z > 2), or at forward speeds μ > 2 CT /2.
This expression compares well with test results, down to an altitude of about one-half rotor radius (ref. 1). The
mean inflow in ground effect (IGE) is thus

cos2
(λi )IGE = 1− (λi )OGE
16z 2
in terms of the inflow out of ground effect (OGE).
23-4.2.4 Linear Inflow Variation
As a first approximation to the nonuniform induced velocity distribution, a linear variation over the disk
is used (for collocation points on the wing set): Δλ = λx x + λy y, where x and y are the coordinates of the
collocation point on the rotor disk. There are contributions to Δλ from forward flight, hub moments, and disk
angular velocity. The linear inflow variation caused by forward flight is
Δλf = λi (κx x + κy y)
where λi is the mean inflow. Typically κx is positive, and roughly 1 at high speed; and κy is smaller in magnitude
and negative (for a counter-clockwise rotating rotor). Both κx and κy must be zero in hover. Based on references
2 to 5, the following models are considered.
15π 15π μ
Coleman and Feingold: κx0 = tan χ/2 =
32 32 μ + λ2 + |λ|
2
κy0 = −2μg
√ √ μ
White and Blake: κx0 = 2 sin χ = 2
μ + λ2
2
κy0 = −2μg
where tan χ = |λ|/μ is the wake angle, and g accounts for the direction of rotation:
0
1 for +y on advancing side, counter-clockwise rotation
g=
−1 for +y on retreating side, clockwise rotation
Extending these results to include sideward velocity gives

Δλf = λi κx0 (x cos ψ0 − y sin ψ0 ) + κy0 (x sin ψ0 + y cos ψ0 )

= λi κx0 (xμx /μ − yμy /μ) + κy0 (xμy /μ + yμx /μ)

= λi (κx0 μx /μ + κy0 μy /μ)x + (−κx0 μy /μ + κy0 μx /μ)y
= λi (κx x + κy y)
Hence the coefficients of the two models are as follows.

15π μx
Coleman and Feingold: κx = fx − fy 2μy g
32 μ2 + λ2 + |λ|
15π μy
κy = −fx − fy 2μx g
32 μ2 + λ2 + |λ|
√ μx
White and Blake: κx = fx 2 − fy 2μy g
μ2 + λ2
√ μy
κy = −fx 2 − fy 2μx g
μ + λ2
2
For flexibility, the empirical factors fx and fy have been introduced (values of 1.0 give the baseline model). Other
options are to input κx and κy directly; or to omit these terms. There is also an inflow variation produced by any
net aerodynamic moment on the rotor disk, which can be evaluated using a differential form of momentum theory.
Finally, a linear inflow variation may be produced in response to pitch or roll rate of the rotor disk (ref. 6). These
two sources give
1
Δλm = (−fm 2CM y x + fm 2CM x y) = λxm x + λym y
λ + μ2
2
Δλr = fq 1.5ωy x − fp 1.5ωx y = λxr x + λyr y

including empirical factors fm , fp , and fq , which are defined in terms of values for hover and for forward flight:
ff μ2 + fh λ2
f=
μ2 + λ2
Note that the denominator of the hub moment term is zero for a hovering rotor at zero thrust; so this inflow
contribution should not be used for cases of low speed and low thrust. Adding the contributions from edgewise
flow, hub moments, and disk angular velocity gives
λx = λi κx + λxm + λxr
λy = λi κy + λym + λyr
the total gradients of the inflow.
23-4.2.5 Differential Momentum Theory
The induced velocity λi can alternatively be calculated using differential momentum theory (ref. 7). Then
λi is no longer uniform, but varies over the rotor disk. This model is intended for axial flow, but can be used in
edgewise flow as well. Consider a rotor system with disk area A, operating at speed V , with an angle α between
V and the disk plane. The induced velocity at the rotor disk is v, and the rotor thrust is T . The momentum theory
result for v in axial flow is
T = 2ρ A|(V +v)|v
The corresponding differential momentum theory result is
dT = 2ρ dA|(V +v)|vF
including a tip loss function F . Here dA = 2πr dr is the area of the annulus at radial station r, dT is the rotor thrust
acting on dA, and v is the induced velocity at r. The blade section load is obtained from the bound circulation:
dT
= L cos φ = (ρU Γ) cos φ = ρuT Γ = ρΩrΓ
dr
blades blades blades blades
(the last step is an approximation in edgewise flow). Considering forward flight as well, the differential momentum
theory result is then
Ω Γ = 4πU vF
blades
where U is the total velocity magnitude at the rotor disk:
U 2 = (V cos α)2 + (V sin α + v)2
The dimensionless result is

CΓ
λi =
2 λ2 + μ2
where the blade bound circulation gives
1
CΓ = Γ/ΩR2
2πF
blades
In this form, the solution for λi follows exactly as for the global momentum theory, with CT replaced by CΓ . The
solution for the inflow is multiplied by κ to obtain λi . Here λi at each radial station r is calculated from the bound
circulation at that radial station. Using the sum of the bound circulation on all blades, as in this expression, gives
a spanwise variation of the induced velocity. Alternatively, an induced velocity that varies azimuthally as well as
spanwise is obtained by using the bound circulation of the wing on which the collocation point is located (times
the number of blades). These two options will be referred to as “span differential momentum theory” and “disk
differential momentum theory.”
The differential momentum theory result for λi includes the corrections for nonideal induced losses in hover
and forward flight; the singularity of momentum theory at ideal autorotation; circulation loop divergence for a
hovering rotor near zero thrust; and ground effect. The linear inflow variation (λx and λy ) can still exist. With the
disk differential momentum theory, the linear inflow produced by hub moments should be suppressed. Differential
momentum theory is only used for collocation points on the wing set; the global momentum theory result for λi is
still used for collocation points off the wing set. The differential momentum theory solution is not available when
the ducted fan model is used.
Prandtl’s correction of actuator disk theory for a finite number of blades (ref. 7) provides the tip loss function
F used in differential momentum theory:
2
F = cos−1 e−πd/s
π
where d is the distance from the tip and s is the spacing between the wake sheets at the tip (both dimensionless).
Hence
πd N
= (1 − r) 1 + 1/λ2
s 2

with N the number of blades; r = x2 + y 2 from the position of the collocation point on the disk plane; and
λ = μz + λi given by global momentum theory. Alternatively, a tip loss factor B can be specified, so
πd 1−r
= ln 2
s 1−B
The tip loss function can also be omitted from the differential momentum theory result (F = 1). When the tip
loss correction is included in the calculation of the induced velocity, the tip loss factor B should not be used by
the wing component.
23-4.2.6 Wake-Induced Velocity
Dimensional velocities are obtained by multiplying the inflow ratio λ by the reference velocity ΩR (the tip
speed):
vi = λi ΩR
vx = λx ΩR
vy = λy ΩR
It is assumed that the velocity is in the direction of the z-axis of the wing axes being used (S or T axes). Thus for
collocation points on the wing set, the wake-induced velocity is:
I
vA = (vi + vx x + vy y) (−I )
I
vA = Kvi (−I )
Each collocation point off the wing set has an empirical factor K, which is the ratio of the velocity at the collocation
point to the mean inflow at the wing set. Thus K accounts for effects such as the fraction of the aerodynamic
component at the collocation point that is affected by the wake (less than 1, zero for no interference); and the
fraction of fully developed wake velocity that is achieved at the collocation point (maximum about 2). The
interference factor K is defined in terms of values for hover and for forward flight, to reflect the extremes of the
wake configuration relative the collocation point. Then K is obtained using:
Kf μ2 + Kh λ2
K = Kf cos2 χ + Kh sin2 χ =
μ2 + λ2
where tan χ = |λ|/μ is the wake angle.
23-4.3 Vortex Ring State
A rotor is operating in vortex ring state (VRS) when it is descending at low forward speed with a vertical
velocity that approaches the value of the wake-induced velocity at the rotor disk. In this condition the rotor tip
vortices are not convected away from the disk rapidly enough, and the wake builds up and periodically breaks away.
The tip vortices collect in a vortex ring, producing a circulating flow down through the rotor disk, then outward and
upward outside the disk. The resulting flow is unsteady, hence a source of considerable low frequency vibration
and possible control problems. For descent at forward speeds sufficiently high that the wake is convected away
from the rotor, vortex ring state does not develop. Vortex ring state encounter can produce a significant increase
in the descent rate of a helicopter or a roll-off of a tiltrotor. The boundary for such behavior is approximately
−1.5 < μz /λh < −0.5 and μ/λh < 1.0. This motion is an instability of the helicopter vertical or tiltrotor roll
dynamics. The instability is a consequence of the form of the rotor inflow as a function of descent rate. Measured
data show that at moderate descent rates (in VRS), the total inflow through the rotor disk λ = μz + λi increases as
the descent rate increases. As the rotor descends into VRS, the energy losses resulting from the recirculating flow
increase, hence the power (total inflow λ) can increase. Where dλ/dμz is negative (roughly μz /λh = −0.5 to
−1.5 for axial flow), the vertical motion (and roll motion of a tiltrotor) is unstable, because an increase in descent
rate at constant collective will produce an increase in total inflow and hence a reduction in thrust — negative
damping. The simple correction described above for the inflow in vortex ring and turbulent wake states eliminates
the singularity of momentum theory at ideal autorotation, but does not result in a negative slope of the inflow as a
function of descent rate.
A model of the rotor mean inflow that includes this character that leads to the unstable flight dynamics
in vortex ring state was developed in Reference 8. It is a parametric extension of momentum theory. This
extended vortex ring state model is an alternative to the simple correction above, used to evaluate the function F
in λi = κλh sF (μ/λh , μz /λh ).
Development of the VRS model begins by establishing a vortex ring state stability boundary as a function
of μ and μz , based on helicopter and tiltrotor flight test data. This stability boundary is where the inflow curve
has zero slope, dλ/dμz = 0. The equation used to define the boundary was chosen for convenience in the model
development. An inflow curve in VRS must be constructed that has zero slope on the specified boundary. The first
step is to eliminate the singularity of momentum theory at ideal autorotation in vertical descent. The result of this
step is referred to as the baseline model. The second step is to create the region of negative slope in vortex ring
state. For both steps, third order polynomials that provide the required behavior of the inflow as a function of μz
are identified. Figure 2 presents the algorithm used to calculate the rotor induced velocity v = λi /λh given values
of Vz = μz /λh and Vx = μ/λh . The following table summarizes the parameters of the model. All velocities in

the model are scaled with λh = CT /2. The algorithm assumes VzA > VzB ; VzD > VzN > VzX > VzE ; and
VzX ≥ VzA .
baseline model VRS model

point parameter value point parameter value
A VzA −1.5 D VzD −0.2
B VzB −2.1 N VzN −0.45
C VxC 0.75 X (Vz + v)N 0.85
X VzX −1.5
X (Vz + v)X 1.25
E VzE −2.0
M VxM 0.95
if Vz ≥ 0 or Vx ≥ VxC , momentum theory is used
baseline curve
end points
VzAID = VzA + 0.2(Vx /VxC )2
VzBID = VzB + 0.2(Vx /VxC )2
and if Vx /VxC > 0.5 then VzBID = VzBID + 0.7(VzAID − VzBID )(2Vx /VxC − 1)3
if VzBID < Vz < VzAID then identify v = bVz + cVz2 + dVz3
matching at VzAID : momentum theory v, dv/dVz
matching at VzBID : momentum theory v
otherwise momentum theory is used
VRS model: if Vz < 0 and Vx < VxM then

stability boundary
VzDID = VzD
VzN ID = 0.5(VzN + VzX ) + 0.5(VzN − VzX )(1 − (Vx /VxM )2 )0.2
VzXID = 0.5(VzN + VzX ) − 0.5(VzN − VzX )(1 − (Vx /VxM )2 )1.5
VzEID = VzE + (VzXID − VzX )
if VzEID < Vz < VzDID then identify Δ(Vz + v) = a + bVz + cVz2 + dVz3
if VzN ID ≤ Vz < VzDID then
match at VzDID : Δ(Vz + v) = dΔ(Vz + v)/dVz = 0
match at VzN ID : Δ(Vz + v) = ((Vz + v)N − (VzN + vN mom ))(1 − (Vx /VxM )6 )0.5
dΔ(Vz + v)/dVz = −(1 + dvbase /dVz )
if VzXID ≤ Vz < VzN ID then
match at VzN ID : as above
match at VzXID : Δ(Vz + v) = ((Vz + v)X − (VzX + vXmom ))(1 − (Vx /VxM )6 )0.5
dΔ(Vz + v)/dVz = −(1 + dvbase /dVz )
if VzEID < Vz < VzXID then
match at VzXID : as above
match at VzDID : Δ(Vz + v) = 0 (not matching slope, so a = 0)
otherwise momentum theory is used
baseline curve slope

dvbase /dVz = dv/dVz of momentum theory or
dvbase /dVz = b + 2cVz + 2cVz2
Figure 23-2 Algorithm to calculate the rotor induced velocity in vortex ring state.
To eliminate the singularity of momentum theory, points A and B are identified on the two branches of
momentum theory for a given Vx , and connected by a third-order polynomial. The coefficients of the polynomial
are identified by matching v and dv/dVz at A, and v at B (with the constant term of the polynomial set to zero).
As Vx increases the points A and B are moved together, so the momentum theory result is used entirely when Vx
is beyond a point C. It is necessary to shape the variation of the points A and B with Vx such that the polynomial
is well behaved (in particular, move the points to the right with increasing Vx , so the polynomial is matched to
the right of the momentum theory peak, where dv/dVz < 0). The stability boundary is specified by the points X
and N for vertical descent, and a point M in forward flight: VzN , VzX , VxM . Appropriate functions are used to
generate a reasonable shape of the boundary, in terms of the variation of the points X and N with Vx . The VRS
model requires an increment Δ(Vz + v) relative the baseline model, defined by the inflow values (Vz + v)N at the
minimum and (Vz + v)X at the maximum. Points D and E are specified, where the VRS model joins the baseline
curve. For each of the three segments of the VRS model (D to N, N to X, X to E) a third order polynomial for
Δ(Vz + v) as a function of Vz is identified by matching v and dv/dVz (except that for X to E the slope is not
matched at E, so the constant term of the polynomial is zero). The final inflow value is v = vbase + f Δvvrs . The
factor f allows the instability in VRS to be reduced or suppressed (0 ≤ f ≤ 1).
23-4.4.1 Perturbation Force and Velocity
Dynamic inflow can be used in the transient and flutter tasks. The uniform inflow parameters from the trim
task are saved. Thus the perturbation component input can be calculated, by subtracting the saved trim values from
the current values: ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
δμx μx μx
⎝ δμy ⎠ = ⎝ μy ⎠ − ⎝ μy ⎠
δμz μz μz trim

δωx ωx ωx
= −
δωy ωy ωy trim
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
δCT CT CT
⎝ δCM y ⎠ = ⎝ CM y ⎠ − ⎝ CM y ⎠
δCM x CM x CM x trim
δz = z − ztrim
vector I (obtained from k I ). The dynamic inflow model uses the trim rotor speed Ω. These perturbation loads
I
and motion are used to calculate the perturbation wake-induced velocity, δvA , which must be added to the trim
value.
23-4.4.2 Aerodynamic Model
The wake-induced velocity perturbation is represented by terms giving uniform and linear variation over the
rotor disk:
δλ = δλu + δλx x + δλy y
so the degree of freedom vector is ⎛ ⎞

δλu
ξ = ⎝ δλx ⎠
δλy
The aerodynamic model relates these inflow components to the perturbation loads and motion of the wing set.
Including a time lag in the inflow response to loading changes, the dynamic inflow equations are:
⎛ ⎞
δCT
1 M y ⎠
LM ξ˙ + ξ = L ⎝ −δ C
Ω
δ CM x
where the effects of velocity perturbations are included through

% &

δCT = κ δCT − CT trim μx δμx + μy δμy + λ δμz
λ2 + μ2
M y = fm δCM y − 1 2
δC λ + μ2 λi δκx
2
M x = fm δCM x + 1 2
δC λ + μ2 λi δκy
2
(refs. 9 and 10). The δκ terms in the moments give velocity perturbations:
Coleman and Feingold:
15π/32
δκx = fx μ2y + λ(λ + signλ μ2 + λ2 ) δμx − μx μy δμy
( μ2 + λ2 + |λ|)2 μ2 + λ2

−μx (λ + signλ μ2 + λ2 ) δμz − fy 2g δμy
15π/32
δκy = −fx −μx μy δμx + μ2x + λ(λ + signλ μ2 + λ2 ) δμy
( μ + λ + |λ|) μ + λ
2 2 2 2 2

−μy (λ + signλ μ2 + λ2 ) δμz − fy 2g δμx
White and Blake: √

2 2
δκx = fx μy + λ2 δμx − μx μy δμy − μx λ δμz − fy 2g δμy
2 2
( μ +λ ) 3
√
2
δκy = −fx −μx μy δμx + μ2x + λ2 δμy − μy λ δμz − fy 2g δμx
2 2
( μ +λ ) 3
actuator disk:
15π/32
δκx = μ2y + λ(λ + signλ μ2 + λ2 ) δμx − μx μy δμy
( μ2 + λ2 + |λ|)2 μ2 + λ2

−μx (λ + signλ μ2 + λ2 ) δμz
15π/32
δκy = − −μx μy δμx + μ2x + λ(λ + signλ μ2 + λ2 ) δμy
( μ2 + λ2 + |λ|)2 μ2 + λ2

−μy (λ + signλ μ2 + λ2 ) δμz
depending on the inflow model.
The effect of the ground on the inflow dynamics is to add a velocity change caused by changes in the rotor
height above the ground:
∂λ λi cos2
δz = δz
∂z 8z 3
from the perturbation of the uniform inflow model for ground effect. Ground effect introduces a positive spring to
the rotor motion: a decrease in the rotor height above the ground produces a decrease in the induced velocity, hence
a rotor thrust increase that acts to oppose the height change. Finally, a linear inflow variation may be produced in
response to pitch or roll rate of the rotor disk (ref. 6). Thus the equations of motion for dynamic inflow are:
⎛ T ⎞ ⎛ (∂λ/∂z) δz ⎞
δC
1 ⎜ M y ⎟ ⎜ ⎟
LM ξ˙ + ξ = L ⎝ −δ C ⎠ + ⎝ fq 1.5δωy ⎠
Ω
δCM x −fp 1.5δωx
including the ground effect and disk rate terms. The dynamic inflow model includes additional factors on the
parameters fx , fy , fm , fp , and fq , so the inflow gradient terms can be suppressed without affecting the trim
solution.
The matrices M and L are required to complete the equations. Except for time, the dynamic inflow equations
are dimensionless. The time lag matrix M is diagonal:
⎡ 8 ⎤
0 0
⎢ 3π ⎥
⎢ 16 ⎥
M =⎢ ⎢ 0 0 ⎥⎥
⎣ 45π ⎦
16
0 0
45π
(refs. 11 to 13). Two models are implemented for the matrix L relating the quasistatic inflow changes to quasistatic
loading changes: a perturbation form of the momentum theory model (the uniform inflow model used for trim);
and an actuator disk model (ref. 10). For the momentum theory model:
⎡ ∂λ ⎤
0 0
⎢ ∂T ⎥
⎢ ⎥
⎢ ∂λ ∂λ ⎥
L = ⎢ κx 0 ⎥
⎢ ∂T ∂M ⎥
⎣ ⎦
∂λ ∂λ
κy 0
∂T ∂M
where % &−1
∂λ CT λ
= 2 λ2 + μ2 + 2
∂T λ + μ2
∂λ 2
=
∂M λ + μ2
2
and κx , κy are obtained from the uniform inflow model of trim. For the actuator disk theory model:
⎡ ∂λ ∂λ ∂λ ⎤
−κx −κy
⎢ ∂T ∂T ∂T ⎥
⎢ ⎥
⎢ ∂λ ∂λ ∂λ ⎥
L = ⎢ κx ⎥
⎢ ∂T ∂M ∂N ⎥
⎣ ⎦
∂λ ∂λ ∂λ
κy
∂T ∂N ∂L
where
15π μx
κx =
32 μ + λ2 + |λ|
2
15π −μy
κy =
32 μ + λ2 + |λ|
2
μ2 + (λ + λi )λ
V =
μ2 + λ2
∂λ 1
=
∂T 2V" #
∂λ 1 4|λ| 4μ2y
= +
∂M V μ2 + λ2 + |λ| ( μ2 + λ2 + |λ|)2
" #
∂λ 1 4 μ2 + λ2 4μ2y
= −
∂L V μ2 + λ2 + |λ| ( μ2 + λ2 + |λ|)2
" #
∂λ 1 4μy μx
=
∂N V ( μ2 + λ2 + |λ|)2
from reference 10, generalized to include the effects of sideward velocity.
23-4.4.3 Wake-Induced Velocity
The equation of motion is solved for the degrees of freedom ξ, then the induced velocity can be evaluated.
Dimensional velocities are obtained by multiplying the inflow ratio by the reference velocity ΩR (the tip speed).
vector I (obtained from k I ). Thus for collocation points on the wing set, the total wake-induced velocity is:
I
vA = [(vi + vx x + vy y)trim + ΩR (δλu + δλx x + δλy y)] (−I )
For collocation points off the wing set, the total wake-induced velocity is:
I
vA = [(Kvi )trim + ΩR (K δλu )] (−I )
using the empirical factor K of the uniform inflow model.
23-4.5 Lagged Inflow Model
For transient conditions there will be a lag in the change of the inflow in response to loading and velocity
changes. This time lag is modelled by using a first order differential equation to calculate the inflow. As for the
dynamic inflow model, the wake-induced velocity perturbation is
δλ = δλu + δλx x + δλy y
so the degree of freedom vector is ⎛ ⎞

δλu
ξ = ⎝ δλx ⎠
δλy
The induced velocity vector λQS = (λu λx λy )TQS is calculated from the instantaneous rotor loading and velocity,
using the uniform inflow model. The corresponding trim value is λtrim . Then the equation of motion is
τ ξ˙ + ξ = λQS − λtrim
The time constant is specified in seconds; in terms of rotor revolutions (τ = τrev (2π/Ω)trim ); or scaled with the
I
hover inflow (τrev = τh /λh ). The wake-induced velocity vA , on and off the wing set, is evaluated from the trim
and perturbation inflow as for the dynamic inflow model.
23-4.6 Ducted Fan
23-4.6.1 Background
Rotor momentum theory (ref. 7) can be extended to the case of a ducted fan. Consider a rotor system with
disk area A, operating at speed V , with an angle α between V and the disk plane. The induced velocity at the
rotor disk is v, and in the far wake w = fW v. The far wake area is A∞ = A/fA . The axial velocity at the fan is
fV z Vz , with fV z accounting for acceleration or deceleration through the duct. The edgewise velocity at the fan
is fV x Vx , with fV x = 1.0 for wing-like behavior, or fV x = 0 for tube-like behavior of the flow. The total thrust
(rotor plus duct) is T , and the rotor thrust is Trotor = fT T . For this model, the duct aerodynamics are defined
by the thrust ratio fT or far wake area ratio fA , plus the fan velocity ratio fV . The mass flux through the rotor
disk is ṁ = ρAU = ρA∞ U∞ , where U and U∞ are the total velocity magnitudes at the fan and in the far wake
respectively:
U 2 = (fV x V cos α)2 + (fV z V sin α + v)2
2
U∞ = (V cos α)2 + (V sin α + w)2
Mass conservation (fA = A/A∞ = U∞ /U ) relates fA and fW . Momentum and energy conservation give
T = ṁw = ρAU∞ w/fA = ρAU fW v

1 w
P = ṁw (2V sin α + w) = T V sin α +
2 2
With these expressions, it is assumed that in forward flight the span of the lifting system equals the rotor diameter
2R; the factor κf accounts for the actual span of the lifting system. Next it is required that the power equals the
rotor induced and parasite loss:
P = Trotor (fV z V sin α + v) = T fT (fV z V sin α + v)
In axial flow, this result can be derived from Bernoulli’s equation for the pressure in the wake. In forward flight, any
induced drag on the duct is being neglected. From these two expressions for power, Vz +fW v/2 = fT (fV z Vz +v)
is obtained, relating fT and fW . With no duct (fT = fV x = fV z = 1), the far wake velocity is always w = 2v,
hence fW = 2. With an ideal duct (fA = fV x = fV z = 1), the far wake velocity is fW = 1. In hover (with or

without a duct), fW = fA = 2fT ; and v = 2/fW vh . The rotor ideal induced power is Pideal = T w/2 = fD T v,
introducing the duct factor fD = fW /2. It is assumed that there is some aerodynamic component that calculates
the duct force, so the rotor inflow component can calculate fT from the rotor and duct loads, and then fW and
fA as required. A simple model of the ducted fan might obtain the duct force from the rotor force and a specified
value of fT ; typically fA can be used to estimate fT for this approach.
23-4.6.2 Uniform Inflow Model
For a ducted fan, the thrust and moments (CT , CM y , CM x ) are calculated from the total load (rotor plus duct).
The thrust ratio fT is calculated from the rotor thrust and total thrust:
Trotor
fT =
T
and similarly for the moment ratios (fM y and fM x ). Once the induced velocity has been calculated, the wake
velocity and area ratios can be evaluated from:

μz
fW = 2 fT − (1 − fT fV z )
λi
5
μ2 + (μz + fW λi )2
fA =
(fV x μ)2 + (fV z μz + λi )2
The wake-induced velocity is obtained from the momentum theory result for a ducted fan:

λ2h = (fW λi /2) (fV x μ)2 + (fV z μz + λi )2
hence
sλ2h /fT μz
fV z μz + λi = +
(fV z μz + λi )2 + (fV x μ)2 fT
In this form, it is possible to solve for λi using the free-rotor expressions given above: replacing λ2h , μz , μ, λ
with respectively λ2h /fT , μz /fT , fV x μ, fV z μz + λi . However, physical problems and convergence difficulties
are encountered with this approach in descent, if an arbitrary value of fT is permitted. From the expression for
fT above, it follows that fT should approach 1/fV z at high rates of climb or descent. To avoid problems with an
arbitrary value of fT , it is assumed that the input value of fT defines the velocity ratio fW = 2fT in descent. So
in descent μz is not replaced by μz /fT .
23-4.6.3 Dynamic Inflow Model
Dynamic inflow can be used in the transient and flutter tasks. For a ducted fan, the trim and perturbation load
coefficients (CT , CM y , and CM x ) are calculated from the total load (rotor plus duct). The thrust ratio fT has the
trim value. The modified thrust perturbation, including the effects of the velocity perturbations, is now:
% &

T = κ δCT − CT trim μx δμx + μy δμy + λ δμz + 2 λ2 + μ2 1 − fT δμz
δC
λ2 + μ2
The terms in the matrix L are calculated using the free rotor expressions given above, except now
% &−1
∂λ 2 2
CT λ
= 2fT λ + μ + 2
∂T λ + μ2
for the momentum theory model; and

fT μ2 + (fT λ + λi )λ
V =
μ2 + λ2
for the actuator disk theory model. In these expressions, (fV x μ) and (fV z μz + λi ) are used for μ and λ.
23-4.6.4 Simple Duct Force Model
In general, some aerodynamic component calculates the forces acting on the duct. A simple model of the
ducted fan can be implemented by specifying the ratio of the rotor load to total load (f ). Then the duct loads can
be obtained from the rotor loads:
1
Lduct = − 1 Lrotor
f
A separate ratio f can be specified for each of the three components of force and three components of moment. A
wing component calculates the aerodynamic force and moment on the rotor blades; these loads are summed and
transformed to obtain the rotor force and moment, in the appropriate axes. Then a differential equation component
(static equations, for scalar summation) can be used to calculate the duct force and moment, in the same axes. This
duct load is the input to the rotor inflow component. The duct load must also be applied to the structure, acting on
a connection at the origin of the wing set axes.
23–5 References
1) Cheeseman, I.C., and Bennett, W.E. “The Effect of the Ground on a Helicopter Rotor in Forward Flight.” ARC
R & M 3021, September 1955.
2) Coleman, R.P.; Feingold, A.M.; and Stempin, C.W. “Evaluation of the Induced-Velocity Field of an Idealized
Helicopter Rotor.” NACA ARR L5E10, June 1945.
3) Mangler, K.W., and Squire, H.B. “The Induced Velocity Field of a Rotor.” ARC R & M 2642, May 1950.
4) Drees, J.M. “A Theory of Airflow Through Rotors and Its Application to Some Helicopter Problems.” Journal
of the Helicopter Association of Great Britain, Volume 3, Number 2, July-September 1949.
5) White, T., and Blake, B.B. “Improved Method of Predicting Helicopter Control Response and Gust Sensitivity.”
Annual National Forum of the American Helicopter Society, May 1979.
6) Keller, J.D. “An Investigation of Helicopter Dynamic Coupling Using an Analytical Model.” Journal of the
American Helicopter Society, Volume 41, Number 4, October 1996.
7) Johnson, W. Helicopter Theory. Princeton University Press, Princeton, New Jersey, 1980.
8) Johnson, W. “Model for Vortex Ring State Influence on Rotorcraft Flight Dynamics.” American Helicopter
Society 4th Decennial Specialist’s Conference on Aeromechanics, San Francisco, January 2004.
9) Johnson, W. “Aeroelastic Analysis for Rotorcraft in Flight or in a Wind Tunnel.” NASA TN D-8515, July 1977.
10) Gaonkar, G.H., and Peters, D.A. “Review of Dynamic Inflow Modelling for Rotorcraft Flight Dynamics.”
AIAA Paper Number 86-0845, May 1986.
11) Peters, D.A. “Hingeless Rotor Frequency Response with Unsteady Inflow.” NASA SP-352, February 1974.
12) Carpenter, P.J., and Fridovich, B. “Effect of a Rapid Blade-Pitch Increase on the Thrust and Induced-Velocity
Response of a Full-Scale Helicopter Rotor.” NACA TN 3044, November 1953.
13) Hohenemser, K.H., and Crews, S.T. “Unsteady Hovering Wake Parameters Identified from Dynamic Model
Tests.” NASA CR 152022, June 1977.
ξ force velocity normal zg wing set bound

duct force ang vel kg position circ
tip plane
motion 1 yes yes no yes no no
interference 0 yes (a) yes (b) yes (a) yes (c) no (d)
velocity
sensor no yes no no
Notes:
a) no if dynamic inflow or lagged inflow model for transient and flutter
b) no for kg if dynamic inflow or lagged inflow model for transient and flutter
c) yes only if collocation points of output and input are same
d) yes if differential momentum theory is used
Figure 23-3 Functionality of rotor inflow component.

Chapter 24
ROTOR DYNAMIC WAKE COMPONENT
24–1 Description
A rotor dynamic wake component calculates the wake induced velocity of a wing set. The wing set consists of
all blades of a rotor. The induced velocity is calculated at collocation points on this wing set, and the interference
velocity at other points can be calculated.
The induced velocity is calculated using the generalized dynamic wake theory developed by Peters and He
(refs. 1 and 2). This is an unsteady wake theory for lifting rotors, based on the acceleration potential for an actuator
disk. The induced inflow at the rotor disk is expressed in terms of a Fourier series azimuthally and polynomials
radially. The result is a system of first-order, ordinary differential equations for the inflow states. Such a finite-state
induced flow model is particularly suitable for rotorcraft aeroelastic analysis.
A nonlinear formulation of the theory is implemented, so the uniform induced velocity term produced by the
rotor thrust is the usual momentum theory result. In the trim task, the static solution for the inflow variables is
obtained (no degrees of freedom). In the transient and flutter tasks, the ordinary differential equations are solved.
Optionally, the induced velocity can be zero in the trim, transient, or flutter task; or the trim induced velocity can
be used in the transient or flutter task.
Figure 1 illustrates the functionality of the rotor dynamic wake component.
Degrees of Freedom
The component can have one degree of freedom vector: the dynamic wake states.
Component Input
a) Aerodynamic interfaces at collocation points on wing set:
1) Location relative wing set frame.
2) Aerodynamic force.
b) Wing set force and moment.
c) Wing set velocity and angular velocity.
d) Wing set plane normal, k I .
e) Tip plane rotation rates.
Component Output
I
290 ROTOR DYNAMIC WAKE COMPONENT

degrees of
location rel wing set frame interference velocity
freedom
aerodynamic force

wing set velocity
wing set angular velocity
wing plane normal
tip plane rotation rates
sensors
Figure 24-1 Functionality of rotor dynamic wake component.

ROTOR DYNAMIC WAKE COMPONENT 291
b) Sensor: dynamic wake parameters.
The component uses velocity and forces in wing set (shaft) axes S, or the wing tip (tip-plane) axes T, depending
on the definition of the input variables. The general convention is that the origin of the axes is at the center of
the wing set, with the x–y plane the wing set “plane”. The z-axis is up (positive lift direction), the x-axis is aft
(positive velocity direction), and the y-axis is to the right. These conventions are used to define the force and
velocity components of the wing set. It is assumed that the thrust and induced velocity are in the z-axis direction
of the axes used (S or T). The orientation of the axes is the same for both clockwise and counter-clockwise rotation
of the blades.
Dimensionless quantities are used in the analysis, based on the air density ρ, rotor radius R, and rotor rotational
speed Ω (from a specified period; the absolute value is used if the rotational speed is negative). The tip speed is ΩR,
and the disk area is A = πR2 . The rotational speed Ω is only used for normalization, but R and ρ are parameters
of the aerodynamic analysis. The radius R defines the wing set span and the disk loading, which determine the
induced velocity in forward flight and axial flight respectively.
a) Location of collocation points on wing set: position r relative origin of wing set frame (S or T axes).
b) Aerodynamic force at collocation points on wing set: force Fqp (S or T axes), for each blade (q) and each
collocation point (p). This is the discretized force (panel load), obtained from the product of the section force and
the aerodynamic panel width; acting at the midpoint of the aerodynamic panel.
c) Wing set force and moment: the total aerodynamic load (force F and moment M ) acting on the entire wing set,
d) Wing set velocity and angular velocity: the wing set velocity relative the air v and angular velocity ω, in S or T
axes. Typically this is the motion of the origin of the wing set frame, from a reference frame component.
e) Wing set plane normal: k I , from a reference frame component.
f) Tip plane rotation rates: (β̇x , β̇y ) relative wing set, from a reference plane component.
24-3.3 Dynamic Wake Model
The generalized dynamic wake theory of Peters and He (refs. 1 and 2) is based on the acceleration potential
for an actuator disk. The induced inflow at the rotor disk is expressed in terms of a Fourier series azimuthally and
polynomials radially, so the inflow velocity is defined by the degree of freedom vector α. The rotor loading is
represented by the generalized force vector τ . Then α is obtained from τ by a first order differential equation:
LM α̇ + α = Lτ
where the derivative matrix is L = ∂α/∂τ , and the diagonal mass matrix M gives the time lag LM . A nonlinear
formulation of the theory is implemented, so the uniform induced velocity term produced by the rotor thrust is
the usual momentum theory result (including corrections for nonideal induced losses in hover and forward flight,
and the singularity of momentum theory at ideal autorotation). It is assumed that the induced velocity is in the
direction of the z-axis of the wing set axes being used (S or T axes). For collocation points off the wing set, the
induced velocity is obtained from the product of the mean inflow at the wing set, and an empirical factor.
24–4 Theory
24-4.1 Geometry, Velocity, and Force
The geometry of the rotor disk is shown in figure 2. The axes are moving with the rotor disk, with the x-axis
downstream and the z-axis in the positive thrust direction. The coordinate ξ is aligned with the velocity vector,
positive upstream. The rotor wake is a skewed cylinder. From the velocity of the wing set relative the air v and
the wing set angular velocity ω, the dimensionless components are calculated as follows:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
−μx β̇x ωx
v ω 1
= ⎝ μy ⎠ + ⎝ β̇y ⎠ = ⎝ ωy ⎠
ΩR Ω Ω
μz 0 ωz
The tip plane rates are included, so ωx and ωy are the angular velocities of the tip plane. The advance ratio μ and
inflow ratio λ are defined as
μ = μ2x + μ2y
λ = λi + μz
where λi is the wake-induced inflow ratio. These are the inplane and normal components of the total velocity
of the air seen by the rotor disk. The total velocity magnitude is V∞ = |v/ΩR|. The wake skew angle is
χ = tan−1 |μ/λ|, measured from the z-axis; or α = tan−1 |λ/μ|, measured from the disk plane. While figure
2 only shows μz , the normal component of the free stream velocity, the wake skew angles are evaluated using
the total normal velocity λ, which includes the wake-induced inflow. The inplane velocity component μ has an
azimuth angle θ0 = − tan−1 (μy /μx ) in the disk plane.
The wake is shown below the rotor disk in figure 2. The case of the wake above the disk (λ < 0) is equivalent
to a rotor with the wake below the disk and negative loading, hence with the signs changed for both λ and Fz .
The relation of the induced velocity to the loading is linear, so both negative and positive loading are allowed.
It follows that the solution is independent of the sign of λ, hence the use of the absolute value in the definition
of the wake skew angles χ and α. The dynamic wake theory is derived using an axis system rotated about the
z-axis by the angle θ0 , hence for μy = 0. To apply the theory, the generalized forces must be evaluated in this
rotated axis system, and the azimuthal rotation by θ0 accounted for when the induced velocity is evaluated from
the inflow states. The theory is developed assuming that the velocity of the rotor disk relative the air is constant,
which implies that the geometry of the wake is also constant. The theory is applied in the transient and flutter tasks
using the instantaneous (probably filtered) values of the rotor disk velocity.
The dimensionless thrust, roll moment, and pitch moment coefficients are obtained from the total aerodynamic
force F and moment M : ⎛ ⎞ ⎛ ⎞
H Mx
F =⎝Y ⎠ M = ⎝ My ⎠
T Mz
V∞ ξ
y
velocity of
rotor
air relative
disk
x
rotor
μx
μy χ rotor
μz wake
ν=1
ν>0 ν>0
η>0
ν=0 η=0 ν=0

x
ν<0 ν<0
ν = –1
Figure 24-2 Geometry of rotor disk.

and
T
CT =
ρA(ΩR)2
Mx
CM x =
ρA(ΩR)2 R
My
CM y =
ρA(ΩR)2 R
The dimensionless lift Fz is obtained from the aerodynamic force at the collocation point:
⎛ ⎞
Fx
Fqp ⎝ Fy ⎠
=
ρΩ2 R4
Fz
This is the discretized force (panel load), obtained from the product of the section force and the aerodynamic panel
width; acting at the midpoint of the aerodynamic panel; for each blade (q) and each collocation point (p). The
position r of the collocation point relative the origin of the wing set (the rotor hub) gives
⎛ ⎞
xR
r = ⎝ yR ⎠
zR
where x and y are the dimensionless coordinates of the collocation point on the rotor disk.
24-4.2 Dynamic Wake Model
24-4.2.1 Equations and Boundary Conditions
Consider the incompressible, potential flow about the rotor disk, with geometry and velocity as shown in
figure 2. Dimensionless quantities are used in the analysis, based on the air density ρ, rotor radius R, and rotor
rotational speed Ω. The equations of mass and momentum conservation for small disturbances are:
∂qi
=0
∂xi
∂qi ∂qi ∂Φ
− V∞ =−
∂t ∂ξ ∂xi
where Φ is the perturbation pressure or acceleration potential; qi the perturbation velocity; and xi the flow field
coordinates. The convection term in the momentum equation has been written in terms of the coordinate ξ:
∂ ∂ ∂
μx − μz = −V∞
∂x ∂z ∂ξ
From the definition of skewed coordinates (x0 , y0 , ξ):

⎛ ⎞ ⎛ ⎞
x x0 − ξ sin χ
⎝y⎠ = ⎝ y0 ⎠
z ξ cos χ
there follows
1
dξ = dz
cos χ
∂ 1 ∂ ∂
= + tan χ
∂z cos χ ∂ξ ∂x
The acceleration potential satisfies Laplace’s equation, ∇2 Φ = 0. The boundary conditions are that the perturbation
pressure is zero at infinity, and there is a pressure discontinuity on the rotor disk. The solution will be obtained by
splitting the acceleration potential into convection and unsteady terms, Φ = ΦV + ΦA , such that
∂ΦV ∂qi
= V∞
∂xi ∂ξ
A
∂Φ ∂qi
=−
∂xi ∂t
It is assumed that these two contributions to Φ each satisfy Laplace’s equation: ∇2 ΦV = 0 and ∇2 ΦA = 0.
An analytical solution can be obtained for a circular wing or disk, by writing Laplace’s equation in ellipsoidal
coordinates, and expanding the potential in associated Legendre functions (spherical harmonics).
24-4.2.2 Ellipsoidal Coordinates and Legendre Functions
The cartesian coordinates (x, y, z) are shown in figure 2. The origin is at the center of the rotor disk, and x–y
is the disk plane. Figure 2 also shows the ellipsoidal coordinates (ν, η, θ), defined as follows:

x= 1 − ν 2 1 + η 2 cos θ

y = 1 − ν 2 1 + η 2 sin θ
z = νη
θ = tan−1 (y/x)

signz
ν= √ 1 − s + (1 − s)2 + 4z 2
2
s = x2 + y 2 + z 2 = 1 − ν 2 + η 2
η = z/ν
The range of the ellipsoidal coordinates is −1 ≤ ν ≤ 1, 0 ≤ η ≤ ∞, 0 ≤ θ ≤ 2π; with signν = signz. The vari-

able θ is the disk azimuth angle, measured from the x-axis. On the rotor disk, η = 0 so ν = ± 1 − (x2 + y 2 ) =
√ √
± 1 − r2 . In the x–y plane outside the rotor disk, ν = 0 so η = r2 − 1. On the z-axis, ν = ±1 so η = |z|.
In ellipsoidal coordinates (and dropping a factor of (ν 2 + η 2 )−1 ), Laplace’s equation ∇2 Φ = 0 is

% & % & % &
∂ ∂Φ ∂ ∂Φ ∂ ν 2 + η2 ∂Φ
(1 − ν 2 ) + (1 + η 2 ) + =0
∂ν ∂ν ∂η ∂η ∂θ (1 − ν 2 )(1 + η 2 ) ∂θ
Using the method of separation of variables, write Φ = Φ1 (ν)Φ2 (η)Φ3 (θ). Then
% & % &
∂ 2 ∂Φ1 m2
(1 − ν ) + − + n(n + 1) Φ1 = 0
∂ν ∂ν 1 − ν2
% & % &
∂ 2 ∂Φ2 m2
(1 + η ) + − n(n + 1) Φ2 = 0
∂η ∂η 1 + η2
∂ 2 Φ3
+ m2 Φ 3 = 0
∂θ2
The solution for Φ3 is periodic if m is an integer. The solution for Φ1 is finite for ν = ±1 (on the z-axis) if
n is an integer (n ≥ m). The solutions are Φ1 = Pnm (ν), the associated Legendre function of the first kind;
Φ2 = Qm n (iη), the associated Legendre function of the second kind; and Φ3 = sin(mθ) or cos(mθ).
The associated Legendre functions Pnm (ν) (n ≥ m) are related to Legendre polynomials Pn (ν):
dm
Pnm (ν) = (−1)m (1 − ν 2 )m/2 Pn (ν)
dν m
The following stable recurrence relation can be used to evaluate Pnm (ν):
m
Pm = (−1)m (2m − 1)!! (1 − ν 2 )m/2
m m
Pm+1 = ν (2m + 1) Pm
(k − m) Pkm = ν (2k − 1) Pk−1
m
− (k + m − 1) Pk−2
m
The double-factorial notation means that for n even/odd, n!! is the product of all even/odd integers less than or
equal to n: 0
n(n − 2)(n − 4) . . . 2 n even
n!! =
n(n − 2)(n − 4) . . . 1 n odd
with 0!! = 1 and 1!! = 1. The first few associated Legendre functions are thus:
P22 = 3(1 − ν 2 )
P11 = −(1 − ν ) 2 1/2
P00 = 1 P21 = −3(1 − ν 2 )1/2 ν
P10 =ν 1
P20 = (3ν 2 − 1)
2
The Legendre polynomial Pn (ν) contains even/odd powers of ν for n even/odd; (1, ν 2 , . . . , ν n ) terms for n even,
(ν, ν 3 , . . . , ν n ) terms for n odd. So dm Pn /dν m has (1, . . . , ν n−m ) terms for n and m either both odd or both
even; and (ν, . . . , ν n−m ) terms for only n or m odd (n + m odd). The factor (1 − ν 2 )m/2 = rm on the rotor disk,
√
where ν = ± 1 − r2 .
The dynamic wake theory uses normalized associated Legendre functions:

m
P n (ν) = (−1)m Pnm (ν)/ρm
n
m
Qn (iη) = Qm m
n (iη)/Qn (i0)
$ 1
2 1 (n + m)!
(ρm
n) = [Pnm (ν)]2 dν =
0 2n + 1 (n − m)!
31 m
With this definition, 0
[P n (ν)]2 dν = 1, and
$ 1
m m
P n (ν)P k (ν) dν = δnk
0
for m + n odd. The first few normalized associated Legendre functions are:
4
2 15
4 P2 = (1 − ν 2 )
8
1 3 4
0 P1 = (1 − ν 2 )1/2 1 15
P0 =1 2 P2 = (1 − ν 2 )1/2 ν
0 √ 2
P1 = 3ν 4
0 5
P2 = (3ν 2 − 1)
4
k
He (reference 1) also gives an explicit expression for the function Ψkj (r) = ν1 P j (ν) on the rotor disk:
1 k j−1
(−1)(q−k)/2 (j + q)!!
Ψkj (r) = P j (ν) = (2j + 1)Hjk rq
ν (q − k)!!(q + k)!!(j − q − 1)!!
q=k,k+2,...
where
(j + k − 1)!!(j − k − 1)!!
Hjk =
(j + k)!!(j − k)!!
√ k
and on the rotor disk η = 0 and ν = 1 − r2 . So Ψkj = ν1 P j is a polynomial in the radial coordinate r: it has rk
to rj−1 terms, for j + k odd and j ≥ k.
24-4.2.3 Potential Solution and Equations of Motion
For a circular wing or disk, Laplace’s equation is written in ellipsoidal coordinates, and the method of
separation of variables is used to obtain a series solution for the acceleration potential, and hence for the rotor
loading:
∞ ∞
Φ= Pnm (ν)Qm m m
n (iη) (cn (t) cos mθ + dn (t) sin mθ)
m=0 n=m+1,m+3,...
∞ ∞
Δp = −2 Pnm (ν)Qm m m
n (i0) (cn (t) cos mθ + dn (t) sin mθ)
m=0 n=m+1,m+3,...
√
On the rotor disk η = 0 and ν = ± 1 − r2 (ν > 0 above the disk and ν < 0 below the disk). The pressure
difference on the disk is thus
Δp = Φ(ν < 0, η = 0) − Φ(ν > 0, η = 0) ∼ Pnm (ν < 0) − Pnm (ν > 0)
It follows that Pnm (ν) with n + m even does not produce a pressure continuity on the disk, since it is an even
function of ν. So these series only contain terms for n + m odd, and Δp = −2Φ(ν > 0, η = 0, θ). Also, Pnm
is only defined for n ≥ m. The functions P (iν) and Q(η) are not used since they do not satisfy the boundary
conditions.
Now normalized associated Legendre functions are introduced, and the coefficients of the series scaled to
give the loading:
∞ ∞
1 m m
Φ=− P n (ν)Qn (iη) (τnmc (t) cos mθ + τnms (t) sin mθ)
2 m=0
n=m+1,m+3,...
∞ ∞
m
Δp = P n (ν) (τnmc (t) cos mθ + τnms (t) sin mθ)
m=0 n=m+1,m+3,...
This is the solution for the pressure field Φ caused by the loading Δp on a circular disk. Separate expansions are
defined for the convection and unsteady terms: ΦV from τ V and ΦA from τ A .
To develop the equations of motion for the dynamic wake, consider the convection and unsteady terms of the
potential. The convection term is integrated upstream to give:
$ ∞
1 ∂ΦV
qi = − dξ
V∞ ξ ∂xi
∂qi ∂ΦA
=−
∂t ∂xi
The downwash w is the normal component of the velocity at the rotor disk, positive downward. So w is the −z
component of the velocity at ξ = 0:
$ ∞
1 ∂ΦV
w= dξ = F [ΦV ]
V∞ 0 ∂z

∂ΦA
ẇ = = E[ΦA ]
∂z η=0
Symbolically then, the inverse operators give the equations of motion of the dynamic wake theory:
E −1 [ẇ] + F −1 [w] = ΦA + ΦV = Φ
The acceleration potential Φ has been discretized in terms of the loading coefficients τ . The downwash velocity
w is discretized in terms of the inflow states α. The downwash is expressed as the series
∞ ∞

w= Ψkj (r) αjkc (t) cos kθ + αjks (t) sin kθ
k=0 j=k+1,k+3,...
k √
The radial functions Ψkj must be linearly independent and complete. The choice Ψkj = P j (ν = 1 − r2 ) gives
k
w = 0 at the blade tips (r = 1, ν = 0), so does not produce good convergence (Ref. 1). Therefore Ψkj = ν1 P j (ν)
is used, which is nonzero at the blade tips.
24-4.2.4 Unsteady Term and Mass Matrix
The mass matrix M (from the operator E) is obtained by substituting the series for ΦA (generalized forces
τ ) and w (inflow states α) into
% &
∂ΦA 1 2 ∂ 2 ∂ A
1 ∂ΦA
ẇ = = 2 η(1 − ν ) + ν(1 + η ) Φ =
∂z η=0 ν + η2 ∂ν ∂η η=0 ν ∂η η=0
Using the orthogonality of first cos mθ and sin mθ, and then of P (ν) gives

1 ∂Qn (iη)
m
π
α̇n = −
mc
τnmcA = τ mcA
2 ∂η 4Hnm n
η=0

m
1 ∂Qn (iη) π
α̇nms = − τnmsA = τ msA
2 ∂η 4Hnm n
η=0
So M c α̇c = τ cA and M s α̇s = τ sA , where

4Hnm
M=
π
is a diagonal matrix.
24-4.2.5 Convection Term and Derivative Matrix
The derivative matrix L (from the operator F ) is obtained by substituting the series for ΦV (generalized forces
τ ) and w (inflow states) into $ ∞
1 ∂ΦV
w= dξ
V∞ 0 ∂z
For the downwash expansion, here it is convenient to use
∞ ∞
k
w= P j (ν) akc ks
j (t) cos kθ + aj (t) sin kθ
k=0 j=k+1,k+3,...
k k
After substituting for the series, multiply by (P j (ν) cos kθ) and (P j (ν) sin kθ); integrate radially and azimuthally
(over ν and θ); and use the orthogonality of cos mθ and sin mθ, and P (ν). Then
kc 1 1 kmc mcV
aj = L τn
V∞ jn
ks 1 1 kms msV
aj = L τn
V∞ jn
where m and k range from 0 to infinity for the cosine terms, and from 1 to infinity for the sine terms; with
1 matrices are:
n = m + 1, m + 3, . . . and j = k + 1, k + 3, . . .. The coefficients of the L
$ 2π $ 1 $ ∞
1 0mc 1 0 ∂Knmc
Ljn = P j (ν) dξ dν dθ
2π 0 0 0 ∂z
$ $ $ ∞
1 kmc 1 2π 1 k ∂Knmc
Ljn = P j (ν) cos kθ dξ dν dθ
π 0 0 0 ∂z
$ $ $ ∞
1 kms 1 2π 1 k ∂Knms
Ljn = P j (ν) sin kθ dξ dν dθ
π 0 0 0 ∂z
where
1 m m
Knmc = − P n (ν)Qn (iη) cos mθ
2
1 m m
Kn = − P n (ν)Qn (iη) sin mθ
ms
2
The integral over the streamwise coordinate ξ becomes
$ ∞ $ ∞
∂ ∂
K(ν, η, θ) dξ = K(x = x0 − ξ sin χ, y = y0 , z = ξ cos χ) dξ
0 ∂z 0 ∂z
$ ∞ $
1 ∂K tan χ ∞ ∂K
= dξ + dz
cos χ 0 ∂ξ cos χ 0 ∂x
$
1 tan χ ∞ ∂
=− K(ξ = 0) + K(x = x0 − z tan χ, y = y0 , z) dz
cos χ cos χ 0 ∂x
1 matrices. However, different radial shape functions have been used for the
He (ref. 1) gives expressions for the L
ẇ and w expansions, so it is necessary to relate the a coefficients to the states α:
%$ 1 &
k k 1 k
aj = P j (ν) P n (ν) dν αnk
0 ν
The result is
1 kmc
αjkc = Ljn τnmcV
V∞
ks 1 kms

αj = Ljn τnmsV
V∞
= 2L matrices.
He (ref. 1) derives the following result for the L
0mc = X m Γ0m
L jn jn

kmc
L jn = X |m−k|
+ (−1) X |m+k|
Γkm
jn

kms
L jn = X
|m−k|
− (−1) X |m+k| Γkmjn
where = min(k, m); X = tan |χ/2|; and

⎧
⎪
⎪ (−1)(n+j−2k)/2 2 (2n + 1)(2j + 1)
⎪
⎪ k + m even
⎪
⎪ (j + n)(j + n + 2)[(j − n)2 − 1]
⎪
⎪ Hnm Hjk
⎨
Γkm
jn = π sign(k − m)
⎪
⎪ k + m odd, j = n ± 1
⎪
⎪
⎪
⎪ 2 Hnm Hjk (2n + 1)(2j + 1)
⎪
⎪
⎩
0 k + m odd, j = n ± 1
24-4.2.6 Ordinary Differential Equations
The ordinary differential equations relating the downwash states α and the loading τ are obtained by substi-
tuting M α̇ = τ A and α = V1∞ Lτ V = Lτ V into the total loading τ = τ A + τ V :
1 c c c
L M α̇ + αc = Lc τ c
Ω
1 s s s
L M α̇ + αs = Ls τ s
Ω
The equations for the cosine and sine states are not coupled. The vectors are α = (αjk ) and τ = (τnm ), were k
and m are the azimuth indices, and j and n are the radial indices (n + m and j + k odd, n ≥ m and j ≥ k. In
these arrays, there are subarrays for each azimuth index; the azimuth indices are the outer loop. The cosine arrays
include the zero-th harmonic.
The number of shape functions is specified by the number of azimuthal harmonics Nψ , and the maximum
order of the radial polynomial Nr . The same number of shape functions is used for both the downwash and the
loading. For each harmonic number m, the Legendre function index n = m + 1, m + 3, . . .. Each radial function
Pnm (ν)/ν is a polynomial in r, with rq terms, q = m, m + 2, . . . , n − 1. Since the minimum power is rm , it
follows Nr ≥ Nψ , Nr ≥ 0, Nψ ≥ 0 are required. If NL is the number of Legendre functions used for harmonic
m, then n = m + 1, m + 3, . . . , m + 2NL − 1 and the maximum power is rm+2NL −2 . Thus the indices in the
arrays have the following variation:
a) cosine: m = 0, 1, . . . , Nψ ; for each m, n = m + 1, m + 3, . . . , m + 2NL − 1

b) sine: m = 1, . . . , Nψ ; for each m, n = m + 1, m + 3, . . . , m + 2NL − 1
where 0
(Nr + 2 − m)/2 Nr and m both odd or both even
NL =
(Nr + 1 − m)/2 otherwise
The total number of states is the sum of NL for each value of m.
24-4.2.7 Extensions
The total velocity normal to the rotor disk is λ = λi + μz . The mean wake-induced inflow is obtained from
√
the first inflow variable: λi = 3α10 . The wake skew angle χ = tan−1 |μ/λ| is evaluated using the total normal
velocity λ, not just μz . Consequently all terms of the derivative matrix L depend on the first inflow variable α10 .
To account for the influence of the induced flow on the total steady flow through the disk, 1/V∞ in L is
replaced by 1/V = 2(∂CT /∂λ), or
μ2 + (λ + λi )λ
V =
μ2 + λ2
for all elements of L except α10 /τ10 .

√
The first inflow variable produces the mean downwash, λi = 3α10 . The first generalized force is obtained
√
from the thrust, τ10 = 23 CT . The derivative matrix term is L00 0 0
11 = 3/8V∞ . The equation α1 = (3/8V∞ )τ1 is
equivalent to
9 CT
λi =
8 2V∞
The momentum theory result for the mean wake-induced velocity is
CT
λi =
2 λ2 + μ2
So to obtain the conventional momentum theory result for the mean downwash, V∞ in L00
11 is replaced by

V = λ2 + μ2
The additional factor of 9/8 occurs because the loading generated by τ10 is not uniform over the disk, but instead
goes to zero at the disk edge. The equation for α10 is nonlinear because of the presence of λ in V . For trim, the
algorithm for λi (CT ) gives
1 9 2
α10 = √ λi CT = √ τ10
3 8 3
This algorithm is described for the rotor inflow component, including corrections for nonideal induced losses in
hover and forward flight (the empirical factor κ), and the singularity of momentum theory at ideal autorotation.
Pitch or roll rate of the rotor disk distorts the rotor wake, and therefore may produce a linear inflow variation
over the disk (ref. 3). This effect is included in the dynamic wake model following reference 4. Wake distortion
terms αwd are added to the equations of motion:
1 c c c
L M α̇ + αc = Lc τ c + αwd
c
Ω
1 s s s
L M α̇ + αs = Ls τ s + αwd
s
Ω
where the only two non-zero terms are

1c fq 1.5ωy
α2wd =
Ψ12 (ν = 0)
fp 1.5ωx
1s
α2wd =− 1
Ψ2 (ν = 0)

and Ψ12 (ν = 0) = Ψ12 (r = 1) = 15/2. Included in αwd are empirical factors fp and fq , which are defined in
terms of values for hover and for forward flight: f = (ff μ2 + fh λ2 )/(μ2 + λ2 ).
24-4.2.8 Limitations
The limitations of the generalized dynamic wake theory include the following. The aerodynamic theory
is incompressible. Only the loads normal to the rotor disk are considered. The rotor loading is represented
by a distribution of acceleration potential (pressure) doublets on a thin, undeformed disk. From the use of the
acceleration potential, it follows that the wake geometry consists of a skewed cylinder, without wake rollup or
distortion. The model has no rolled-up tip vortices, and so no blade-vortex interaction.
There are also several weaknesses in the fluid dynamic theory. That both ΦA and ΦV satisfy Laplace’s
equation (instead of just the sum) is an assumption. The acceleration potential Φ is defined for all space and time
by the loading τ , hence the velocity q is defined for all space and time (not just the downwash w from ordinary
differential equations). The flow is incompressible, so time lag effects can only come from the vortex wake.
Two different series are used for the downwash w. The loading representation does not include a function that
corresponds to a uniform disk loading.
24-4.3 Rotor Loading and Generalized Forces
The generalized forces τ must be evaluated from the pressure difference Δp acting on the rotor disk. The
loading has been written as the following series
m
Δp = P n (ν) (τnmc (t) cos mθ + τnms (t) sin mθ)
m=0 n=m+1,m+3,...
√ m 0 √
where ν = 1 − r2 on the rotor disk. P n is the product of ν and a polynomial in r. In particular, P 1 = 3ν and
1
P 2 = 15/2νr. Note that this loading is always zero at the edge of the rotor disk. There is no loading function
that corresponds to a uniform disk loading.
k k
To obtain the loading coefficients, multiply by (P j (ν) cos kθ) and(P j (ν) sin kθ); integrate radially and
31
azimuthally (over ν and θ); and use the orthogonality of cos mθ and sin mθ, and P (ν). Note that 0 (. . .)dν =
31
0
(. . .)(r/ν)dr. Thus
$ 2π $ 1
0c 1
τn = Δp rΨ0n (ν) dr dθ
2π 0 0
$ $
mc 1 2π 1
τn = Δp rΨmn (ν) cos mθ dr dθ
π 0 0
$ $
ms 1 2π 1
τn = Δp rΨmn (ν) sin mθ dr dθ
π 0 0
1 m
where Ψm n = ν P n (ν). The integral over r and θ is an integral over the entire rotor disk, at time t. These
generalized forces are defined relative a disk x-axis that is aligned with the rotor inplane velocity direction. The
inplane velocity component has an azimuth angle θ0 = − tan−1 (μy /μx ) in the disk plane. So θ must be replaced
by (θ − θ0 ) in the above expressions.
The rotor loading Δp is generated by the pressures acting at time t on a finite number of blades. The
aerodynamic lift force on blade q is Fq (r, t). The collocation point location on the rotor disk is (x, y), from which

r = x2 + y 2 and the blade nominal azimuth θq = tan−1 (y/x) are obtained. Then

Δp = Fq (r) η = r(θ − θq )
blades
where a chordwise loading distribution (η) has been introduced; η is the chordwise coordinate, such that θ =
3
θ0 + tan−1 (η/r) ∼
= θ0 + η/r. By definition (η)dη = 1. Then
$ 1
1
τn0c = Fq 0
m Ψn (ν) dr
2π 0
blades
$ 1
1
τnmc = Fq m
m Ψn (ν) cos m(θq − θ0 ) dr
π 0
blades
$ 1
1
τnms = Fq m
m Ψn (ν) sin m(θq − θ0 ) dr
π 0
blades
where m is an integral of (η). The discretized aerodynamic lift force on blade q at collocation point p is Fqp (t),
so
1
τn0c = Fqp Ψ0n (ν)
2π
blades coll pt
1
τnmc = n (ν) cos m(θq − θ0 )
Fqp Ψm
π
blades coll pt
1
τnms = n (ν) sin m(θq − θ0 )
Fqp Ψm
π
blades coll pt
A concentrated chordwise loading gives = δ(η), and m = 1 for all harmonics. So m = 1 has been used here.
The radial function Ψm
nis evaluated at the nominal span station of the collocation point, rAp .
√
The first three generalized forces are evaluated using Ψ01 = 3 and Ψ12 = 15/2, hence are related to the
aerodynamic thrust, pitch moment, and roll moment acting on the rotor disk:
√
0c 3
τ1 = CT
42
15
τ21c = (−CM y cos θ0 + CM x sin θ0 )fm
2
4
15
τ21s = ( CM x cos θ0 + CM y sin θ0 )fm
2
including the empirical factor fm , which is defined in terms of values for hover and for forward flight. Optionally
these three generalized forces can be calculated directly from the input CT , CM y , and CM x , instead of from the
discretized forces on the blades.
24-4.4 Wake-Induced Velocity
The downwash w is evaluated over the rotor disk from the inflow states α. In the trim task, the static solution
for the inflow states is obtained (no degrees of freedom). In the transient and flutter tasks, the ordinary differential

equations are solved. The collocation point location is (x, y), from which r = x2 + y 2 and θ = tan−1 (y/x)
are obtained. The inflow states are defined relative a disk x-axis that is aligned with the rotor inplane velocity
direction. The inplane velocity component has an azimuth angle θ0 = − tan−1 (μy /μx ) in the disk plane. So θ
must be replaced by (θ − θ0 ) in the downwash expressions. The downwash is evaluated from the following series:

w= Ψkj (r) αjkc (t) cos k(θ − θ0 ) + αjks (t) sin k(θ − θ0 )
k=0 j=k+1,k+3,...
k √
where Ψkj = ν1 P j (ν), and ν = 1 − r2 on the rotor disk. The function Ψm
n is a polynomial in r. In particular,
0
√ 1
m
Ψ1 = 3 and Ψ2 = 15/2r. The radial function Ψn is evaluated at the nominal span station of the collocation
point, rAp .
Dimensional velocities are obtained by multiplying the downwash w by the reference velocity ΩR (the tip
speed). It is assumed that the velocity is in the direction of the z-axis of the wing axes being used (S or T axes).
Thus for collocation points on the wing set, the wake-induced velocity is:
I
vA = w ΩR (−I )
I
vA = Kλi ΩR (−I )
√
where λi = 3α10 is the mean downwash. Each collocation point off the wing set has an empirical factor K,
which is the ratio of the velocity at the collocation point to the mean inflow at the wing set. Thus K accounts
for effects such as the fraction of the aerodynamic component at the collocation point that is affected by the wake
(less than 1, zero for no interference); and the fraction of fully developed wake velocity that is achieved at the
collocation point (maximum about 2). The interference factor K is defined in terms of values for hover and for
forward flight, to reflect the extremes of the wake configuration relative the collocation point. Then K is obtained
using K = (Kf μ2 + Kh λ2 )/(μ2 + λ2 ).
24-4.5 Dynamic Inflow
The three-term dynamic wake theory corresponds to the commonly used dynamic inflow model. The down-
wash series consists of terms that are uniform and linear over the rotor disk:
w = Ψ01 α10 + Ψ12 (α21c cos θ + α21s sin θ)

4
√ 15
= 3α10 + r(α21c cos θ + α21s sin θ)
2
= wu + wx r cos θ + wy r sin θ
√ √
using Ψ01 = 3 and Ψ12 = 15/2r. The generalized forces τ10c = ( 3/2)CT , τ21c = − 15/2CM y , and

τ21s = 15/2CM x produce the rotor loading
0 1
Δp = P 1 τ10 + P 2 (τ21c cos θ + τ21s sin θ)
4 !
√ 0 15
= 3τ1 + 1c 1s
r(τ2 cos θ + τ2 sin θ) 1 − r2
2

3 15
= CT + r(−CM y cos θ + CM x sin θ) 1 − r2
2 2
0 √ √ 1 √
using P 1 = 3 1 − r2 and P 2 = 15/2r 1 − r2 . The equations of motion are
LM α̇ + α = Lτ + αwd

1c
where α2wd = 1.5ωy 1s
2/15 and α2wd = −1.5ωx 2/15. The matrices are:
⎡4 ⎤
0 0
⎢π ⎥
⎢ 8 ⎥
M =⎢
⎢0 0 ⎥
⎥
⎣ 3π ⎦
8
0 0
3π
⎡ ⎤
3 π
−X √ 0
⎢ 4 2 10 ⎥
1 1 ⎢
⎢ π 5
⎥
⎥
L= L= ⎢X√ (1 − X 2 ) 0 ⎥
2V 2V ⎢ 10 8 ⎥
⎣ 5⎦
0 0 (1 + X 2 )
8
where X = tan(χ/2), so
μ
X=
( μ + λ2 + |λ|)
2

2 μ2 + λ2
1 + X2 =
( μ2 + λ2 + |λ|)
2|λ|
1 − X2 =
( μ + λ2 + |λ|)
2
In terms of the dynamic inflow states and loading, the equations are:
⎡ 8 ⎤
0 0 ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
⎢ 3π ⎥ ẇu wu CT 0
⎢ 16 ⎥
L∗ ⎢
⎢ 0 0 ⎥ ⎝ ⎠ ⎝
⎥ ẇx + wx = L
⎠ ∗⎝
−CM y ⎠ + ⎝ 1.5ωy ⎠
⎣ 45π ⎦ ẇy w CM x −1.5ωx
16 y
0 0
45π
with ⎡ ⎤
9 3π μ
− 0
⎢ 16 8 μ + λ2 + |λ|
2 ⎥
⎢ ⎥
1 ⎢
⎢ 3π μ 75 |λ| ⎥
⎥
L = ⎢ 8 2
∗ 0 ⎥
V ⎢ μ + λ2 + |λ| 16 μ2 + λ2 + |λ|
⎥
⎢ ⎥
⎣ 75 μ2 + λ2 ⎦
0 0
16 μ2 + λ2 + |λ|
For comparison, the dynamic inflow model of Pitt and Peters (ref. 5) gives the same mass matrix, and
⎡ ⎤
1 15π μ
− 0
⎢ 2 64 μ2 + λ2 + |λ| ⎥
⎢ ⎥
⎢
1 ⎢ 15π μ |λ| ⎥
∗ 4 0 ⎥
L = ⎢ 64 2 + λ2 + |λ| 2 + λ2 + |λ| ⎥
V ⎢ μ μ ⎥
⎢ ⎥
⎣ 2
μ +λ 2 ⎦
0 0 4
μ2 + λ2 + |λ|
for the derivative matrix. The larger numerical constants in the dynamic wake L matrix are a consequence of the
√
additional factor of 1 − r2 in the loading representation.
24–5 References
1) He, C.J. “Development and Application of a Generalized Dynamic Wake Theory for Lifting Rotors.” Georgia
Institute of Technology, Ph.D. Thesis, July 1989.
2) Peters, D.A., and He, C.J. “Finite State Induced Flow Models — Part II: Three-Dimensional Rotor Disk.”
Journal of Aircraft, Volume 32, Number 2, March-April 1995.
3) Keller, J.D. “An Investigation of Helicopter Dynamic Coupling Using an Analytical Model.” Journal of the
American Helicopter Society, Volume 41, Number 4, October 1996.
4) Hamers, M., and Basset, P.-M. “Application of the Finite State Unsteady Wake Model in Helicopter Flight
Dynamic Simulation.” European Rotorcraft Forum, the Netherlands, September 2000.
5) Gaonkar, G.H., and Peters, D.A. “Review of Dynamic Inflow Modelling for Rotorcraft Flight Dynamics.”
AIAA Paper Number 86-0845, May 1986.
ξ force velocity normal wing set aero

ang vel position force
tip plane
motion 1 yes yes no no yes (c)
interference 0 yes (a) yes (a) yes yes (b) yes (a) (c)
velocity
sensor no yes
Notes:
a) no for transient and flutter (except velocity)
b) yes only if collocation points of output and input are same
c) no if one use force
Figure 24-3 Functionality of rotor dynamic wake component.

Chapter 25
WING WAKE COMPONENT
25–1 Description
A wing wake component calculates the wake induced velocity of a wing set. It is called a “nonuniform inflow”
component. There may be a corresponding uniform inflow component; which component gives a nonzero velocity
result then depends on the specified model, and perhaps on the loop solution level. There is one wake per wing.
There is a separate set of influence coefficients for the contribution of each wing to the velocity at a collocation
point. The interference velocity calculated by the component is the sum of the velocities from all wings in the set.
The induced velocity is calculated from a vortex wake model, with discretized vorticity elements. The velocity
depends on the wing circulation at past times. The relation is described by an integral equation, which requires
an implicit solution. The induced velocity is calculated at collocation points on this wing set, and the interference
velocity at other points can be calculated.
The wake model is usually based on a vortex lattice (straight-line segments) approximation for the wake. A
small viscous core radius is used for tip vortices. Vortex sheet elements can be used to represent the inboard wake,
but it is usually sufficient (and more efficient) to approximate sheets by line segments, with a large core radius to
eliminate large velocities. Nonplanar, quadrilateral sheet elements are available if needed. The wake influence
coefficients are calculated for incompressible flow. The induced velocity is calculated at an arbitrary collocation
point: on this wing, on other wings of this wing set, on other aerodynamic components, or in the flow field. The
circulation can be assumed periodic. The influence of the ground can be included in the wake-induced velocity
calculation, through the use of image elements in the wake model. Wake geometry components can also include
the ground plane influence. For communication with an external aeroacoustic analysis, the induced velocity can
be calculated excluding vortex elements inside the computational domain of the external analysis.
The wake rollup process is modelled. The tip vortices can roll up at both wing tips or only at one tip.
Eventually the tip vortex has the strength of the wing peak bound circulation at the time that the wake elements
were created. The possibility of two bound circulation peaks (left and right peaks of opposite sign) is included
in the rollup model. A number of parameters define entrainment and stretching in the tip vortex rollup process.
The far wake trailed vorticity can be divided into several spanwise panels, if necessary to provide more detailed
structure of the inboard vorticity. The spanwise location of the tip vortex at the generating wing is also prescribed
(implemented in the wing and wake geometry components). Wing-vortex interaction loading is calculated using
second-order lifting-line theory (three-quarter chord collocation point), or just an artificially large vortex core size.
The vortex core radius can be constant, or grow with wake age, and can include a term that scales with the trailed
vorticity moment. A large core radius can be used for the velocity induced on the inboard part of the wing, in order
to suppress wing-vortex interaction there (as observed in experiments on rotors). For axial flow, an extended far
308 WING WAKE COMPONENT
wake model can be used. For lifting-line theory, the collocation points on this wing set can be at the quarter chord
or three-quarter chord. The lifting line can be at the quarter chord position of the wing, or can be approximated
by a straight line.
inflow, and trim. The dynamic inflow model calculates perturbations from trim (which must be a uniform or
nonuniform model). For the nonuniform inflow model, it is possible to still use uniform inflow for collocation
points off the wing set (a more general classification of collocation points can be handled by using separate
components). The models available depend on the solution task:

b) Transient: zero, uniform inflow, nonuniform inflow, dynamic inflow, trim.
c) Flutter: zero, uniform inflow, dynamic inflow, trim.
For the trim task, the model can be specified for each level of a designated wake loop. For the “trim” option in
the transient task, the wake component returns zero velocity, so the “trim” option must be implemented by using
a part solution method = trim.
Figure 1 illustrates the functionality of the wing wake component.
Component Input
a) Aerodynamic interfaces at collocation points: location relative inertial frame, rPI .
b) Influence coefficients (from this component).
I
c) Wake geometry: rW for each wing.
d) Bound circulation: Γ at all span stations, for each wing.
far wake panels; and centroid data (Γ, rC , rG , f , δ, set) for far wake panel edges; for each wing.
f) Wake vorticity strength (from this component).
g) Wing set plane normal, k I .
I
h) Box axis positions, rbox .
Component Output
I
b) Wake vorticity strength: G for each wing.
c) Influence coefficients: C I , for each collocation point and each wing.
d) Sensor: span station of circulation peaks (rmax , rL , rR ) for entire wing and multiple far wake panels, for each
wing (nonuniform inflow parameters).
25-3.1 Vortex Wake Analysis
25-3.1.1 Wake Structure and Rollup
A three-dimensional wing trails the bound circulation Γ into a wake. Figure 2 illustrates the wake of a rotor
WING WAKE COMPONENT 309

location rel inertial frame interference velocity
wake geometry
bound circulation influence coefficients
bound circulation peaks wake vorticity strength
wing plane normal sensors

box axis positions
Figure 25-1 Functionality of wing wake component.

bound
circulation
de
rotor bla
tip vortex
inboard
vortex
sheet
root
vortex
Figure 25-2 Trailed and shed vorticity in rotor wake.

blade. The spanwise variation of Γ produces trailed vorticity, parallel to the local free stream direction at the time
it leaves the wing. Time variation of Γ produces shed vorticity, oriented spanwise. The strength of the trailed
and shed vorticity is determined by the span and time derivatives of Γ at the time that the wake element was
created. The bound circulation typically drops quickly to zero at the wing tips. The trailed sheet therefore has a
high strength (proportional to the spanwise derivative of Γ) at the wake outer edges, and quickly rolls up to form
concentrated tip vortices. The rollup process, which is also influenced by the tip geometry, produces a line vortex
with a small core radius, hence large velocities.
The vorticity in the tip vortex is distributed over a small but finite region, called the vortex core, because of
viscosity of the fluid. The core radius is defined at the maximum tangential or circumferential velocity. Figure
3 shows the influence of the core radius on peak velocity. The vortex core is an important factor in the induced
velocity character since it limits the maximum velocities near the tip vortices. There is only limited experimental
data on the core size. The strong, concentrated tip vortices are the dominant feature of the wake. The tip vortices
produce a highly nonuniform flow field. Thus any aerodynamic component encountering this flow field will see
large vortex-induced loadings. Between the tip vortices there is an inboard sheet of trailed vorticity (which may
roll up to form additional vortices, usually much weaker than the tip vortices). With unsteady loading, this inboard
sheet has shed vorticity, from the time derivative of Γ. For a rotor, the bound circulation peak occurs near the blade
tip, and the gradient of bound circulation is low at the blade root; so the blade tip vortex is strong and concentrated,
while the root vortex is weaker and more diffuse.
was created. This geometry is distorted by the self-induced velocity of all the wakes in the system, as well as the
influence of wings and bodies on the air velocity. In axes moving with the wing, the wake geometry also reflects
the past history of the wing position.
A calculation of the wake rollup from first principles is needed, but it is a difficult problem, beginning to be
attacked using the methods of computational fluid dynamics. The vortex core is largely formed at the wing trailing
edge, so the problem is not the inviscid rollup of a vortex sheet. Moreover, discretization of the wake for a rollup
calculation is not easy; indeed it may not be well-posed even for an inviscid, two-dimensional problem. Hence
the rollup problem involves three-dimensional, unsteady, viscous fluid dynamics. Here the wake rollup process
is modelled not calculated, meaning that the structure and properties of the rolled-up wake are determined from
assumptions and input parameters, and from the spanwise distribution of the bound circulation where the wake was
created. It is important that the model account for the influence of the strength of the tip vortex when it encounters
a following wing; the influence of the core radius when the vortex is fully rolled up; and the effect of the vortex
rollup on the tip load of the generating wing.
The wake model must encompass the rollup implied by typical circulation distributions on wings. Generally
it is assumed that the lift is concentrated at the tip of the wing. So the trailed vorticity strength is high at the edge
of the wake, and the vortex sheet quickly rolls up into a concentrated tip vortex. The formation of the vortex is
influenced by the tip geometry, with the core largely formed by the time the vortex leaves the trailing edge. The
rolled-up tip vortex quickly attains a strength nearly equal to the maximum bound circulation. The tip vortex
has a small core radius, depending on the wing geometry and loading. A tip vortex may form at both tips (for a
nonrotating wing), or at just one tip (for a rotor blade). This is the assumed picture of the rollup process upon
which the wake model is based.
potential vortex
distributed vorticity, core radius rc = r0
1.5
distributed vorticity, core radius rc = 3r0 / 4
distributed vorticity, core radius rc = r0 / 2
distributed vorticity, core radius rc = r0 / 3
circumferential velocity, v/(Γ/2 π r0)
1.0
0.5
0.0
0.0 1.0 2.0 3.0
distance from vortex center, r/r0
Figure 25-3 Tip vortex core radius and peak velocity.

In relating the rolled-up wake structure to the bound circulation distribution, it is necessary to consider single-
peak and dual-peak cases. Often the bound circulation has the same sign along the entire wing span; this is the
single-peak case. There may be local minima and maxima in Γ, as from wing-vortex interactions, but generally
there is a monotonic increase and decrease in Γ. The rolled-up wake is constructed solely from the magnitude and
position of the peak bound circulation.
The wing loading at a given time need not be all positive or all negative. For example, it is common for
a helicopter rotor in high speed forward flight to encounter negative lift on the advancing tip, particularly in the
second quadrant. With a flapping rotor, the net pitch and roll moment on the hub must be small (zero if there is no
flap hinge offset). In forward flight, the lift capability on the retreating side is limited by the combination of low
dynamic pressure and stall of the airfoil. Consequently the lift on the advancing side must also be small, in order
to maintain roll balance. At sufficiently high speed, the lift on the advancing tip can become negative. Large twist
of the blade (built-in or elastic) increases the negative loading. The vortex-induced loading on following blades
shows that this negative loading produces substantial negative trailed vorticity in the wake. Over much of the disk,
Γ is still positive all along the span. However, there is a range of azimuth on the advancing side with a negative
peak near the tip and a positive peak inboard. Figure 4 illustrates the wake structure.
Thus the bound circulation can be positive over part of the wing span, and negative over the remainder; this
is the dual-peak case. Here one tip vortex forms with negative strength (opposite sign as normal rollup), with a
sheet of positive trailed vorticity created between the two peaks, which perhaps also rolls up to some extent. Again
there may be local minima and maxima in Γ, which are assumed to have no effect on the rollup. For the dual-peak
case, the loading is assumed to have one region of positive circulation and one region of negative circulation. The
rolled-up wake is constructed from the magnitude and position of the left and right bound circulation peaks.
The wake structure (single or dual-peak) assumed when the influence coefficients are calculated may not
correspond to the wing loading distribution, either because the solution has not converged or because the wrong
wake model was specified. If the influence coefficients are calculated for a dual-peak model and the loading is
actually single-peak, then the analysis can find a bound circulation value for the other peak that gives the same
induced velocity as the single-peak model. If the influence coefficients are calculated for a single-peak model and
the loading is actually dual-peak, then some approximation will result. With both tip vortices rolled up, one has
the wrong sign and magnitude. With only the right tip vortex rolled up, the single-peak model gives the correct
tip vortex strength if it uses the value of the right peak (if it uses the peak of maximum magnitude, the tip vortex
strength can be wrong again). The inboard wake sheet has the wrong structure and strength in both cases. It is
essential to get the tip vortex strength correct. Sometimes the inboard wake structure (including possible secondary
rollup) is important as well.
There may be features of the inboard wake structure that must be modelled. For example, a trailing-edge
flap or a rapid change in the wing chord can produce a step change in the bound circulation, which might produce
significant rollup of the inboard trailed vorticity (figure 5). The wake model is constructed by specifying the
span stations of such inboard rollup, based on the geometry of the wing. Hence the far wake trailed vorticity is
divided into several spanwise panels, with the single-peak or dual-peak model applied to each panel. The wake
structure is often more complicated than considered by this component. Such structure has too much detail to be
accommodated by a modelling approach; a calculation of the rollup is certainly needed.
The effects of the rollup process on the tip loading of the generating wing should be considered. Most such
bound circulation
(ψ = 180; age = 0)
g
win
tip vortex
(age = 90)
trailed and
shed vorticity
(age = 180)
Figure 25-4 Rotor wake rollup structure.

bound
circulation
de
rotor bla
tip vortex
root
vortex
rolled up trailed vorticity
Figure 25-5 Rotor wake with multiple rolled-up trailed vorticity.

effects are beyond lifting-line theory, but at least the spanwise location of the rollup should be modelled in extreme
cases. The core is largely formed by the time the tip vortex reaches the trailing edge, even for a tapered tip. The
center of the core is found somewhat inboard of the physical tip, typically 1% span inboard for a rectangular
planform and more for a highly tapered tip. In lifting-line theory, this implies that there is no bound circulation
of the wing tip outboard of the rolled-up location, with a significant effect on the spanwise loading distribution if
the tip vortex forms at least 3–5% span inboard. This effect can be modelled in the wing component by forcing
the bound circulation to be trailed into the tip vortex at the proper (prescribed) span station. In addition, the wake
geometry component must account for the smaller effective span of the wing when constructing the undistorted
geometry of the far wake.
25-3.1.2 Wing-Vortex Interaction
The airloads produced by wing-vortex interaction depend on numerous physical effects, including the extent
of the tip vortex rollup; the tip vortex strength; the size of the viscous core; the distorted wake geometry; lifting-
surface effects on the induced loading; and possibly even vortex bursting, vortex-induced stall on the wing, or
wing-induced geometry changes.
The viscous core of a line vortex plays a fundamental role in determining the velocity near the vortex. Consider
the tangential or circumferential velocity v about a line vortex, at a distance r from the line. The core radius rc is
defined as the distance r at which the maximum value of v is encountered. For a potential line vortex (no core),
v = Γ/2πr, where the strength Γ is some fraction of the maximum bound circulation (as determined by the rollup
process). For small r, viscosity reduces the magnitude of v, by spreading the vorticity over a nonzero domain
instead of a line. A Rankine vortex core produces solid body rotation of the fluid inside rc , by having a uniform
vorticity distribution concentrated entirely within the core radius. The maximum tangential velocity with a Rankine
core is vmax = Γ/2πrc . For a distributed vorticity core, one simple possibility has the circulation proportional to
r2 /(r2 + rc2 ), so half the vorticity is outside the core radius (ref. 1). With this core the maximum tangential velocity
is vmax = Γ/4πrc . The Rankine, Scully, and several other models of the vorticity distribution are implemented
in the induced velocity calculation. Figure 6 shows the influence of core type on the velocity. Measurements of
the velocity distributions about tip vortices show that the maximum tangential velocity is much less than Γ/2πrc ,
indicating that a substantial fraction of the vorticity is outside the core radius. Hence the distributed core model is
more realistic.
The vortex-induced loads depend on the normal velocity at the wing surface, produced by a vortex at a
distance h above or below the wing. For a potential vortex (no core) the maximum normal velocity (peak velocity)
is Γ/4πh. For h = 0 (the extreme case), the peak velocity equals the maximum tangential velocity, which depends
on the core type and rc as described above. For nonzero separation distance, the Rankine core gives a peak velocity
√
equal to the potential value when h is greater than rc / 2, while the distributed core gives a peak velocity that is
always less than the potential value. In general, the peak velocity is proportional to Γ/rc and a function of h/rc ,
so the core size is a critical parameter of the model, controlling the peak loads. For h = 0 the peak velocity of
a distributed core is 50% that of the Rankine core; for h = rc , the ratio is 70.7%. Hence using the proper core
model is important for accurate prediction of the airloads.
Based on correlation with measured wing-vortex interaction loads, the present analysis requires a core radius
rc of about 20–25% chord, using the distributed core model and second order lifting-line theory (three-quarter
chord collocation point). Limited data suggests that the actual viscous core has a radius of 10–20% chord (full
1.5
potential vortex
Rankine core
Scully core
Bagai-Leishman core
power law core (n = .75)
Oseen core
circumferential velocity, v/(Γ/2 π rc)
1.0
0.5
0.0
0.0 1.0 2.0 3.0
distance from vortex center, r/rc
Figure 25-6 Tip vortex core types.

scale).
In close wing-vortex encounters, the induced loading varies rapidly along the span. First-order lifting-line
theory overpredicts such loading, especially if the span and age resolution in the wake are not small enough. Second-
order lifting-line theory or lifting-surface theory is needed for accurate prediction of wing-vortex interaction loads
(as well as the airloading for swept tips, yawed flow, and low aspect-ratio wings). It is found that lifting-surface
effects reduce the peak induced loading by 20-40% for a vortex-wing separation equal to 25% of the chord (for
various interaction angles). This effect can be approximated by increasing the viscous core size by about 15%
chord. Thus with first order lifting-line theory (quarter chord collocation point), the core radius rc should be
35–40% chord.
The peak-to-peak value of the vortex-induced loading dominates the measured and calculated airloads. Con-
sider what factors influence the peak-to-peak loading. There are physical factors:
a) The rollup process, at the generating wing and in the wake. This process produces the
strength and core size of the tip vortex at the encounter with a following wing. The strength
is less than or equal to the peak bound circulation, and the core size is typically 10–20% of
the chord. If it is assumed that the strength equals the peak bound circulation when in fact the
strength is less (for example, with incomplete rollup by the time the vortex reaches a following
wing), then the analysis overpredicts the loading.
b) Lifting-surface effects, from the low effective aspect-ratio of a close interaction. If such
effects are ignored, as with first-order lifting-line theory, then the analysis overpredicts the
loading.
c) Possibly vortex bursting, vortex-induced stall on the wing, local distortion of the vortex
geometry by the wing, and compressibility and viscous effects in general.
There are computational factors:
a) Wake geometry. If the calculated wing-vortex separation is too large, as when a rigid wake
geometry is used, then the analysis underpredicts the loading.
b) Span and age discretization in the wake. The discretization of the wake in span and age
should be about equal to the wing-vortex separation, which has an effective minimum equal
to the core radius. Typically the span and especially the age resolution used are too large for
close wing-vortex interactions. In such cases the analysis overpredicts the loading.
c) Span and time discretization of the loading. Typically the time resolution of the calculated
airloads is too large. In such cases the analysis underpredicts the loading.
There are measurement factors:
a) Unsteadiness and noise in the data. If in nominally periodic loading the time of occurrence
of the wing-vortex interaction changes from revolution to revolution, then an averaging process
reduces the measured peak loads, and the analysis overpredicts the loading.
The core size rc is a convenient parameter with which to control the amplitude of the calculated wing-vortex
interaction loads, since it determines the maximum tangential velocity about the vortex (inversely proportional to
rc ). Moreover, the core size is seldom either measured or calculated, so it is an input parameter of the analysis.
The approach thus is to model all effects possible in the theory, as accurately as possible; and then use the value of
rc to account for the actual viscous core radius, and also for all phenomena of the interaction that are not otherwise
modelled (or are inaccurately modelled). A goal for the development of better models is that the vortex core size
represent the actual physical core (10–20% chord) and nothing else.
Blade-vortex interaction on helicopter rotors in low-speed forward flight introduce additional considerations.
It has been observed that when vortex-induced loads are calculated using the core size that gives good correlation
at the blade tip, the strength of the blade-vortex interaction is significantly overpredicted for inboard radial stations.
An examination of measured low speed airloads indicates that the vortex-induced loading is high when the blade
first encounters the vortex, on the advancing side; decreases inboard as the blade sweeps over the vortex, on the
front of the disk; and then recovers again on the retreating side. The core size can be increased for collocation
points on the inboard part of the blade in order to eliminate (but not explain) this problem. Evidently there is some
phenomenon limiting the loads. Among the possibilities that have been proposed are the following.
a) Local distortion of the vortex geometry.

b) Bursting of vortex core, induced by the blade. The bursting must propagate upstream in
order to affect the blade-vortex interaction loading, but must not propagate too far.
c) Interaction of the vortex with the trailed wake it induces behind the blade, with the effect
of diffusing and reducing the circulation in the vortex.
d) Local flow separation produced by high radial pressure gradients on the blade. This is an
unsteady separation, introducing delays needed to reduce the blade-vortex interaction only on
the inboard part of the blade.
e) Viscous effects on the interaction, because the blade-vortex separation is much smaller on
the front of the disk than on the sides.
Note that (b), (c), and (d) do not provide a mechanism for the loads to recover on the retreating side; and that
(b) and (c) would also affect subsequent interactions of the vortex element. Increasing the core size for inboard
collocation points is a simple way to model the effects on the airloads, but no explanation of the physics is intended.
The exact physical mechanism involved remains speculative. More detailed measurements of the aerodynamics,
including the wake geometry, are needed to explore this phenomenon.
25-3.1.3 Nonuniform Inflow Calculation
The nonuniform inflow is calculated by integrating the Biot-Savart law over the wake. Hence the induced
velocities depend on the strength and geometry of the wake vorticity. The strength is defined by the span and
time variation of the bound circulation. The geometry can be obtained from rigid, prescribed, or free models.
Because of the complex wake geometry that can be produced by the wing motion, it is not possible to evaluate
the integrals analytically, even with no distortion. A direct numerical integration is not satisfactory either, because
the large variations in the integrand (produced by wing-vortex interactions) require small step sizes for accuracy.
Hence the wake is modelled by a set of discrete elements. The Biot-Savart law can be integrated analytically for
each element, and the total velocity obtained by summing contributions from all elements. This approach is both
accurate and efficient.
The tip vortices are modelled as a connected series of straight line segments, with some kind of core repre-
sentation. This is a good model for the most important part of the wake. The inboard trailed and shed vorticity
is typically modelled as a vortex mesh or lattice, or possibly vortex sheet elements. With a lattice (line segment)
model of the inboard wake, a large core size is needed, not as a representation of a physical effect, but to produce an
approximation for the sheet element, by eliminating any singularities in the velocity for close passage of following
wings. Note that a large-core line segment might also represent partially rolled-up inboard trailed vorticity.
Options for modelling the inboard wake elements include vortex sheets, either nonplanar-quadrilateral or
planar-rectangular; or line segments in the middle of the elements, with a large core to eliminate singularities. The
nonplanar-quadrilateral element is computationally expensive, and planar-rectangular elements introduce problems
with mismatched edges. Moreover it is found that line segments give about the same results as sheet elements.
Hence line segments are normally used, producing a vortex lattice model. Such a model for the inboard wake is
not as good as the tip vortex representation, but the inboard wake is not as important either, so a more approximate
model is acceptable. If the inboard wake model is important for a case, then a better model for the rollup process
will be needed too (to define the vorticity distribution over the inboard wake surface).
Modelling the wake by a set of discrete vortex elements introduces the following approximations: replacing
the curvilinear geometry by a series of straight line or planar elements; simplified distribution of strength over
an individual element (constant or linear); and sheets replaced by lines, or a planar-rectangle approximation for a
nonplanar-quadrilateral. A practical model must balance the accuracy and efficiency of such approximations.
A discretized wake model is developed by assuming that the wing bound circulation is known at discrete
points along the span and in the past. Thus the bound circulation Γ(r, t) is calculated at the aerodynamic span
stations rAi , i = 1 to M , and at the wake ages τk = k Δτ , k = 0 to K. A linear variation of Γ between these
points means the wing generates a wake of sheet elements behind the wing (figure 7). The strength of the trailed
and shed vorticity in an element can be obtained from Γ at the time that the vorticity was created. Thus the strength
can be characterized by the bound circulation corresponding to the four corners of the element (figure 8). The
element leading edge corresponds to Γ at (t − τk ), and the trailing edge to Γ at the earlier time (t − τk+1 ). A
spanwise change in Γ produces trailed vorticity δ. A linear variation of Γ means that the strength δ is constant
spanwise, but linear along its length. A time change in Γ produces shed vorticity γ. A linear variation of Γ means
that the strength of γ is constant in age, but linear along its length. The wake geometry provides the locations of
the four corners of the element.
An economical approximation for the sheet element is line segments, with a large core to avoid velocity
singularities near the line. Hence a vortex lattice model is obtained by collapsing all sheet elements to finite-
strength line segments (figure 9). The line segments are in the center of the sheet element, so the points at
which the induced velocity are calculated (collocation points) are at the midpoints of the vortex lattice grid, both
spanwise and in time. Locating the collocation points midway between the trailers is a standard practice, to avoid
the singularities at the lines. Locating the collocation points midway in time is required to correctly obtain the
unsteady aerodynamic effects of the shed wake.
With the sheets collapsed to lines, the strength of the line segments vary linearly along their length. A further
approximation is stepped (piecewise constant) variation in strength. Then instead of a linear variation along the
line, there is a jump in the strength at the center of the element, where the shed and trailed lines cross. This wake
model corresponds to a stepped distribution of the bound circulation, both spanwise and in time. There is little
reason to use such a stepped variation in strength, since the linear variation is as easy to implement.
Γ (rAM , t) τ=0
x sheet panels
e
ort (except tip vortex)
ndv
b ou
Γ (rA1, t)
τ = τk
vortex wake
panels
near
wake
tip
far vortex
wake
Figure 25-7 Wake model constructed from sheet panels (right tip vortex rollup).
Γ (rAi , t - τk ) Γ (rA(i+1) , t - τk )
rAi rA(i+1)
τk
δ
τk +1
Γ (rAi , t - τk +1) Γ (rA(i+1) , t - τk +1)
Figure 25-8 Vortex wake element.

dashed lines =
original sheet
rtex panels
vo
und
bo solid lines =
line segments
near
wake
tip
far vortex
wake
Figure 25-9a Wake model with line segments replacing sheet panels (right tip vortex rollup).
x
orte vortex line
dv segments
un
bo
near
wake
tip
vortex
far
wake
Figure 25-9b Wake model constructed from line segments (right tip vortex rollup).
Consider the wake model for collocation points on the wing that generated the wake. The wake directly
behind this wing is called the “near wake” here (the terms “near wake” and “far wake” are used in many ways
in the literature). The implementation of lifting-line theory involves primarily this region of the wake. The most
important requirement is to model the detailed variation of the wake strength, both spanwise and in time, in order
to accurately get the classical three-dimensional (Prandtl) and unsteady (Theodorsen) effects of the wake on the
wing loading. The rollup process is less important for the near wake.
A vortex lattice is used for the near wake, rather than sheet elements. Discretization of the wake is better
behaved with line segments than with sheet elements. With sheet elements, numerical difficulties arise from the
edge and corner singularities (particularly when planar-rectangular elements are used) and the fact that nonplanar-
quadrilateral elements can not be integrated analytically. Also, the most important case of the downwash from
the two sheet elements adjoining the collocation point would require a higher-order element or some other special
treatment.
The rollup process is not considered in modelling the near wake, except that the effect of the rollup on the
tip loading of the generating wing should be modelled. The spanwise location of the tip vortex formation can be
prescribed, forcing the wing to trail the remaining bound circulation into the wake at that span station. This effect
is implemented in the wing component.
For lifting-line theory, the geometry of the collocation points involves placing them at the three-quarter chord
in the local flow direction, in order to implement second-order lifting-line theory. Note that the collocation points
are part of the geometry of the wake rather than the wing (part of the outer problem not the inner problem). The
wake geometry is obtained from the position of the lifting line at the current and past time steps. Thus the wake
geometry can be used to get the local flow direction for the collocation points. A consequence of this approach is
that the collocation points are automatically kept away from trailed line segments of the near wake. In cases of
highly distorted geometry, especially near a reverse flow boundary, some close encounters between the collocation
points and wake vortex lines can still occur. The calculation can use a core size for the near wake line segments in
order to avoid singularities, but this core size has no physical significance, and indeed its value must not influence
the solution for the loading.
Optionally a straight lifting line can be used to generate the wake geometry, ignoring curvature of the actual
quarter chord. First order lifting-line theory (quarter chord collocation point) should use a straight lifting line
for consistency, since sweep effects are higher order. In this case, the wing geometry is generated by a straight
line from left tip to right tip, interpolated to the aerodynamic collocation points as required. Often the wing tip
locations are on an appropriate reference line, not the actual tips. This approximation has nothing to do with the
aerodynamic interfaces (velocity and force) between the structure and the aerodynamic components.
Consider the wake when it reaches collocation points on a following wing or other aerodynamic component.
This wake region is called the “far wake” here. For the far wake, it is most important to model the rollup
process. Since the wake quickly rolls up at the outer edges to form concentrated tip vortices, the tip vortices play
a dominant role in the aerodynamics, and must be modelled well. The inboard vorticity plays a lessor role, so
more approximations are acceptable. The wake rollup process is modelled rather than calculated, which means
that the structure and strength of the far wake are obtained from assumptions, based on the form of the spanwise
distribution of the bound circulation. While the bound circulation is calculated at many spanwise points on the
wing, the entrainment and stretching that occurs during the rollup process means that such detailed knowledge
of the vorticity strength is not available in the far wake. Hence the far wake model is constructed using the
bound circulation at a small number of points, typically just the circulation peaks. More information on the bound
circulation distribution can be retained by dividing the far wake trailed vorticity into several spanwise panels, and
using the circulation peaks in each panel.
Consider the single-peak case, when the bound circulation has the same sign (positive or negative) spanwise
for a given time. The spanwise maximum of the bound circulation is Γmax . It is assumed that in the far wake
(where the rollup process is complete) there are rolled-up tip vortices with strength equal to the value of Γmax at
the time that the wake element was created. If only the right tip vortex is rolled up, there is corresponding negative
trailed vorticity with total strength −Γmax in the inboard sheet. The tip vortex model then is a line segment with
this strength, and a small core radius (which is an input parameter). Any error in the assumed strength, because
the vortex is partially or over-rolled up, will be compensated for by the value of the core radius. In the absence of a
calculation of the stretch and rollup of the vorticity, the detailed structure of the inboard vortex sheet is unknown.
Therefore the inboard portion of the wake is modelled as a single sheet element with trailed and shed vorticity. If
both tip vortices roll up, this element has no trailed vorticity. If only the right tip vortex is rolled up, it is assumed
that the strength of the trailed vorticity is constant spanwise, with a total value of −Γmax . This is an efficient
model, consisting of only two or three elements with a total of three line segments at each age, which minimizes
computation; and depending only on Γmax , which minimizes storage.
This modelling approach can be extended to cases of a partially rolled-up tip vortex, or a dual-peak circulation
distribution. The rollup process may not be complete when the tip vortex encounters the following wing. Then
the strength of the tip vortex is less than Γmax , which reduces the vortex-induced loads. It is possible to introduce
a prescribed rollup, such that the tip vortex has only a fraction of the maximum strength. The remainder of the
trailed vorticity outboard of the peak is still in the inboard wake, so the inboard sheet has both positive and negative
trailed vorticity, divided at a span station that corresponds to the peak. Similarly, the stretching of the inboard wake
caused by the rollup can be prescribed. Hence the far wake model includes parameters that define the entrainment
and stretching in the tip vortex rollup process.
Consider the dual-peak case, when there are two peaks of opposite sign (left peak ΓL and right peak ΓR ) in the
spanwise distribution of the bound circulation. It is assumed that there is only one zero crossing of the circulation
strength between the left and right tips. If both tip vortices roll up, they are assumed to have strengths ΓL and
ΓR . These tip vortices have the same sign, and the inboard sheet has trailed vorticity of the opposite sign and total
strength ΓL − ΓR . If only the right tip vortex is rolled up, it is assumed to consist of the trailed vorticity outboard of
the right peak, hence strength ΓR . The inboard sheet has then both positive and negative trailed vorticity, divided
at a span station corresponding to the left peak. The trailed vorticity between the two peaks may partially roll up
as well, which can be modelled by using a line segment with a physically-meaningful core radius.
The far wake trailed vorticity can be divided into several spanwise panels, in order to provide more detailed
structure of the inboard vorticity. In particular, rolled up trailed vorticity can be modelled at the boundaries between
panels. The far wake panels are defined by specifying the panel boundaries at span stations rBi , i = 1 to M − 1
for M panels (where rB0 and rBM are always the wing tips). The specification of panel boundaries depends
primarily on the wing geometry, such as the edges of a trailing-edge flap, or changes in chord (figure 5). Hence
these boundaries are defined as part of the wing component. A single-peak model or dual-peak model is then used
for each of the far wake panels.
It is difficult to use this approach to introduce more detail in the model. The lack of a calculation of the rollup
process is a serious limitation. More detail usually means more empirical parameters to describe the structure. It
is hardly worth the effort to model the partially rolled-up wake, since there is seldom enough information to select
the modelling parameters, and the input value of the core radius compensates for errors in the tip vortex strength.
The dual-peak model is important however, since if a single-peak model is applied to the dual-peak circulation
distribution it can actually result in the wrong sign of the tip vortex strength. It is also necessary to be concerned
about cases with two or more zero crossings the spanwise distribution of Γ; local minima and maximum in Γ from
wing-vortex interaction; and rapid changes of Γ with time. Such cases need a calculation of the rollup, not an
empirical model.
25-3.1.4 Description of the Wake Model
The wake behind a wing is divided into near wake and far wake regions (figure 9). The wake panel extending
in age from τk to τk+1 is in one of these regions. The near wake model is only used for collocation points on
the wing that generated the wake. The elements in the near wake are always modelled using line segments. The
tip vortices in the far wake are modelled using line segments, and the inboard wake can be modelled using line
segments or sheet elements. The far wake wake region can have both tip vortices rolled up, just one tip vortex
rolled up, or neither; consists of one or more spanwise panels; and can use a single-peak or dual-peak model for
each panel. Partial entrainment and partial stretching can be included in the tip vortex rollup process. The extent
of the far wake region can be different for collocation points on and off the wing set.
The model for the near wake region is illustrated in figure 10. The wing bound circulation distribution
generates a wake panel consisting of sheet elements, and each sheet element is replaced by line segments. The far
wake models can be described in terms of an equivalent circulation distribution (figures 11 to 13). The structure of
the wake is actually determined by the rollup process, which here is based on assumptions and the bound circulation
peaks. The equivalent circulation distribution is that which would directly generate the rolled-up wake structure.
Although this equivalent distribution does not really exist, it is consistent with the convention of describing the
wake strength in terms of the bound circulation at the four corners of an element.
Figure 11 illustrates the far wake models for both tip vortices rolled up, or just the left or right tip vortex
rolled up, or neither; and for single-peak and dual-peak circulation distributions. Here a single spanwise panel is
considered, so the edges of the panel are the wing tips. The structure of the wake panels follows from the equivalent
circulation distribution: a step in circulation generates a line segment for the rolled-up tip vortex; a linear spanwise
change in circulation generates a sheet element (which may then be modelled by line segments); and a constant
spanwise circulation generates a sheet element with only shed vorticity. The dual-peak distributions are drawn as
if the right circulation peak is negative, ΓR < 0; the implementation allows arbitrary sign for both ΓL and ΓR .
In figure 11a, complete entrainment is assumed, so the tip vortex has a strength equal to the bound circulation
peak; and complete stretching is assumed, so the inboard vorticity extends not to the position of the peak but to
the tip vortex. The vorticity strength in the far wake region depends only on the peak bound circulation Γmax for
a single-peak model, or ΓL and ΓR for a dual-peak model. In the dual-peak model, the trailed vorticity between
the two peaks may partially roll up as well, which can be modelled by using a line segment with a physically-
meaningful core radius. For a more general model, it is assumed that the rollup process takes place over the wake
age τ = 0 to τRU = KRU Δτ . With partial stretching, the inboard vorticity extends to span station ρ, ρ varying
linearly from the circulation peak position r at τ = 0 to the tip position at τRU . With partial entrainment, the
Γ2
Γ1 ΓM
Γ=0 Γ=0
left rAi right
wing wing
tip tip
M+1 sheets
(sheet modelled by lines)
Figure 25-10 Wing circulation distribution and resulting wake panel model for near wake region.
single-peak model dual-peak model
actual wing circulation distribution

Γmax ΓL
left right left rR right

wing wing wing wing
tip r max tip tip rL tip
ΓR
equivalent circulation distribution
tip vortices rolled up:
Γmax ΓL
both
ΓR
Γmax ΓL
right
rL
ΓR
Γmax ΓL
left
rR
ΓR
Γmax
ΓL
neither
rR
r max rL
ΓR
resulting wake panels:

step = line (rolled up tip vortex)
linear = sheet
constant = shed sheet
= optional rolled up inboard trailed vorticity
Figure 25-11a Equivalent circulation distribution for far wake region;

complete entrainment and complete stretch in rollup.

Γmax ΓL

wing wing wing wing
ΓR
tip vortices rolled up:
Γmax ΓL
both
ρ
R
ρ
L
ΓR
Γmax ΓL
right
ρ
R
ρ rL
max
ΓR
Γmax ΓL
left
rR
ρ ρ
max L
ΓR
Γmax
ΓL
neither
rR
r max rL
ΓR
stretch of inboard vorticity because of rollup:

ρ = r at τ = 0, ρ = tip at τ = τ RU
Figure 25-11b Equivalent circulation distribution for far wake region;

complete entrainment and partial stretch in rollup.

Γmax ΓL

wing wing wing wing
ΓR
tip vortices rolled up: Γ ΓL
max
both fR Γmax
f LΓmax fL ΓL
ρ
R
r max ρ fR ΓR
L
Γmax
ΓL ΓR
right fR Γmax
ρ
R
ρ rL fR ΓR
max
Γmax ΓL ΓR
left fL Γmax fL ΓL
rR
ρ ρ
max L
Γmax ΓR
ΓL
neither
rR
r max rL
ΓR
entrainment of trailed vorticity in rollup:

f = f B at τ = 0, f = f E at τ = τ RU
(f = 0 no rollup, f = 1 complete rollup)
Figure 25-11c Equivalent circulation distribution for far wake region;

partial entrainment and partial stretch in rollup.
rolled up not rolled up
rBi rBi

(single-peak model or dual-peak model for each panel)
entire
wing
ρR
ρmax ρmax ρL
individual
far wake
panels
adjacent panels not rolled up and not wing tip:

independent sheet edge at peak of adjacent
panel ( ρ not r)
Figure 25-12 Equivalent circulation distribution for far wake region;

multiple spanwise panels

Γmax ΓL

wing wing wing wing
ΓR
single Γmax ΓL
panel
fR Γmax fL ΓL
f LΓmax
ρ
R
ρ ρ fR ΓR
max L
ΓR
tip vortex not rolled up: ρ = r and f = 0
(single peak: ρ = r only if neither tip rolled up)
rollup process: ρ = tip and f = 1 for complete rollup

stretch: ρ = r at τ = 0, ρ = tip at τ = τ
RU
entrainment: f = f B at τ = 0, f = f E at τ = τ RU
ρ = tip: omit inboard element

f = 0: omit tip vortex line
multiple
panels
not rolled up and not wing tip:

sheet edge at peak of adjacent panel (ρ not r)
Figure 25-13 Equivalent circulation distribution for far wake region; general
model.
tip vortex strength is a fraction f of the peak circulation, f varying linearly from fB at τ = 0 to fE at τ = τRU
(f = 0 for no rollup, f = 1 for complete rollup). The prescribed parameters KRU , fB , and fE define the rollup
process. The equivalent circulation distributions are shown in figure 11b for partial stretch of the inboard vorticity
because of the rollup; and in figure 11c for as well partial entrainment of vorticity in the rollup.
Figure 12 illustrates the far wake models with the trailed vorticity divided into several spanwise panels. The
single-peak model or dual-peak model is applied to each panel. Whether the tip vortices roll up has already been
specified. With multiple panels it is also necessary to specify whether the trailed vorticity rolls up at the boundaries
between panels. If rollup occurs at these boundaries, then the panels are independent of each other. The tip vortex
stretching model is also applied at these boundaries, but the partial entrainment model is not used here. If rollup
does not occur (and the panel edge is not a wing tip), then to obtain a piecewise linear distribution of equivalent
circulation over the entire wing, it is necessary to extend the sheet edge not to the panel edge, but to the position
(perhaps stretched) of the peak on the adjacent panel.
A general model for the far wake, encompassing all the cases in figures 11 and 12, is defined in figure 13.
If a tip vortex does not rollup, then f is zero and ρ is at the wing tip. Otherwise f and ρ are calculated from the
parameters defining the rollup process. If the factor f is zero, then the tip vortex line segment is omitted. If the
position ρ is at the tip, then a sheet element is omitted. The positions (ρmax , ρL , ρR ), which are obtained from the
positions of the bound circulation peaks (rmax , rL , rR ), may break the inboard sheet into two or three elements.
Hence (rmax , rL , rR ) are part of the wake geometry description, needed to calculate the influence coefficients.
These influence coefficients are used to evaluate the wing loading, which then provides new values of the location
of the circulation peaks. Therefore (rmax , rL , rR ) may be quantities that must be converged in the wake geometry
iteration. A particular wake model does not depend on all of these quantities however. For example, the dual-peak
model with just the right tip vortex rolled up depends only on rL (figure 11a); and the single-peak model with one
or both tip vortices rolled up depends on none of the peak positions (figure 11a).
An aspect of the single-peak and dual-peak models that must be considered further is the consequences of
using one model for a loading distribution that is more appropriate to the other model. The single-peak or dual-
peak model (together with the circulation distribution from the last loading evaluation) is used to calculate the
influence coefficients; then these influence coefficients and the current circulation distribution are used to evaluate
the induced velocity. Figure 14 illustrates the possible wake models with dual-peak loading and only the right tip
vortex rolled up. The single-peak model based on ΓR at least gets the correct sign and magnitude of the tip vortex
strength. The single-peak model based on Γmax is not appropriate for this example, but is an option required
when the loading is closer to a single-peak distribution. Figure 15 considers the case where the current circulation
distribution is dual-peak, but the influence coefficients were calculated using the single-peak model. This case can
occur because the use of the single-peak model has been prescribed; or because the circulation distribution from
the last iteration was single-peak (the option of using a dual-peak model with an arbitrary peak position is not
implemented). With only the right tip vortex rolled up, it is usually best to evaluate the induced velocity using the
right peak ΓR in place of Γmax (as shown in figure 14). For all of the rollup options, the trailed vorticity has the
wrong sign on the inboard sheet. With both tip vortices rolled up, one tip vortex has the wrong sign as well. Hence
it is important to use the dual-peak model when the calculated bound circulation has both positive and negative
values. Figure 16 considers the case where the current circulation distribution is single-peak, but the influence
coefficients were calculated using the dual-peak model. In this case a reasonable result is obtained by evaluating
the induced velocity using Γmax for ΓL and ΓR if both tip vortices are rolled up (or neither); or using Γmax for
ΓR and interpolating ΓL to rL if only the right tip vortex is rolled up. For all of the rollup options, the strengths
of the tip vortices are correct.
In general, the inertial frame I is the only common frame for the wake geometry and all collocation points.
Thus the induced velocity is calculated in I axes. The analysis assumes that the locations of the collocation points
and the wake geometry are measured relative the origin of I, in I axes.
The wake geometry and hence the influence coefficients may be constant (or periodic) in a frame moving with
the wing, even for cases involving unsteady loading. However, if the wing is moving relative to the inertial frame,
then the wake geometry in I axes is not constant (or periodic). If only the relative positions of the collocation
points and wake geometry are important, then it is still possible to use the wake geometry at just a single time. In
general, the wake geometry and influence coefficients must be calculated over the full range of time required by
the trim and transient solution procedures.
The wake consists of rolled-up tip vortices and an inboard sheet of the remaining vorticity, extending behind
the wing. The analysis can model cases where the wake rolls up at both tips. For a rotor, only the vortex from the
blade tip rolls up significantly. A rolled-up vortex from the root of the blade is seldom observed. The analysis can
thus model cases where the wake rolls up only at the right wing tip, which is appropriate for a counter-clockwise
rotating rotor. For clockwise rotation of the rotor, the convention used is that the span stations are still ordered
from root to tip (now from right to left), and the “right” and “left” rolled-up vortices are still at the tip and root.
So the definitions of “right” and “left” have been reversed. Reversing the order of the span stations changes the
sign of both trailed and shed vorticity in the wake. Hence the sign of the influence coefficients must be changed
when the order is from right to left.
CAMRAD II can be used with an external aeroacoustic analysis. The external analysis typically employs the
methods of computational fluid dynamics, possibly with a limited computational domain that does not encompass
the entire wing wake. The external analysis may also not calculate the effects of the structural dynamic motion
on the aerodynamic boundary conditions. Figure 17 illustrates the communication between CAMRAD II and an
external aeroacoustic analysis. The basic procedure, known as loose coupling, has the following steps: following
steps:
a) Run CAMRAD II to obtain the converged trim solution; save the wing loads (lift, drag,
and moment, including any prescribed increments) as ctotal ; and calculate the wing motion
using the appropriate structural dynamic sensors. Depending on the requirements of the
external analysis, may also calculate a partial angle-of-attack αP at the wing. The wake-
induced velocity used to evaluate αP is calculated excluding the wake vorticity that is in the
computational domain of the external analysis; and αP includes the effects of the structural
motion.
b) Run the external aeroacoustic analysis, using the wing motion (or αP ) to calculate the wing
loads cext . Calculate the loading increments (Δc)n+1 = (Δc)n + (cext − ctotal )n .
c) Run CAMRAD II to obtain the converged trim solution, using the prescribed loading
BLADE BOUND CIRCULATION
ΓL
rL
ΓR
FAR WAKE STRUCTURE
dual-peak model
ΓR
ρ GI
single-peak model, using outboard peak

ΓR
single-peak model, using peak with largest magnitude
ΓL
Figure 25-14 Wake models with dual-peak loading.

influence coefficients: circulation: resulting

single-peak model dual-peak approach approximation
tip vortices
rolled up:
both ρ
R use Γmax
rmax ρ rmax
L
right ρ use ΓR
R
ρmax
ρmax rL
left
rR use ΓL
ρmax ρ ρmax
L
neither
rR use Γmax
rmax rL rmax

influence coefficients calculated using single-peak model but circulation is
dual-peak.
influence coefficients: circulation: resulting

dual-peak model single-peak approach approximation
tip vortices
rolled up:
both ρ ΓL = ΓR = Γmax
R
ρ rmax ρ ρ
L L R
right ρ ΓR = Γmax
R
interpolate
rL ρmax ΓL to rL rL ρ
R
left
rR ΓL = Γmax
interpolate
ρ ρmax ΓR to r R ρ rR
L L
neither
rR ΓL = ΓR = Γmax
rL rmax rL rR

influence coefficients calculated using dual-peak model but circulation is
single-peak.
CAMRAD II
blade total loads

cl total cdtotal cmtotal prescribed
motion load
(including
increments
(perhaps increments) Δcl Δcd Δcm
partial
angle of
attack) Δcn+1 = Δcn + (cext – ctotal )
external loads cl ext cd ext cm ext
external aeroacoustic analysis
Figure 25-17 Communication with an external aeroacoustic analysis (loose coupling).

increments Δc.
d) Repeat the procedure as required to obtain a consistent solution from CAMRAD II and
the external analysis. The procedure has converged when the wing loading calculated by
CAMRAD II equals the wing loading calculated by the external analysis (so Δc would be the
same in the next cycle).
Often the external analysis requires a higher resolution in span, time, and wake age than CAMRAD II needs
to obtain a converged trim solution for the coupled aerodynamic and structural dynamic behavior. Hence it is
appropriate to calculate high resolution quantities in a separate loop in the trim task, using fixed wake geometry
and structural motion (and fixed circulation if the partial angle-of-attack is needed), after the converged solution
for the structural dynamic response has been obtained.
Simply setting the loads in the wing component to the values calculated by the external analysis would
not account for changes in the angle-of-attack as the wing motion and wake effects are updated. Therefore the
component uses the following expression for the lift coefficient:
c (α) = c ext (αold ) + c α (α − αold )

=c + c int (α) − c old (αold )
ext (αold )

= c int (α) + c ext (αold ) − c old (αold )
=c int + Δc
where c ext is the lift obtained from the external analysis (evaluated at αP ); c int is the lift calculated by this
component; and c old is the lift calculated by this component in the previous cycle (when αP was evaluated). Similar
expressions are used for the drag and moment coefficients. Hence communication with an external aeroacoustic
analysis can be implemented by introducing prescribed increments in the section aerodynamic coefficients: Δc ,
Δcd , Δcm .
If the external aeroacoustic analysis accounts for the entire flow field, then it requires just the wing motion
from CAMRAD II. Otherwise, the communication between CAMRAD II and the external aeroacoustic analysis
is based on the partial angle-of-attack, so that the wake in the computational domain of the external analysis is not
used twice. The external analysis may account for just the wake vorticity directly behind the wing; for tip vortices
from other wings as well; or for all vorticity inside the computational domain. This component must implement
the same options in excluding wake elements during the calculation of the induced velocity. A wing component
can then use this induced velocity to calculate αP .
The computational domain of the external analysis is here approximated by a box, defined by six planes.
The box axes (orthogonal) are specified by three vectors: two points on axes, plus the origin; or two unit vectors,
plus the origin. The box is generated by two planes normal to each axis. Computational domains of other shapes
could be similarly analyzed. However, for use by this component, it is only necessary that the domain boundary
be correct where it intersects the wake; and probably αP is not very sensitive to small errors in that intersection.
Hence the box may be a sufficient approximation for most computational domains encountered.
a) Location of collocation points: position rPI relative origin of inertial frame, in I axes. The position of a
collocation point on the wing set might be obtained from a wake geometry component. The position of a point
off the wing set is obtained directly from a structural dynamic component. For collocation points on the wing
generating the wake, the near wake model is used. For collocation points on the wing set generating the wake, a
correction for inboard wing-vortex interaction can be applied.
b) Wake geometry: the standard interface for wake geometry components. The wake geometry at time t is described
I
by the position of a set of nodes at age τk : rW k as a function of time t. The points required are the locations of
the tip vortices, the inboard sheet edges, and the aerodynamic collocation points (panel midpoints plus wing tips):
⎧
⎪ rT Lk left tip vortex
⎪
⎪
⎨ rT Rk right tip vortex
rW k = rSLk left edge inboard sheet
⎪
⎪
⎪
⎩ rSRk right edge inboard sheet
rAik wing span stations, i = 0 to M + 1
for k = 0 to K. The wake extent K can be different for the tip vortices, inboard sheet, and span stations. The
geometry at the wing span stations is only required for the near wake model. There is a set of (rT Lk , rT Rk , rSLk ,
rSRk ) values for each panel of the far wake trailed vorticity.
c) Bound circulation: Γ at all span stations (midpoints of the aerodynamic panels), for each wing.
d) Bound circulation peaks: magnitude (Γmax , ΓL , ΓR ) and position (rmax , rL , rR ) for entire wing and multiple
far wake panels; and centroid data (Γ, rC , rG , f , δ, set) for far wake panel edges; for each wing.
e) Wing set plane normal: k I , from a reference frame component. A filter component can be used to get the mean
normal. A filtered k I will not be a unit vector. This vector is required to construct the extended far wake for axial
flow. The normal to the wing set axes S, or the wing tip axes T can be used.
I
f) Box axis positions: rbox , from a structural dynamic component or a reference frame component. The box is the
computational domain of an external aeroacoustic analysis. The definition consists of three vectors: two points
on axes, plus the origin; or two unit vectors, plus the origin.
25-3.5 Induced Velocity and Influence Coefficients
The induced velocity is evaluated by integrating over all vorticity in the wake. For a fixed wake geometry
and incompressible flow, the Biot-Savart law gives a linear relation between the induced velocity and the wake
strength. The wake strength is linearly related to the wing bound circulation at the current and past times. Hence
the induced velocity produced by one wing at collocation point P takes the form:
$$
v(t) = C(t, τ ; r)Γ(t − τ ; r) dr dτ
where Γ is the wing bound circulation, and C the influence coefficients (calculated from the wake geometry and
modelling assumptions). Here C is actually a differential operator, since the wake strength depends on the time
and span derivatives of the bound circulation. The wing span variable is r. At time t, a wake element and its
position in the wake is identified by the wake age τ , such that t − τ is the time that the element was created. Hence
the strength of that element depends on the bound circulation at t − τ , and the induced velocity is evaluated by an
integral equation. The age goes from τ = 0 at the current wing position, to τ = ∞ far behind the wing.
The bound circulation is discretized spanwise, being evaluated at the aerodynamic span stations rAi , i = 1
to M . The wake must have a consistent spanwise discretization. The strength of the near wake depends on the
circulation at all span stations, but in the rolled-up wake the strength depends only on the peak circulation values
(left and right peaks, or maximum). The wake age is discretized by using the geometry and strength only at a set of
ages τk = k Δτ , k = 0 to K, for fixed wake age increment Δτ . Thus the integral equation for the wake-induced
velocity becomes:
% KF
I
vA (t) = CL (t, τk )ΓL (t − τk ) + CR (t, τk )ΓR (t − τk )
wings panels k=0
KN M &
+ CN (t, τk ; rAi )Γ(t − τk ; rAi )
k=0 i=1
including a summation over all wings of this component. If the far wake trailed vorticity is divided into several
spanwise panels, then there is a summation over all panels as well. The induced velocity is calculated at a set of
time values t , as determined by the solution procedure. Generally the influence coefficients should be evaluated
at the same time values, and the bound circulation is evaluated at t − τ as required.
The influence coefficients at time t consist of a set of values at age τk : Ck as a function of time, in I axes.
The coefficients required are from the far wake (right and left circulation peaks) and near wake (circulation at all
span stations): ⎧
⎪ CRk far wake, right peak
⎪
⎪
⎨ CLk far wake, left peak
Ck = rGIk span station of inboard circulation peak
⎪
⎪
⎪
⎩ rGOk span station of outboard circulation peak
CN ik near wake, wing span stations, i = 1 to M
for k = 0 to KF for the far wake, and k = 0 to KN for the near wake. There is a set of (CRk , CLk , rGIk , rGOk )
values for each panel of the far wake trailed vorticity. The vector of the influence coefficients (component input
and output) consists of all these quantities, for the velocity at one collocation point from the wake of one wing.
The vector is calculated as a function of time, as defined by the solution procedure. The quantities present in the
influence coefficient vector depend on the wake model. All quantities are present for collocation points on the
wing and a dual-peak model. In other cases:
a) For collocation points off the wing set, a different far wake extent KD is used.
b) For a single-peak model, CL and rGI and rGO are not used.
c) For collocation points off the wing that generated the wake, CN is not used.
d) For periodic circulation, the maximum age in the influence coefficient vector is one period.
The quantities rGI and rGO have the following values for single-peak and dual-peak circulation distributions, with
the various rollup options.
dual-peak single-peak
rGI rGO rGI rGO
right tip rollup rL ρR < rLtip ρmax
left tip rollup rR ρL < rLtip ρmax
both or neither rL rR < rLtip rmax
Note that rGI is always at an aerodynamic panel midpoint, hence never at a wing tip. A value of rGI outside the
wing span indicates the single-peak model was used for the influence coefficients at this wake age. The influence
coefficient vector stores the values of rGI and rGO because they may be needed to evaluate the induced velocity.
The component sensor gives access to rL and rR .
The wake vorticity strength at time t consists of a set of values at age τk : Gk as a function of time. The wake
strength is calculated from the wing bound circulation. The far wake requires the peak values of the circulation,
and the near wake requires the circulation at all span stations:
⎧
⎪ GRk = ΓL (t − τk ) far wake, right peak
⎨
GLk = ΓR (t − τk ) far wake, left peak
Gk =
⎩ GM k = Γmax (t − τk )
⎪ far wake, maximum
GN ik = Γ(t − τk ; rAi ) near wake, wing span stations, i = 1 to M
for k = 0 to KG = max(KF , KD ) for the far wake, and k = 0 to KN for the near wake. For periodic circulation,
the maximum age required is one period. There is a set of (GRk , GLk , GM k ) values for each panel of the far
wake trailed vorticity. The vector of the wake vorticity strength (component input and output) consists of all these
quantities, for the wake of one wing. The vector is calculated as a function of time, as defined by the solution
procedure.
The wake-induced velocity could be calculated directly from the wake geometry and the bound circulation.
The component has the option of calculating the influence coefficients internally, whenever the velocity is evaluated,
instead of using separate interfaces. Internal calculation of the influence coefficients reduces the memory required,
and also is more efficient for the transient task. For the trim task however, the capability to separately calculate
the influence coefficients and the wake vorticity strength is required for efficiency. The influence coefficients need
to be updated only when the wake geometry changes significantly during the solution process. The wake vorticity
strength need be evaluated only for each wing, not for each velocity collocation point. The wake age increment
Δτ must be the same as used in the wake geometry description. The wake age increment is not necessarily related
to the increment in time t that the solution procedure uses. Often the solution procedures are defined such that the
bound circulation Γ and the induced velocity v are solved using the same time increment, which is the same as the
wake age increment. For the general case, the standard methods for interpolating system response allow Γ and v
to be evaluated at arbitrary times.
The influence coefficient vector is calculated at a set of time values t , as determined by the solution procedure.
The vector consists of the influence coefficients at a set of age values τk . When the influence coefficients are used
to evaluate the velocity, the following access methods are possible:
a) Direct access: C used only at times t identical to the calculated values t .

b) Constant: C at the first t used for all time t.
c) Evaluate at time: C evaluated at the required time t from the response calculated at times
t (standard linear interpolation in time).
d) Interpolate: velocity interpolated to the required time t (rather than interpolating C).
The solution procedure can be defined so direct access of the influence coefficients in the subvector is proper.
Direct access is most efficient, but requires coordination of the solution procedures for the influence coefficients
and the induced velocity. Specifically, the time and time step used to calculate the velocity must be the same as
the time and time step used to calculate the influence coefficients. With direct access, the response for C must
not have a nominal or reference. Note that if the wake geometry is constant in a frame moving with the wing, and
only the relative positions of the collocation points and wake geometry are important, then the constant option can
be used even if the geometry is not actually constant in inertial axes. Evaluating the influence coefficients at an
arbitrary time is the standard way for components to communicate, but is an approximation (unless it is actually
equivalent to direct access). Interpolating the velocity rather than the influence coefficients is least efficient, but
is necessary if the velocity is to be calculated at times that are independent of the times at which the influence
coefficients were calculated.
Similarly, when the wake vorticity strength is used to evaluate the velocity, it can be obtained by direct access,
by evaluating at the required time, or by interpolation. The option of assuming that the wake vorticity strength is
constant is not appropriate (and not implemented).
The standard procedures for evaluating the system response interpolate the influence coefficient and wake
vorticity strength vectors in time. It is the velocity that should actually be interpolated:
C(t , τk ) and Γ(t − τk ) used to evaluate v(t )

C(t +1 , τk ) and Γ(t +1 − τk ) used to evaluate v(t +1 )
then v(t ) and v(t +1 ) interpolated to v(t)
If instead the influence coefficients and wake vorticity strength are interpolated:
C(t , τk ) and C(t +1 , τk ) interpolated to C(t, τk )

Γ(t − τk ) and Γ(t +1 − τk ) interpolated to Γ(t − τk )
then C(t, τk ) and Γ(t − τk ) used to evaluate v(t)
The two results are not quite the same. Specifically, consider interpolating
v(t ) = C(t , τk )Γ(t − τk )

k=0
to arbitrary time t. Find such that t ≤ t ≤ t +1 ; and calculate the weights w and w +1 for linear interpolation.
The interpolated velocity is:
v(t) = w v(t ) + w +1 v(t +1 )

% & % &
=w C(t , τk )Γ(t − τk ) + w +1 C(t +1 , τk )Γ(t +1 − τ k )
k=0 k=0
% &
= w C(t , τk )Γ(t − τk ) + w +1 C(t +1 , τk )Γ(t +1 − τk )
k=0
% &

= w C(t , τk ) + w +1 C(t +1 , τk ) w Γ(t − τk ) + w +1 Γ(t +1 − τk )
k=0
% &

+ ww +1 C(t +1 , τk ) − C(t , τk ) Γ(t +1 − τk ) − Γ(t − τk )
k=0
using w +1= 1 − w . Interpolating the influence coefficients and wake vorticity strength instead is equivalent to
neglecting the second line in the last result. For the option to interpolate the velocity, the influence coefficients and
wake vorticity strength are required at t and t +1 . The influence coefficients are evaluated at times t = t0 + Δt.
First is found such that t ≤ t ≤ t +1 :
% &
t − t0
=
Δt
t − t0
w +1 = − =1−w
Δt
(brackets indicating integer truncation). The velocity is evaluated separately at t and t +1 . The wake vorticity
strength vector contains the bound circulation as a function of past time (wake age τk ). Hence the standard
evaluation of the response at t or t +1 , which uses linear interpolation in time, gives the correct result. Since t
are the times of the influence coefficient response, the influence coefficients are obtained by direct access. Then
v(t) = w v(t ) + w +1 v(t +1 ) =w C(t , τk )Γ(t − τk ) + w +1 C(t +1 , τk )Γ(t +1 − τk )

k=0 k=0
is the required velocity.
25–4 Theory
25-4.1 Induced Velocity Calculation
The induced velocity at a collocation point rP must be evaluated at time t, from the influence coefficients at t
and the bound circulation at t and all past time. The bound circulation is discretized spanwise, being evaluated at
the aerodynamic span stations rAi , i = 1 to M . The wake has a consistent spanwise discretization. The strength
of the near wake depends on the circulation at all span stations, but in the rolled-up wake the strength depends
only on the peak circulation values (left and right peaks, or maximum). The wake age is discretized by using
the geometry and strength only at a set of ages τk = k Δτ , k = 0 to K, for fixed wake age increment Δτ . The
wake age increment can be specified directly, or in terms of an azimuth increment Δφ: Δτ = |Ω| Δφ, where the
rotational speed Ω is obtained from a specified period. Thus the wake-induced velocity is:
% KF
vAI
(t) = CL (t, τk )ΓL (t − τk ) + CR (t, τk )ΓR (t − τk )
wings panels k=0
KN M &
+ CN (t, τk ; rAi )Γ(t − τk ; rAi )
k=0 i=1
including a summation over all wings of this component, and perhaps over all far wake panels. The terms present
depend on the wake model. All terms are present for collocation points on the wing, a dual-peak model, and left
or right tip rollup. In other cases the evaluation of vA must consider the following.
a) For collocation points off the wing set, a different far wake extent KD is used. For collocation points off the
wing that generated the wake, CN is not used. For periodic circulation, the maximum age in the summation is one
period.
b) For a single-peak model, CL does not exist, and CR is used with the maximum bound circulation Γmax .
Optionally with only the right tip vortex rolled up, ΓR can be used instead of Γmax (if two peaks are found), so that
at least the tip vortex strength is correct in the case of dual-peak loading. Similarly, with only the left tip vortex
rolled up, ΓL can be used instead of Γmax .
c) For a dual-peak model and both tip vortices rolled up (or neither), if only one peak is present in the bound
circulation distribution, then ΓL = ΓR = Γmax is used.
d) For a dual-peak model and only the right tip vortex rolled up, if only one peak is present in the bound circulation
distribution, then
ΓR = Γmax

rGI − rbound
ΓL = Γmax
rGO − rbound
are used to evaluate the velocity. Here rbound = rLtip is the span station of the left tip; and rGI and rGO are
the peak locations used to calculate the influence coefficients (stored with C). The interpolated value of ΓL is
intended to make the dual-peak model give the single-peak result (although actually rGI at both the leading edge
and trailing edge of the wake panel affect the induced velocity produced by the circulation at both the leading edge
and trailing edge). Similarly with only the left tip vortex rolled up:
ΓL = Γmax

rbound − rGI
ΓR = Γmax
rbound − rGO
are used, where rbound = rRtip is the span station of the right tip. With multiple far wake panels, and no rollup
at an interior boundary, the span station at the circulation peak in the adjacent panel replaces the tip span station.
Thus for right rollup, rbound = rGIprev , where rGIprev is the peak of the previous panel (for which the right
vortex is not rolled up); if rGIprev is outside the wing span (single-peak model) rGOprev is used instead. Hence
depending on the configuration at the panel tip that is not rolled up:
left not rolled up right not rolled up

wing tip rLtip rRtip
interior, dual-peak rGI previous panel rGI next panel
interior, single-peak rGO previous panel rGO next panel
gives the value of rbound used to interpolate the bound circulation. Figure 18 outlines the evaluation of the
bound circulation for the induced velocity calculation. Note that ΓL = 0 (with ΓR = Γmax ) identifies single-peak
circulation distribution; and rGI outside the wing span identifies single-peak influence coefficients in the dual-peak
model.
Lifting-line theory is applicable for sweep that varies slowly along the span (unless a special deriviation
and implementation is used), so kinks in the quarter-chord locus can result in incorrect prediction of the induced
velocity. An empirical correction is introduced to compensate: a factor Fsk on the near-wake contribution to the
induced velocity vA at collocation point rAk . This factor is proportional to the change in sweep, calculated as
described for the lifting line wing component.
25-4.2 Influence Coefficient Calculation
25-4.2.1 Process
The induced velocity at a collocation point rP from the wake of one wing is evaluated at time t by summing
over all elements in the wake model:
K−1
v(t) = Δv(t)
k=0 span
where
a) the sum in k is over the wake age, of extent K (maximum of KN , KF , and KD );

b) the spanwise sum is over all wake elements in the panel from age τk (leading edge) to τk+1
(trailing edge);
c) Δv is the velocity produced by an individual wake element.
single-peak model (only CR exists)

single-peak circulation: Γmax
dual-peak circulation (ΓL = 0):
both tip vortices rolled up (or neither): Γmax
right tip vortex rolled up: ΓR (option Γmax )
left tip vortex rolled up: ΓL (option Γmax )
dual-peak model (CR , CL , rGI , rGO exist)
single-peak influence coefficients (rGI outside wing span; CL not used)
single-peak circulation: Γmax
dual-peak circulation (ΓL = 0):
both tip vortices rolled up (or neither): Γmax
right tip vortex rolled up: ΓR
left tip vortex rolled up: ΓL
dual-peak influence coefficients
dual-peak circulation (ΓL = 0): ΓL , ΓR
single-peak circulation:
both tip vortices rolled up (or neither): ΓL = ΓR = Γmax
right tip vortex rolled up: ΓR = Γmax , interpolate ΓL
left tip vortex rolled up: ΓL = Γmax , interpolate ΓR
Figure 25-18 Outline of evaluation of bound circulation for induced velocity calculation.
The strength of an element is defined in terms of the bound circulation at its corners, so the velocity is a linear
combination of the circulation:
Δv(t) = ΔC Γ(t − τ ; r)
and the contributions ΔC can be added to the appropriate influence coefficients. For panels in the near wake
region, there are M + 1 wake elements, where M is the number of aerodynamic span stations; and ΔC contributes
to the near wake influence coefficients CN of the circulation at rAi . For panels in the far wake region, there are
2–5 wake elements, and ΔC contributes to the far wake influence coefficients CL and CR of the peak circulation.
Optionally, it may be assumed that the circulation is periodic, with a period equal to J wake age increments.
No change to the analysis of the wake elements is required, but now k is cyclically reduced to the range 0 to
k = J − 1 when identifying the influence coefficient for ΔC. Thus the storage required for the far wake influence
coefficients is compressed to one period, and the computation required to evaluate the induced velocity is reduced.
The geometry at time t is required to analyze the wake. The position of the collocation point rP is known,
and whether it is on or off the wing that generated the wake, or on or off the wing set. The wake geometry is
evaluate circulation peak positions

evaluate collocation point position
zero all influence coefficients
near wake
evaluate wake geometry at wing: rte at τ = 0
wake age loop: kte = 1 to KN
kle = kte − 1
wake geometry at leading edge: rle = rte
evaluate wake geometry at trailing edge: rte at τ = kte Δτ
near wake ΔC
far wake, each trailed vorticity panel
evaluate wake geometry at wing: rte at τ = 0 (or KN )
if collocation point not on this wing: bound vortex ΔC
wake age loop: kte = 1 (or KN +1 ) to KF
kle = kte − 1
wake geometry at leading edge: rle = rte
evaluate wake geometry at trailing edge: rte at τ = kte Δτ
far wake ΔC
if axial flow model and kte ≥ KF − J + 1: extended far wake ΔC
Figure 25-19 Outline of process for calculating influence coefficients.
described by the position of a set of nodes at age τk . The points available are:
⎧
⎪
⎪
⎪
⎪
⎪
⎩ rSRk right edge inboard sheet
where the wing span stations correspond to the collocation points at the midpoints of the aerodynamic panels, plus
the wing tips. There is a set of (rT Lk , rT Rk , rSLk , rSRk ) values for each panel of the far wake trailed vorticity.
Figure 19 is an outline of the process for calculating the influence coefficients. The wake panel extending
from kle to kte consists of one or more vortex elements. The extended far wake model may not be used at all.
For collocation points off this wing set, the far wake extent is KD ; other parameters such as core size and vortex
element modelling options are also different for collocation points on and off the wing set.
25-4.2.2 Wake Panels
Each wake panel consists of a number of line segments and sheet elements. Figure 20 shows the convention
for describing the strength and geometry of an element. The geometry is defined by vectors to each of the numbered
points (r1 , r2 , r3 , and r4 ) and a vector to the collocation point (rP ). Associated with each numbered point is the
bound circulation that the wing had when the vortex element was created (Γ1 , Γ2 , Γ3 , and Γ4 ). Then the velocity
line segment
1 leading edge, τk
Γ1
Γ2
2 trailing edge, τk +1
sheet element
Γ3 Γ1
3 1 leading edge, τk
Γ4 Γ2
4 2 trailing edge, τk +1
left edge right edge
Figure 25-20 Description of vortex line segment and sheet element.

induced by a vortex line segment is:

Δv = Γ1 Δv1 + Γ2 Δv2
The velocity induced by a vortex sheet element is:
Δv = Γ1 Δvt1 + Γ2 Δvt2 + Γ3 Δvt3 + Γ4 Δvt4

= (Γ1 − Γ3 )Δvt1 + (Γ2 − Γ4 )Δvt2
for the trailed vorticity, and
Δv = Γ1 Δvs1 + Γ2 Δvs2 + Γ3 Δvs3 + Γ4 Δvs4

= (Γ1 − Γ2 )Δvs1 + (Γ3 − Γ4 )Δvs3
for the shed vorticity.
The wake panel with leading edge at τk is in the near wake or far wake region. The models for these wake
regions are discussed above. Figure 21 describes the elements in a near wake panel. The geometry is described
by positions at the wing tips (rA0 and rA(M +1) ) and at the aerodynamic span stations (rAi , for i = 1 to M ). The
panel consists of M + 1 sheet elements. Each sheet element is modelled by a trailed line segment and a shed line
segment.
Figure 22 describes the far wake model, for single-peak and dual-peak circulation distributions. The equivalent
circulation distributions are obtained from figure 13. The geometry is described by positions at the tip vortices
(rT L and rT R ), and at the edges of the inboard sheet (rSL and rSR ). The positions at span stations ρmax , ρL ,
and ρR are obtained as required by linear interpolation between the inboard sheet edges. With multiple far wake
panels, if the trailed vorticity is not rolled up and not at the wing tip, then the sheet edge is at the peak of the
adjacent panel. The panel consists of tip vortex line segments, and sheet elements for the inboard wake.
For the dual-peak model, the inboard trailed vorticity can optionally be modelled as a sheet (not rolled up)
or as a line segment with an appropriate core (rolled up). If the bound circulation distribution does not have two
peaks, the single-peak model is used instead for this wake panel.
For collocation points off this wing, the velocity caused by the wing circulation is also included. The wing
is modelled as a bound vortex, consistent with the far wake model (figure 22). The geometry is described by
positions at the edges of the inboard sheet (rSL and rSR ). The positions at span stations ρmax , ρL , and ρR are
obtained as required by linear interpolation. The bound vortex consists of one or two vortex line segments. A core
radius (default value equal to 25% of the wing mean chord) is used to avoid excessive velocities near the bound
vortex. This representation of the flow about the wing is appropriate for the far field only, and is not accurate close
to the wing surface.
The cases with partial or no stretching in the rollup process all have the same number of inboard wake sheet
elements or bound vortex segments (two for the single-peak model, three for the dual-peak model). The tip vortex
line does not exist if f = 0; the trailed vorticity middle sheet of the dual-peak model may be rolled up; and with
multiple far wake panels, the sheet edge is extended to the peak of the adjacent panel (if not rolled up and not at
the wing tip). Cases with complete stretching in the rollup process (identified by ρmax , ρL , or ρR at the panel
boundary) require special treatment since the inboard wake sheet element next to the rolled up tip vortex does not
exist. In these cases, partial entrainment (defined by f < 1) is included in the far wake model, but not in the bound
vortex.
0 Γ1 Γ2 ΓM 0
le
te
rA0 rA1 rA2 rAM rA(M+1)
rA(i -1) rEi rAi
3 1 leading edge
2
trailed
1 2
shed
1
4 2 trailing edge
Figure 25-21 Description of near wake panel.

equivalent bound circulation

Γmax
fL Γmax fRΓmax
left wing tip ρ max right wing tip
bound vortex
fLΓmax Γmax fRΓmax
1 2 1 2 wing
rSL ρ max rSR
far wake
fL Γmax Γmax fRΓmax
2 3 1 3 1 1 le
1 4 2 4 2 2 te
rTL rSL ρ max rSR rTR
Figure 25-22a Description of far wake panel, single-peak model.

equivalent bound circulation
ΓL
fLΓL
ρR right wing tip
left wing tip ρL

fRΓR
ΓR
bound vortex
fLΓL ΓL ΓR fRΓR
1 2 1 2 1 2 wing
rSL ρL ρR rSR
far wake
fLΓL ΓL ΓR fRΓR
2 3 1 3 1 3 1 1 le
1 4 2 4 2 4 2 2 te
rTL rSL ρL ρR rSR rTR
Figure 25-22b Description of far wake panel, dual-peak model.

25-4.2.3 Panel Geometry
The geometry of the near wake panel element is shown in figure 21. The positions of the four corners
correspond to the aerodynamic span stations of the wing rAi (at the wing panel midpoints), except for the first
element (left edge at wing tip rA0 ) and the last element (right edge at wing tip rA(M +1) ). In the near wake, the
vortex sheet element is always modelled by line segments in the middle of the element. The end points of the
trailed line are on the element leading and trailing edges, interpolated to the wing aerodynamic panel edge rEi (so
the collocation points are always equidistant from the nearest trailed lines):
r 2 = wr1 + (1 − w)r3
r 1 = wr2 + (1 − w)r4
where
rEi − rA(i−1) ΔrA(i−1)
w= =
rAi − rA(i−1) ΔrAi + ΔrA(i−1)
with ΔrAi the wing panel width. However for the first near wake element, r = rA0 ; and for the last near wake
element, r = rA(M +1) . The end points of the shed line are on the middle of the element side edges, and moved
1/4 chord aft of the collocation point:
1 r2 − r1
r 2 = (r1 + r2 ) + di
2 |r2 − r1 |
1 r4 − r3
r 1 = (r3 + r4 ) + di−1
2 |r4 − r3 |
where di = 3ci /4, using the chord ci at the aerodynamic span station (di = ci /4 if the collocation points are at
the quarter chord rather than the three-quarter chord).
Optionally the bound vortex (panel leading edge for zero wake age) can be straightened relative the collocation
point for the induced velocity from the near wake. Consider the leading edge position rL and trailing position rT
corresponding to span station rAi . The direction of the trailed line at the collocation point is a = rLP − rT P . The
direction of the inboard part of the bound vortex is b = rLIB −rL0 (using rLIB near the root). Then m = a−( a·b b·b )b
is a vector perpendicular to the bound vortex (vector b), in the wake plane (vectors a and b). Then
m · (rL − rLP )
rLs = rL − (rL − rT )
m · (rL − rT )
extrapolates the leading edge point to a plane through rLP , perpendicular to the wake plane (m · (rLs − rLP ) = 0).
The wake model may require the panel geometry corresponding to a span station ρ between the wing tips (or
panel boundaries) where the inboard wake is divided into separate sheet elements. The position at ρ is obtained
by linear interpolation between the positions of the inboard sheet edges:
r = wrSL + (1 − w)rSR
where
rRtip − ρ
w=
rRtip − rLtip
and rLtip and rRtip are the span stations of the left and right tips (or panel boundaries). These equations extrapolate
the positions of the inboard sheet edges if ρ is outside the panel span (as required with multiple far wake panels,
and no rollup at the panel boundaries). The calculation of ρ begins with the span stations of the circulation peaks.
The position of the circulation peak is the average of the values at the panel leading and trailing edges:
1
ρL = [ρL (t − τk ) + ρL (t − τk+1 )]
2
1
ρR = [ρR (t − τk ) + ρR (t − τk+1 )]
2
If only one peak is found at the trailing edge and the left peak is the maximum at the leading edge, then
1
ρL = [ρL (t − τk ) + ρmax (t − τk+1 )]
2
1
ρR = [ρR (t − τk ) + rRtip ]
2
The positions for other cases are found in a similar manner. If only one peak is found at both the leading edge and
the trailing edge, then the single-peak model is used. If a tip vortex does not roll up, then the circulation factor f is
zero and the position ρ is at the wing tip. Otherwise f and ρ are calculated from the parameters defining the rollup
process. The rollup process is assumed to take place over the wake age τ = 0 to τRU = KRU Δτ . With partial
entrainment, the circulation factor f varies linearly from fB at τ = 0 to fE at τ = τRU . With partial stretching,
the position ρ varies linearly from the circulation peak position r at τ = 0 to the wing tip position rtip at τ = τRU .
Thus
f = (1 − w)fB + wfE
ρ = (1 − w)r + wrtip
where w = τk /τRU . If the factor f is zero, then the tip vortex line segment is omitted. If the position ρ is at the
tip, then a sheet element is omitted.
25-4.2.4 Vortex Core Radius
A vortex core radius must be specified for the tip vortices. Different values of rc can be used for collocation
points on and off the wing set. Core radii are also required for the near wake and bound vortex line segments, to
avoid singular velocities. If the inboard trailed wake is rolled up at panel boundaries or in the dual-peak model,
a core radius is required. Core radii are required for line segments that are used to approximate a sheet element.
The analysis implements default values for most of these quantities.
Several models are implemented for the vorticity distribution in the vortex core. Consider the tangential or
circumferential velocity v about a line vortex of strength Γ, at a distance r from the line. The core radius rc is
defined as the distance r at which the maximum value of v is encountered. For a potential line vortex (no core),
v = Γ/2πr. For small r, viscosity reduces the magnitude of v, by spreading the vorticity over a nonzero domain
instead of a line. With a finite core, let f (r)Γ be the circulation at r (f (0) = 0 and f (∞) = 1); so v = f Γ/2πr.
Then for fixed strength Γ and core radius rc , the maximum velocity about the vortex is determined by the value of
f at rc . The peak velocity magnitude is reduced as the vorticity outside of rc is increased.
A Rankine vortex core produces solid body rotation of the fluid inside rc , by having a uniform vorticity
distribution concentrated entirely within the core radius. The circulation factor is f = min(r2 /rc2 , 1). A Rankine
core produces the largest possible peak velocity for a given Γ and rc . For a distributed vorticity core, one simple
possibility has the circulation proportional to f = r2 /(r2 + rc2 ), so half the vorticity is outside the core radius.
This is the Scully model (ref. 1). Vatistas (ref. 2) defined a power-law core using f = r2 /(r2n + rc2n )1/n , which
gives a peak velocity of v = 2−1/n (Γ/2πrc ). This peak velocity increases as the exponent n increases. The
Bagai-Leishman model has n = 2 (ref. 3). Oseen (ref. 4) obtained the solution of the Navier-Stokes equations for
decay of a laminar vortex:
Γ 2
Γ 2 2

v= 1 − e−r /4νt = 1 − e−ar /rc
2πr 2πr
√
introducing the core radius rc2 = a4νt (a = 1.2564 is the solution of ea = 1 + 2a; so rc = 2.2418 νt). Thus for
the Oseen model, the circulation factor is
2 2

f = 1 − e−ar /rc
At the core radius, f = 2a/(1 + 2a). Figure 6 showed the influence of core type on the velocity. Note that the
Bagai-Leishman and Oseen models give similar velocity distributions.
The vortex core radius can be constant, or it can grow with wake age. The Oseen solution of the Navier-Stokes
√
equations for decay of a laminar vortex gives rc = 2.2418 ντ , where τ is the wake age (time since the vortex
element was created). This expression is generalized by allowing an arbitrary exponent, and adding an initial
radius at zero age. Thus the power-law model for core radius growth is
n
rc rc0 τ
= +
c c τ1
where τ1 is the wake age where the radius equals the mean chord c. A maximum value of the core radius can
also be specified. The linear and square-root models use n = 1 and n = 0.5 respectively, with an appropriate
τ1 . The Oseen model for core radius growth has n = 0.5 and τ1 = c2 /(2.24182 ν) (so the dimensionless age is
proportional to the Reynolds number). The Oseen model implies very slow growth of the core radius however. The
Bagai-Leishman model for core radius growth introduces an additional factor δ, such that τ1 = c2 /(2.24182 νδ)
with n = 0.5. A value of δ = 104 gives τ1 on the order of one rotor revolution, with which the core radius
increases enough to produce good convergence of rotor wake geometry calculations (ref. 3). However, a value of
δ = 10 gives good correlation with the vortex core growth measured on a model rotor (ref. 5).
The vortex core radius can also include a term that scales with the trailed vorticity moment. Hence the general
expression for the core radius rc is:
n
rc0 τ
rc = c +c + fM rG
c τ1
where rG is the second moment about the centroid (the radius of gyration) of the trailed vorticity, evaluated at the
time the vortex element was created (t − τ ). This moment is obtained by integrating the vorticity for all adjacent
trailed lines of the same sign. Equating the moment of the trailed vorticity created behind the wing to the moment
of the rolled up vortex implies that the constant fM should be on the order of 1.0 (depending on the vorticity
distribution in the vortex core).
The loading on a helicopter rotor in low-speed forward flight exhibits the influence of some phenomenon that
suppresses vortex-induced loads on inboard collocation points of the blade (on the front of the rotor disk). The core
size can be increased for collocation points on the inboard part of the blade in order to model this phenomenon.
Outboard of a specified span station r2 , the core size rcO is used (the normal tip vortex core); inboard of a specified
span station r1 , the core size rcI is used (a much larger core); in between r1 and r2 , the core size is obtained by
linear interpolation. So for a collocation point at span station rAi :
6r rAi ≥ r2
cO
rc = rcI rAi ≤ r1
wrcI + (1 − w)rcO r1 < rAi < r2
where
r2 − rAi
w=
r2 − r1
This core size is used for tip vortex elements in the far wake region, for collocation points on this wing set.
Increasing the core size is a simple way to suppress the vortex-induced loading, but no explanation of the physics
of the phenomenon is intended by this approach.
25-4.2.5 Axial Flow Wake Model
An accurate calculation of the induced velocity of a rotor in axial flight requires consideration of the wake
very far from the rotor disk. The detailed far wake model described above must be used near the rotor disk (for
typically four or five spirals), but thereafter a more approximate model is acceptable. Hence an efficient extended
far wake model is available for axial flow. The extended far wake model can only be used for a rotor: only the
right tip vortex is rolled up; and the circulation is periodic, with a period equal to J wake age increments. The
extended far wake model is intended to be used with a single far wake panel.
The extended far wake is constructed by extrapolating the last revolution of the far wake (the last J wake
panels). It is assumed that the wake is convected in the direction of the wing set plane normal k I (preferably wing
tip axes T, but possibly wing set axes S). The axial distance h that the wake is convected in one revolution (J wake
age increments) is calculated from the tip vortex geometry at the beginning and end of the last revolution of the
far wake:

h = − (k I )T rTI R (τK ) − rTI R (τK−J )
The basic approach is to spread the vorticity of a far wake panel over an axial distance h, and then construct the
extended far wake equivalent to L spirals. Assuming that each of the L spirals has the same strength requires that
the circulation be periodic. Figure 23 shows the vortex elements of the extended far wake, corresponding to one
panel in the last revolution of the far wake model; and figure 24 describes the transformation from the far wake
model to the extended far wake model. The tip vortex elements are spread vertically to form a vortex sheet with
axial and circumferential components. There is a corresponding axial root vortex from the inboard trailed vorticity.
The shed vorticity is spread vertically to form a vortex sheet. These wake elements extend axially a distance Lh
beyond the last revolution of the far wake. The strength of the vorticity in the extended far wake is obtained from
the peak bound circulation:
Γk = Γmax (t − τk )
or from ΓR for the dual-peak model. The geometry is described by positions at the tip vortex (rT R ), and at the
edges of the inboard sheet (rSL and rSR ).
The geometry of the sheet element on the wake boundary is obtained from the position at the tip vortex,
2
last revolution of
far wake model
τk
3
τk +1
1
axial
root 1
vortex
3
inboard shed
vortex sheet
axial and
circumferential
vortex sheet on
wake boundary
2
4
4
2
Figure 25-23 Description of extended far wake panel (for axial flow).
wake boundary LΓk +1 LΓk 0 Γk +1+ Γk

-
2J
3 1 3 1
Δτ
Γk
h
Γk +1
4 2 4 2
0 0 0 Γk +1+ Γk
-
2J
root vortex Γk +1+ Γk

2
2J
Γk +1+ Γk
1
2J
shed wake LΓk LΓk
3 1
Γk
Γk +1
4 2
LΓk +1 LΓk +1
h = axial convection in one revolution
L = number of revolutions in extended far wake
J = number of wake age increments in one revolution
Figure 25-24 Development of extended far wake model.

rk = rT R (τk ):
1
r1 = rk − hk
2
1
r3 = rk+1 − h k
2
r2 = r1 − Lh k
r4 = r3 − Lh k
An approximation as a planar rectangle is sufficient. The induced velocity from the axial vorticity is given by the
trailed vorticity of this sheet element, with
1
Γ1 = Γ2 = − (Γk+1 + Γk )
2J
Γ3 = Γ4 = 0
The induced velocity from the circumferential vorticity is given by the shed vorticity of this sheet element, with
Γ1 = LΓk
Γ2 = 0
Γ3 = LΓk+1
Γ4 = 0
The tip vortex is spread axially over a distance h, giving a circumferential vortex sheet strength of γ = Γ/h. The
slope of the wake spiral gives an axial vortex sheet strength of γ(h/J)/w = Γ/Jw, where w is the width of the
sheet. The above convention for the circulation at the four corners produces sheet strengths of the required values.
Note that the axial sheet strength should be Γk /Jw on the 1–2 edge, and Γk+1 /Jw on the 3–4 edge (figure 24).
However, the sheet element available has the strength varying linearly along the length of the vorticity only (from
the 1–3 edge to the 2–4 edge in this case). Hence a constant strength of (Γk+1 + Γk )/2Jw is used instead.
The geometry of the sheet element for the shed vorticity is obtained from the average position at the edges of
the inboard sheet:
1
rL = rSL (τk+1 ) + rSL (τk )
2
1
rR = rSR (τk+1 ) + rSR (τk )
2
so
1
r1 = r R − h k
2
1
r3 = r L − h k
2
r2 = r1 − Lh k
r4 = r3 − Lh k
An approximation as a planar rectangle is sufficient. The induced velocity is given by the shed vorticity of this
sheet element, with
Γ1 = Γ3 = LΓk
Γ2 = Γ4 = LΓk+1
The shed vorticity of total strength (Γk − Γk+1 ) is spread axially over a distance h. Note that a roughly horizontal
shed vortex sheet of strength (Γk − Γk+1 )/w (where w is the width of the sheet) is being approximated by a
vertical sheet of strength (Γk − Γk+1 )/h (figure 24). The above convention for the circulation at the four corners
produces sheet strength of the required value.
The inboard trailed vorticity is in an axial root vortex. The geometry of the line segment is obtained from the
average position at the edge of the inboard sheet:
1
r2 = rL − h k
2
r1 = r2 − Lh k
The induced velocity is given by this line with

1
Γ1 = Γ2 = (Γk+1 + Γk )
2J
to match the strength of the axial vorticity on the wake boundary.
25-4.3 Vortex Line Segment
The velocity produced by a straight, finite-length vortex line segment is required. Figure 25 shows the
configuration. The line segment extends from point 1 to point 2 in space, and the velocity is required at point P.
The vortex strength varies linearly from Γ1 at point 1 to Γ2 at point 2. The convention for positive circulation is
about the vector from point 1 to point 2. The geometry is defined by the position vectors r1 and r2 , from the ends
of the line segment to P. The geometry is actually specified by vectors from a common origin to points 1, 2, and
P, from which r1 and r2 can be calculated. The position vectors are in arbitrary axes F, and hence the induced
velocity is obtained in axes F. The velocity induced by a vortex line segment is required in the form
The calculation requires a core radius rc , and specification of the modelling options. The Biot-Savart law gives
the induced velocity produced by this line segment:
$
1 Γ r × dσ
Δv = −
4π r3
where r is the vector from the element dσ on the segment, to the point P; and r = |r|. The coordinate σ is measured
along the vortex segment, from s1 to s2 :
1 1
s1 = (r1 · r2 − r12 ) = r1 · (r2 − r1 )
s s
1 1
s2 = (r22 − r1 · r2 ) = r2 · (r2 − r1 ) = s1 + s
s s
where s is the length of the segment:
s2 = |r1 − r2 |2 = r12 + r22 − 2r1 · r2
Write r = rm − σ1 , where rm is the minimum distance from the vortex line (including its extension beyond the
end points of the segment) to the point P, and 1
is the unit vector in the direction of the vortex:
1 1
rm = 2
r1 (r22 − r1 · r2 ) + r2 (r12 − r1 · r2 ) = (r1 s2 − r2 s1 )
s s
1
1
= (r1 − r2 )
s
r1
1
r Γ1
r2
dσ
2 Γ2
Figure 25-25 Vortex line segment.

The vectors rm and 1

are perpendicular, and
s = |r1 s2 − r2 s1 |2 = r12 r22 − (r1 · r2 )2

2 2
rm
The vortex strength varies linearly along the segment:

1 1
Γ= Γ1 (s2 − σ) + Γ2 (σ − s1 ) = (Γ1 s2 − Γ2 s1 ) + σ(Γ2 − Γ1 ) = Γm + σΓs
s s
It follows that
$
r1 × r2 s2
Γm + σΓs
Δv = 2 3/2
dσ
4πs s1 (rm + σ2 )
% &σ=s
r1 × r2 Γm σ/rm 2
− Γs 2
= 2 + σ )1/2
4πs (rm 2 σ=s1
% &
r1 × r2 s2 s1 1 1
= 2
Γm − − Γs r m
2
−
4πsrm r2 r1 r2 r1
% 0 8 0 8&
r1 × r2 s2 s2 s1 2
rm 1 1 s1 s2 s1 2
rm 1 1
= 2
Γ1 − + − − Γ2 − + −
4πsrm s r2 r1 s r2 r1 s r2 r1 s r2 r1
= Γ1 Δv1 + Γ2 Δv2
is the induced velocity of the vortex line segment with linearly varying circulation. Consider also a line segment
with a stepped distribution of the circulation. The distance from the midpoint of the segment to the point P is
1
r3 = (r1 + r2 )
2
and the midpoint is located at
1 1 1
s3 = (s1 + s2 ) = 2 (r22 − r12 ) = s1 + s
2 2s 2
The line segment has constant strength Γ1 from s1 to s3 , and constant strength Γ2 from s3 to s2 . Applying the
above result (with Γs = 0) to both pieces of the line segment gives

r1 × r3 s3 s1 r3 × r2 s2 s3
Δv = 2
Γ1 − + 2
Γ2 −
4π(s/2)rm r3 r1 4π(s/2)rm r2 r3
% 0 8 0 8&
r1 × r2 s3 s1 s2 s3
= 2
Γ1 − + Γ2 −
4πsrm r3 r1 r2 r3
= Γ1 Δv1 + Γ2 Δv2
for the induced velocity of the vortex line segment with stepped variation of circulation. The influence of the
vortex core is accounted for by multiplying the induced velocity of the line segment by the circulation factor f :
2
Rankine core: f = min(rm /rc2 , 1)
2 2
Scully core: f = rm /(rm + rc2 )
2

Bagai-Leishman core: f = rm / rm 4 + r4
c
2 2n
power-law core: f = rm /(rm + rc2n )1/n
2 2

Oseen core: f = 1 − e−arm /rc
The core radius rc is the location of the maximum tangential velocity. Thus the vortex core is accounted for by
2 2
using the factor f /rm instead of 1/rm in the expression for Δv.
25-4.4 Vortex Sheet Element
The velocity produced by a nonplanar, quadrilateral vortex sheet element is required. Figure 26 shows the
configuration. The four corners of the sheet element are at points 1 to 4 in space, and the velocity is required at
point P. The strength of the sheet vorticity is defined in terms of the wing bound circulation associated with the
four corners (Γ1 , Γ2 , Γ3 , and Γ4 ). The geometry is defined by the position vectors r1 , r2 , r3 , and r4 , from the
corners to P. The geometry is actually specified by vectors from a common origin to points 1, 2, 3, 4, and P, from
which the vectors from the corners to P can be calculated. The position vectors are in arbitrary axes F, and hence
the induced velocity is obtained in axes F. The velocity induced by a vortex sheet element is required in the form
Δv = Γ1 Δvt1 + Γ2 Δvt2 + Γ3 Δvt3 + Γ4 Δvt4

= (Γ1 − Γ3 )Δvt1 + (Γ2 − Γ4 )Δvt2
Δv = Γ1 Δvs1 + Γ2 Δvs2 + Γ3 Δvs3 + Γ4 Δvs4

= (Γ1 − Γ2 )Δvs1 + (Γ3 − Γ4 )Δvs3
for the shed vorticity. The calculation requires a sheet thickness dvs or core radius rc , and specification of the
modelling options.
For the general case of a nonplanar-quadrilateral element, the induced velocity is evaluated by numerical
integration. For the case of a planar-rectangular element the integrations can be performed analytically. The
planar-rectangular element can be used as an approximation for the general element. The edges of adjacent
elements will not join with this approximation, but the amount of computation required is reduced. As a further
approximation, the sheet can be replaced by trailed and shed line segments, with large core-radii to avoid high
induced velocities near the lines.
The sheet geometry is defined by the vectors s and t joining the midpoints of the sides (figure 26). These
vectors always intersect; the vector from the intersection to P is r0 . In terms of the vectors from the four corners
to P:
1
r0 = (r1 + r2 + r3 + r4 )
4
1
t = (r2 + r4 − r1 − r3 )
2
1
s = (r3 + r4 − r1 − r2 )
2
1
u = (r2 + r3 − r1 − r4 )
2
where u is the vector between the midpoints of the diagonals. The sheet surface is described by the coordinates σ
and τ , each varying from − 1/2 to 1/2; the origin σ = τ = 0 is at r0 . Then the vector from (σ, τ ) on the sheet to P
is
r = r0 − σs − τ et
= r0 − σes − τ t
= r0 − σes − τ et + 2στ u
r0
Γ3 Γ1
3 1 τ = 1/2
t
s
τ=0
Γ4 Γ2
4 2 τ = -1/2
σ = -1/2 σ=0 σ = 1/2
Γ1 - Γ3
2
3 1
large-core line
Γ4 - Γ3 Γ2 - Γ1 segment
approximation
1 2 for vortex sheet
element
4 2
1
Γ2 - Γ4
Figure 25-26 Vortex sheet element.

where the vectors in the trailed and shed directions are

et = t + 2σu
es = s + 2τ u
The bound circulation corresponding to the four corners is known, but not the actual vorticity distribution over the
sheet. A linear variation of the trailed vorticity δ(τ ) and shed vorticity γ(σ) is assumed:
1
δ=− (Γt + 2τ Γu )
|es |
1
γ=− (Γs + 2σΓu )
|et |
where
1
Γs = (Γ2 + Γ4 − Γ1 − Γ3 )
2
1
Γt = (Γ3 + Γ4 − Γ1 − Γ2 )
2
1
Γu = (Γ2 + Γ3 − Γ1 − Γ4 )
2
and
1 1
Γt + 2τ Γu = −(Γ1 − Γ3 )( + τ ) − (Γ2 − Γ4 )( − τ )
2 2
1 1
Γs + 2σΓu = −(Γ1 − Γ2 )( + σ) − (Γ3 − Γ4 )( − σ)
2 2
So % &
es et
ω dA = γ +δ |es | dσ|et | dτ
|es | |et |

= (Γs + 2σΓu )es − (Γt + 2τ Γu )et dσ dτ
is the differential vorticity on the sheet. The Biot-Savart law gives the induced velocity produced by this sheet
element: $
1 r×ω
Δv = − dA
4π r3
$ 1/2 $ 1/2
1 dσ dτ
=− [(Γs + 2σΓu )(r × es ) − (Γt + 2τ Γu )(r × et )]
4π −1/2 −1/2 r3
and hence $$
1 1 dσ dτ
Δvt1 = −( + τ )(r × et ) 3
4π 2 r
$$
1 1 dσ dτ
Δvt2 = −( − τ )(r × et ) 3
4π 2 r
$$
1 1 dσ dτ
Δvs1 = ( + σ)(r × es ) 3
4π 2 r
$$
1 1 dσ dτ
Δvs3 = ( − σ)(r × es ) 3
4π 2 r
The velocity from an infinitesimally thin vortex sheet has a logarithmic singularity near the edges. To avoid this
singularity, r3 in the integrand denominator is replaced by
(r2 + d2vs )3/2
The parameter dvs may be considered a viscous core size (compare with the role of a core radius for a line segment).
The introduction of dvs also improves the convergence of the numerical integration when the collocation point is
close to the sheet surface.
For a planar-rectangular element, the velocity integrals can be evaluated analytically. The planar assumption
means u = 0; the rectangular assumption means s · t = 0. Hence define
ρ = r0 − σs − τ t
= rm − (σ − σm )s − (τ − τm )t
where the minimum distance between P and the plane of the sheet gives
1
σm = r0 · s
|s|2
1
τm = 2 r0 · t
|t|
rm = r0 − σm s − τm t
Note that rm · s = 0 and rm · t = 0. So
ρ2 = ρ2m + (σ − σm )2 s2 + (τ − τm )2 t2
ρ2m = rm
2
+ d2vs
ρ × s = rm × s − (τ − τm )t × s
ρ × t = rm × t − (σ − σm )s × t
Then the induced velocity can be written as the planar-rectangular result plus a correction term:
$$
1 1 r × et ρ×t
Δvt1 = (Δvt1 )PR + −( + τ ) − dσ dτ
4π 2 r3 ρ3
and similarly for Δvt2 , Δvs1 , and Δvs3 . If the element is actually a planar rectangle, then the correction term is
zero (r × et /r3 = ρ × t/ρ3 ). If the element is a nonplanar quadrilateral, the expressions given for ρ, (ρ × s),
and (ρ × t) are still used, and the correction term is evaluated by numerical integration. As a approximation,
the correction term can be neglected. The integrated expressions for the velocity induced by a planar-rectangular
element are: $$
1 1 dσ dτ
(Δvt1,2 )PR = −( ± τ )(ρ × t)
4π 2 ρ3
% & % &
rm × t 1 s×t 1
=− ( ± τm )I1 ± I2 + ( ± τm )I3 ± I4
4π 2 4π 2
$$
1 1 dσ dτ
(Δvs1,3 )PR = ( ± σ)(ρ × s)
4π 2 ρ3
% & % &
rm × s 1 t×s 1
= ( ± σm )I1 ± I3 − ( ± σm )I2 ± I4
4π 2 4π 2
where $$
dσ dτ 1 (σ − σm )(τ − τm )st
I1 = = tan−1
ρ3 ρm st ρm ρ
$$
(τ − τm ) dσ dτ 1
I2 = 3
= 2 ln (ρ − s(σ − σm ))
ρ st
$$
(σ − σm ) dσ dτ 1
I3 = 3
= 2 ln (ρ − t(τ − τm ))
ρ s t
$$
(σ − σm )(τ − τm ) dσ dτ ρ
I4 = =− 2 2
ρ3 s t
each evaluated for σ = − 1/2 to 1/2 and τ = − 1/2 to 1/2. For the general case of a nonplanar quadrilateral, the
correction terms of the form $ 1/2 $ 1/2
I= f (σ, τ ) dσ dτ
−1/2 −1/2
are evaluated numerically. Using trapezoidal integration with M equally spaced intervals in σ and τ gives
M M
1
I= f (eM M M M
i , ek ) wi wk
M2 i=0 k=0
where
i 1
eM
i = −
M 2
'
wiM = 1/2 i = 0 or i = M
1 otherwise
Now
2M M M
f (e2M 2M
i ) wi = f (eM M
i ) wi + f (oM
i )
i=0 i=0 i=1
where
2i − 1 1
oM
i = −
2M 2
Then doubling the number of integration steps gives
2M 2M
1
I2M = f (e2M 2M 2M 2M
i , ek ) wi wk
4M 2 i=0 k=0
2M
" M M
#
1
= f (e2M M M
i , ek ) wk + f (e2M M
i , ok ) wi2M
4M 2 i=0 k=0 k=1
" M M M M
1
= f (eM M M M
i , ek ) wi wk + f (oM M M
i , ek ) wk
4M 2 i=0 k=0 i=1 k=0
M M M M
#
+ f (eM M M
i , ok ) wi + f (oM M
i , ok )
i=0 k=1 i=1 k=1
1
= IM + JM
4
M
" M
#
1 M
JM = f (oM M
i , ek ) + f (eM M
k , oi ) wkM + f (oM M
i , ok )
4M 2 i=1
k=0 k=1
The process is started with M = 1, and repeated until the velocity converges:
|I2M − IM |2 < 2 |ΔvPR |2
( is an input tolerance), or until the number of steps reaches a limit (such as M = 512). An unconverged result can
be very wrong, so it is important that be small enough. The need for a very small value of , and the possibility
of reaching the limit on M , can be controlled to some extent by the choice of dvs .
An economical approximation is to replace the vortex sheet by a line segments, with either a linear or a
stepped circulation distribution, and a large core size to eliminate the high induced velocity near the lines. The
strength and position of the line segments are determined from the circulation and position of the four corners of
the sheet:
Γ2line = Γ1 − Γ3
1
r2line = (r1 + r3 )
2
1
r1line = (r2 + r4 )
2
1
r2line = (r1 + r2 )
2
1
r1line = (r3 + r4 )
2
for the shed vorticity (figure 26). Hence the line segment velocity increments give for the trailed sheet vorticity
Δvt1 = Δv2line
Δvt2 = Δv1line
and for the shed sheet vorticity

Δvs1 = −Δv2line
Δvs3 = −Δv1line
The core radius can be specified arbitrarily; or rc = |s|/2 can be used for the trailed vorticity, and rc = |t|/2 for
the shed vorticity.
25-4.5 Circular-Arc Vortex Segment
The velocity produced by a circular-arc vortex line segment is required by the wake geometry components
(although not by this wing wake component). A piecewise-straight model of a vortex line is usually satisfactory, but
to calculate the self-induced velocity of the line its curvature must be included. Figure 27 shows the configuration.
The line segment extends from point 1 to point 2 in space, and the velocity is required at point P. Here it is assumed
that P is located at one end of the segment, hence the velocity is required at point 1 or at point 2. In order to define
a curved line, a third point is required, either point 0 (before point 1) or point 3 (after point 2). A circular arc of
radius R is defined by points (0,1,2) or (1,2,3); and the vortex line segment is that portion of the arc extending
from point 1 to point 2. The vortex strength varies linearly from Γ1 at point 1 to Γ2 at point 2. The convention for
positive circulation is about the vector from point 1 to point 2. The geometry is defined by the position vectors r1
and r2 from a common origin to the ends of the line segment; plus the vector r0 or r3 to a third point. The position
vectors are in arbitrary axes F, and hence the induced velocity is obtained in axes F. The velocity induced by a
vortex line segment is required in the form
The calculation requires a core radius rc , and specification of the modelling options. A non-zero core radius is
needed in order to obtain a finite induced velocity on the line segment; and this velocity depends on the distribution
of tangential and axial velocities in the viscous core. Bliss (ref. 6) shows that the correct induced velocity is
3
Γ2
2
R
Γ1
1
Δθ
0
n
y
2
3
P at 1:
2 dσ
R
a θ
x
1 1
c
b
2
d
x
0 θ
P at 2: R
dσ
−y 1
Figure 25-27 Circular-arc vortex segment.

obtained from a line of concentrated vorticity (zero core radius) if the integration over the line stops a cutoff
distance d before the collocation point. Assuming that the ratio of the core radius to the arc radius is small, the
required cutoff distance is:
rc 1 −A+C
d= e2
2
The quantities A and C follow from the kinetic energy in the viscous core:
$ r
A = lim rv 2 dr − ln r
r→∞ 0
$ ∞
C= 2rw2 dr
0
Here r is the distance from the center of the core (scaled with the core radius rc ); and v and w are the swirl and
axial velocities (scaled with the circulation Γ/2πrc ). The axial velocity will be ignored, hence C = 0 here. For
the vorticity distributions considered here, the dimensionless velocity v and the integral A are as follows.
1
Rankine core: v = min(r, 1/r) A= 4
Scully core: v = r/(r2 + 1) A = − 12

√
Bagai-Leishman core: v = r/ r4 + 1 A=0
power-law core: v = r/(r2n + 1)1/n A∼

= 1 (n+3)(n−2)
4 (n+1)n
2

Oseen core: v = 1 − e−ar /r A= 1
2 (γ + ln a2 ) = 0.0562
(where γ ∼
= 0.5772 is Euler’s constant). Hence depending on the core model, the cutoff distance is:
⎧1
⎪ 2 e1/4 = 0.6420
⎪ Rankine
⎪
⎪
⎪
⎪
⎪
⎪
1
2 e = 1.3591 Scully
⎪
⎪
⎨
1 1/2
d/rc = 2 e = 0.8244 Bagai-Leishman
⎪
⎪
⎪
⎪ 1 e−A+1/2
⎪
⎪ power
⎪
⎪2
⎪
⎪
⎩ √1 e(1−γ)/2 = 0.7793 Oseen
2a
The Biot-Savart law gives the induced velocity produced by this line segment:
$
1 Γ r × dσ
Δv = −
4π r3
where r is the vector from the element dσ on the segment, to the point P; and r = |r|. The coordinate σ is
measured along the vortex segment. The circular arc has radius R and angular length Δθ (figure 27). The vector
n is normal the plane of the arc. Consider the case with the collocation point P at end point 1. Distance along the
vortex segment is defined by the angle θ, measured from point 1. Then
⎛ ⎞
1 − cos θ
r = R ⎝ − sin θ ⎠
0
⎛ ⎞
− sin θ
dσ = R dθ ⎝ cos θ ⎠
0
θ
r = 2R sin
2
θ
r × dσ = −nR2 dθ 2 sin2
2
The vortex strength along the segment is assumed to be:
sin θ/2 θ
Γ = Γ1 + (Γ2 − Γ1 ) = Γm + Γθ sin
sin Δθ/2 2
which is a linear variation for small Δθ (for analytical integration, a factor of sin θ/2 is needed instead of θ/2). It
follows (assuming small d/R) that
$ Δθ % &
n dθ dθ
Δv ∼
= Γm + Γθ
8πR d/R 2 sin θ/2 2
% &θ=Δθ
n θ θ
= Γm ln tan + Γθ
8πR 4 2 θ=d/R
% &
∼ n 4R Δθ θ
= Γm ln tan + Γθ
8πR d 4 2
% &
n 4R Δθ Δθ/2 Δθ/2
= Γ1 ln tan − + Γ2
8πR d 4 sin Δθ/2 sin Δθ/2
= Γ1 Δv1 + Γ2 Δv2
is the induced velocity of the circular-arc vortex segment with linearly varying circulation (P at 1). Note that the
self-induced velocity of a vortex ring is given by twice the above result for Δθ = π:
nΓ 4R
Δv = ln
4πR d
Consider also a line segment with a stepped distribution of the circulation: constant strength Γ1 from θ = 0 to
Δθ/2, and constant strength Γ2 from θ = Δθ/2 to Δθ. Thus
$ Δθ
n dθ
Δv ∼
= Γ
8πR d/R 2 sin θ/2
" θ=Δθ/2 θ=Δθ #
n θ θ
= Γ1 ln tan + Γ2 ln tan
8πR 4 θ=d/R 4 θ=Δθ/2
% &
∼ n 4R Δθ tan Δθ/4
= Γ1 ln tan + Γ2 ln
8πR d 8 tan Δθ/8
% &
n 4R Δθ tan Δθ/4 tan Δθ/4
= Γ1 ln tan − ln + Γ2 ln
8πR d 4 tan Δθ/8 tan Δθ/8
= Γ1 Δv1 + Γ2 Δv2
is the induced velocity of the circular-arc vortex segment with stepped variation of circulation (P at 1). Next
consider the case with the collocation point P at end point 2. Distance along the vortex segment is defined by the
angle θ, measured from point 2 (figure 27). Then
⎛ ⎞
1 − cos θ
r = R ⎝ sin θ ⎠
0
⎛ ⎞
sin θ
dσ = R dθ ⎝ cos θ ⎠
0
and the expressions for r and r × dσ are unchanged. The vortex strength along the segment is assumed to be:
sin θ/2 θ
Γ = Γ2 + (Γ1 − Γ2 ) = Γm + Γθ sin
sin Δθ/2 2
Thus it is only necessary to switch Γ1 and Γ2 in the expressions for the induced velocity:
% &
n 4R Δθ Δθ/2 Δθ/2
Δv = Γ2 ln tan − + Γ1
8πR d 4 sin Δθ/2 sin Δθ/2
is the induced velocity of the circular-arc vortex segment with linearly varying circulation (P at 2); and
% &
n 4R Δθ tan Δθ/4 tan Δθ/4
Δv = Γ2 ln tan − ln + Γ1 ln
8πR d 4 tan Δθ/8 tan Δθ/8
is the induced velocity of the circular-arc vortex segment with stepped variation of circulation (P at 2). To complete
the evaluation of the induced velocity, the quantities n, R, and Δθ must be obtained from the geometry of the
circular arc. Consider the arc defined by points (0,1,2). The chord vectors a and b are
a = r2 − r1
b = r1 − r0
(figure 27). The normal vector is

b×a
n=
|b × a|
where
|b × a|2 = a2 b2 − (a · b)2
The induced velocity will be set to zero if a is parallel to b (|b × a| = 0); or if the length of a, b, or a + b is
zero. Next the vectors c and d are constructed. The vector c is perpendicular to both a and n; the vector d is
perpendicular to both b and n. Hence
c= a2 b − (a · b)a
c

d = d b2 a − (a · b)b
The scalar factors c and d are obtained from the requirement that c and d meet at the center of the arc:
a b
+c=− +d
2 2
The a and b components of this equation give two relations
a2 (a · b)
=− + d a2 b2 − (a · b)2
2 2
(a · b) b2
+ c a2 b2 − (a · b)2 = −
2 2
which can be solved for c and d:
1 (a · b) + b2 1 b · (a + b)
c =− =−
2 a2 b2 − (a · b)2 2 |b × a|2
1 a2 + (a · b) 1 a · (a + b)
d = =
2 a2 b2 − (a · b)2 2 |b × a|2
The arc radius R is obtained from a b

R = + c = − + d
2
2 2
a2 b2
= + c2 = + d2
4 4
a2 b2 a2 + 2(a · b) + b2
=
4 a2 b2 − (a · b)2
a2 b2 |a + b|2
=
4 |b × a|2
Hence
ab |a + b|
R=
2 |b × a|
n b×a
=
2R ab |a + b|
The angular length of the arc Δθ is
Δθ a |b × a|
sin = =
2 2R b |a + b|
so 4
Δθ a2 |b · (a + b)|
cos = 1− =
2 4R2 b |a + b|
Δθ sin Δθ/2 |b × a|
tan = =
4 1 + cos Δθ/2 b |a + b| + |b · (a + b)|
Δθ ab |a + b|
2R tan =
4 b |a + b| + |b · (a + b)|
and 4
tan Δθ/4 1 + cos Δθ/4 Δθ
= =1+ 1 + tan2
tan Δθ/8 cos Δθ/4 4
Finally consider the arc defined by points (1,2,3). Now the chord vectors a and b are
a = r3 − r2
b = r2 − r1
The expressions for n, c, d, and R are unchanged. The angular length of the arc Δθ is
Δθ b |b × a|
sin = =
2 2R a |a + b|
so a and b are interchanged in cos Δθ/2, tan Δθ/4, and 2R tan Δθ/4.
25-4.6 Ground Effect
The influence of the ground can be included in the wake-induced velocity calculation, through the use of
image elements in the wake model. Wake geometry components can also include the ground plane influence.
The boundary condition imposed by the ground is zero normal wake-induced velocity at the ground plane. This
constraint is satisfied by introducing an image element for every vortex element in the wake model. The image
element is created from the vortex element by changing the sign of the strength, and reflecting the position across
the ground plane. This reflection means changing the sign of the normal component of the position relative the
ground. So for a point r on a vortex element, the image point is at
rimage = r + 2hkg
where h = hOAGL − rT kg is the distance of r above the ground plane; hOAGL is the distance above ground level
of the origin of the frame in which r is measured; and kg is the downward normal vector of the ground plane. The
inertial frame I is used for the wake analysis.
A wake element can be too near or below the ground plane. This can occur with rigid or prescribed geometry
that does not include the effect of the ground; or because of numerical effects in the free geometry calculation.
Here “near” the ground plane is defined by a distance δ above ground level (δ can be zero or negative). Optionally,
the analysis neglects any part of a vortex element (and the corresponding part of the image) that is below h = δ.
Consider a vortex line segment extending from r1 to r2 . The height above ground level is h1 and h2 at the ends
of the segment. If r2 is below the ground and r1 above (h2 < δ and h1 ≥ δ), then r2 is replaced by
h1 − δ δ − h2
r2new = r2 + r1
Δh Δh
where Δh = h1 − h2 . This new point is on the original line (r2new = σr2 + (1 − σ)r1 ), and gives h2new = δ.
Similarly, if r1 is below the ground and r2 above (h1 < δ and h2 ≥ δ), then r1 is replaced by
h2 − δ δ − h1
r1new = r1 + r2
Δh Δh
where Δh = h2 −h1 . If both ends are above the ground, no replacement is made. If both ends are below the ground,
the velocity from the vortex line segment and its image is zero. With this option, the free wake geometry behaves
reasonably near the ground plane. The above replacement is made only for vortex line segments. Consequently,
an image element in ground effect is only considered for line segments, not for sheet elements.
For communication with an external aeroacoustic analysis, the induced velocity can be calculated excluding
vortex elements inside the computational domain of the external analysis. The following vortex elements can be
tested for exclusion:
a) The wake behind the wing of the collocation point.

b) The wake behind the wing of the collocation point, plus tip vortices from all wings in the
set.
c) All wake elements.
The wake behind the wing of the collocation point is considered only if the wake age at the panel leading edge
is less than a specified value, kle ≤ kbox (to allow for the case of a wing flying over its own wake). The bound
vortices and the elements in the extended far wake are assumed to be outside the computational domain, hence
are never excluded. Only vortex line segments are considered, so vortex sheet elements can not be used for any
inboard wake panels inside the computational domain.
The computational domain of the external analysis is here approximated by a box, defined by six planes. The
orthogonal box axes are specified by three vectors, in one of the following forms:
a) Two unit vectors, plus the origin: u1 , u2 , and r0 ; typically obtained from a reference frame
component.
b) Two points on axes, plus the origin: r1 , r2 , and r0 ; typically obtained from a structural
dynamic component. The corresponding unit vectors are then proportional to u1 = r1 − r0
and u2 = r2 − r0 .
These three vectors must be defined using the same axes and origin as for the wake geometry (the inertial frame
here). From this input, orthogonal unit vectors of the computational domain are calculated as follows:
u1
1 =
|u1 |
12
u
2 = 12 = (I − 1 T1 )u2
u
|1
u2 |
3 =
1 2
Each of the two axes specified is identified as the positive or negative x, y, or z-axis. Hence the signs of m are
changed as required, and each m is associated with one of the three axes. The box is generated by two planes
normal to each axis, specified by the intercepts of the planes with the x, y, and z axes. From this information, each
plane can be described by an outward normal ni , and one point on the plane oi (i = 1 to 6; m = 1 to 3):
normal ni point oi
m r0 + m dmax
−m r0 + m dmin
where dmin and dmax are the minimum and maximum intercepts of the axis corresponding to m . The i-th plane
is the locus of points r such that
(r − o)T n = 0
Then an arbitrary point r is outside the plane if
(r − o)T n > 0
and inside the plane if

(r − o)T n < 0
The box interior is the region of points r that are inside all six planes. The wake-induced velocity is calculated
as the sum of the velocities produced by straight, finite-length line segments. A line segment extends from r1 to
r2 . In order to obtain the induced velocity excluding wake elements inside the computational domain, each line
segment is tested as follows. If both r1 and r2 are inside the box, then the entire line segment is excluded (the
velocity increment is zero). If r1 or r2 is outside the box, then the intersection of the line with each of the six
planes is examined. The line segment or its extension intersects a plane at
rI = r1 (1 − σ) + r2 σ
where
(o − r1 )T n
σ=
(r2 − r1 )T n
(which gives (rI − o)T n = 0). If (r2 − r1 )T n = 0, then the line and plane are parallel, and there is no intersection.
The intersection lies within the line segment if 0 ≤ σ ≤ 1; and on the boundary of the computational domain if rI
is on or inside each of the other five planes. If so, then rI is identified as the intersection s1 or s2 (if respectively r1
or r2 is outside this plane). These intersections identify portions of the line segment inside the box, which are to
be excluded. If there are no such intersections, then the entire line segment is outside the box, and is included. If
there is one such intersection, then the velocity of the segment portion from r1 to s1 , or from s2 to r2 , is included.
If there are two such intersections, then the velocity of the segment portions from r1 to s1 , and from s2 to r2 , are
both included.
25–5 References
1) Scully, M.P. “Computation of Helicopter Rotor Wake Geometry and its Influence on Rotor Harmonic Airloads.”
Massachusetts Institute of Technology, ASRL TR 178-1, March 1975.
2) Vatistas, G.H.; Kozel, V.; and Mih, W.C. “A Simpler Model for Concentrated Vortices.” Experiments in Fluids,
Volume 11, Number 1, April 1991.
3) Bagai, A., and Leishman, J.G. “Rotor Free-Wake Modeling Using a Pseudo-Implicit Technique — Including
Comparisons with Experimental Data.” Journal of the American Helicopter Society, Volume 40, Number 3, July
1995.
4) Panton, R.L. Incompressible Flow. John Wiley and Sons, New York, 1984.
5) Ham, Y.O.; Leishman, J.G.; and Coyne, A.J. “On the Turbulent Structure of a Tip Vortex Generated by a Rotor.”
American Helicopter Society Forum, June 1996.
6) Bliss, D.B.; Teske, M.E.; and Quackenbush, T.R. “A New Methodology for Free Wake Analysis Using Curved
Vortex Elements.” NASA CR 3958, December 1987.
wing wing wing wing

wake circulation circulation vorticity
geometry peaks strength
wing no yes (a) yes (a) X

vorticity
strength
collocation no no no yes
point
velocity
wing sensor no no yes (a) no
wing/coll pt yes (a) no yes (a) no

influence
coefficient
collocation normal wing/coll pt box axis

point influence position
location coefficient
wing no no no no
vorticity
strength
collocation no no yes (b) no

point
velocity
wing sensor no no no no
wing/coll pt yes (b) no (c) X no (d)

influence
coefficient
Notes:
a) yes only if wings of output and input are same
b) yes only if collocation points of output and input are same
c) yes if axial flow
d) yes if box exists
Figure 25-26 Functionality of wing wake component.

Chapter 26
WAKE GEOMETRY COMPONENTS
26–1 Description
All wake geometry components share common characteristics, reflected in the standard features of the im-
plementation of the components. The wake geometry components calculate the position of the wake vorticity in
space. The undistorted geometry is obtained from the motion of the wing: a wake element is convected by the
wind, from the position in the air at which it was created. This geometry is distorted by the self-induced velocity
of all the wakes in the system, as well as the influence of wings and bodies on the air velocity. In axes moving with
the wing, the wake geometry also reflects the past history of the wing position. The distorted wake geometry is
calculated for a set of wings. Specifying the set defines what mutual interaction between wakes is accounted for in
the wake geometry. If the set does not consist of all wakes in the system, then some interaction is being neglected.
The wake geometry model can be rigid, prescribed, or free. A rigid model calculates the wake geometry distortion
from the mean induced and interference velocity at the wing. A prescribed wake geometry is obtained from an
empirical model, based on measurements. A free wake geometry is obtained by calculation. The wake geometry
model may depend on the loop solution level.
In summary, the following wake geometry models can be used: rigid, free, and trim distortion. The models
available depend on the solution task:
a) Trim: zero, rigid, free.

b) Transient: zero, rigid, free, trim distortion.
c) Flutter: not applicable.
transient task, the component calculates the geometry from the transient undistorted geometry, plus the saved trim
distortion. For the “zero” option, no output is calculated.
26-2.1 Character of Wake Geometry
The wake geometry structure is typically dominated by the tip vortices, consisting of the rolled up trailed
vorticity arising from the tips of the wings. The basic geometry is created by the motion of the wing relative to the
air. For a wing moving at a fixed speed without rotation, the basic geometry consists of straight lines extending
behind the wing. For a rotor, the basic geometry consists of a helix below each blade (convected downward by the
mean induced velocity), skewed aft in edgewise flight.
380 WAKE GEOMETRY COMPONENTS
As at the wings, the induced velocity throughout the flow field is highly nonuniform. Thus the actual position
of the wake element, determined by the integral of the local convection velocity, is highly distorted from the
basic geometry. In some systems the wake geometry distortion is not important, or can be adequately represented
by just considering convection by the mean induced velocity. The wake geometry is important when a wing or
body (perhaps the wing that generated the wake) passes close to the wake. The wake-induced loading in such
configurations is very sensitive to the separation between the wings and tip vortices. Hence it is necessary to
consider the self-induced distortion of the wake, particularly the tip vortices.
Calculating the free wake geometry requires evaluating the induced velocity at node points in the wake.
The models and methods used are similar to those for calculating the induced velocity at the wing set, but here
the velocity is needed for points throughout the flow field, hence at an order of magnitude more points than the
collocation points on the wings. Much more attention to efficiency is therefore required for the free wake problem.
26-2.2 Lifting Line Wing
For a lifting-line theory, the wake geometry is based on the motion of the bound vortex (lifting line) through
the air. The lifting line is assumed to be at the quarter chord of the wing. For constructing the wake geometry, the
shift to the geometric three-quarter chord in reverse flow is ignored.
The geometry of the collocation points involves placing them at the three-quarter-chord in the local flow
direction, in order to implement second-order lifting-line theory. Note that the collocation points are part of the
geometry of the wake rather than the wing (part of the outer problem not the inner problem). The wake geometry
is obtained from the position of the lifting-line at the current and past time steps. Thus the wake geometry can
be used to get the local flow direction for the collocation points. A consequence of this approach is that the
collocation points are automatically kept away from trailed line segments of the near wake. In cases of highly
distorted geometry, especially near a reverse flow boundary, some close encounters between the collocation points
and wake vortex lines can still occur. The calculation can use a core size for the near wake line segments in order
to avoid singularities, but this core size has no physical significance, and indeed its value must not influence the
solution for the loading.
Optionally a straight lifting-line can be used to generate the wake geometry, ignoring curvature of the actual
quarter chord. First order lifting-line theory (quarter-chord collocation point) should use a straight lifting-line
for consistency, since sweep effects are higher order. In this case, the wing geometry is generated by a straight
line from left tip to right tip, interpolated to the aerodynamic collocation points as required. Often the wing tip
locations are on an appropriate reference line, not the actual tips. This approximation has nothing to do with the
aerodynamic interfaces (velocity and force) between the structure and the aerodynamic components.
26-2.3 Description of Geometry
In general, the inertial frame I is the only common frame for the wake geometry and all collocation points.
Thus the wake geometry is calculated as the position relative the origin of I, in I axes. The initial position is where
the wake element is created, such as the wing trailing edge or lifting line. Then the wake element is convected by
the wind, and by the wake self-induced velocity. Usually the wing is itself moving relative to the inertial axes. The
wake geometry may be constant (or periodic) in a frame moving with the wing, even for cases involving unsteady
loading. However, if the wing is moving relative to the inertial frame, then the wake geometry in I axes is not
constant (or periodic). If only the relative positions of the collocation points and wake geometry are important,
WAKE GEOMETRY COMPONENTS 381
then it is still possible to use the wake geometry at just a single time. In general, the wake geometry must be
calculated over the full range of time required by the trim and transient solution procedures.
I
The wake geometry is generated by points along the wing span: rQ as a function of time t. The points
required are the locations of the wing tips and the aerodynamic collocation points (panel midpoints):
6r left wing tip
TL
rQ = r T R right wing tip
rAi collocation points, i = 1 to M
Locations at the wing (trailing edge or lifting line) are calculated using the standard aerodynamic interface of
structural dynamic components, evaluated at the current time and past time as required. The wake consists of
rolled up tip vortices and an inboard sheet of the remaining vorticity, extending behind the wing. The tip vortices
and inboard sheet have distinct geometries, requiring separate description. The tip vortices are described by lines
arising from the left and right wing tips. The inboard sheet is a surface, described by a set of spanwise points
starting at the wing tips (the sheet edges) and at the span stations of the collocation points. The tip vortex can form
on the generating wing at a span station rT V inboard of the wing tip. The effect of rT V on the loading and near
wake is implemented in the wing component. In the wing components, rT V is an input parameter. Separately, the
effect of rT V on the far wake can be included in the geometry of the tip vortices and sheet edges, by assuming they
are described by lines arising from span stations rT V rather than from the tips. In the wake geometry components,
rT V is an input parameter, or is calculated assuming that the centroid of the trailed vorticity is conserved (Betz
rollup). The point rQ at span station rT V is obtained by interpolation from the input rQ at the wing tips and the
aerodynamic collocation points.
At time t, a wake element and its position in the wake is identified by the wake age τ , such that (t − τ ) is the
time that the element was created. The age goes from τ = 0 at the current wing position, to τ = ∞ far behind
the wing. The wake geometry is discretized by calculating it only at a set of wake ages τk = k Δτ , k = 0 to K,
for fixed wake age increment Δτ . The extent of the calculated geometry K can be different for the various span
stations, as required by the wake model. The wake geometry at time t is described by the position of a set of nodes
I
at age τk : rW k as a function of time. The points required are the locations of the tip vortices, the inboard sheet
edges, and the aerodynamic collocation points (panel midpoints plus wing tips):
⎧
⎪
⎪
⎪
⎪
⎪
⎩ r SRk right edge inboard sheet
for k = 0 to K. There is a set of (rT Lk , rT Rk , rSLk , rSRk ) values for each panel of the far wake trailed vorticity.
The tip vortices and inboard sheet edges are generated by the wing tip locations (optionally including the effect of
the rollup location rT V ). If the wake is not rolled up at a wing tip or panel boundary, then rT is not used (rT will
be equal to rS ). At the boundary between adjacent panels, rT should be the same if the wake is rolled up there,
while rS should be the same if the wake is not rolled up. The wake geometry at the wing span stations is used for
the near wake region (hence does not include the effect of rT V ). The vector of the wake geometry (component
input and output) consists of all these positions. The vector is calculated as a function of time, as defined by the
solution procedure.
The wake geometry vector is calculated at a set of time values t , as determined by the solution procedure.
The wake geometry vector describes the position at a set of age values τk . When the wake geometry is used, the
following access methods are possible:
a) Direct access: rW used only at times t identical to the calculated values t .

b) Constant: rW at the first t used for all time t.
c) Evaluate at time: rW evaluated at the required time t from the response calculated at times
t (standard linear interpolation in time, for fixed wake age).
d) Interpolate: rW interpolated to the required time t for fixed wake element (rather than fixed
wake age).
The solution procedure can be defined so direct access of the wake geometry in the subvector is proper. Direct
access is most efficient, but requires coordination of the solution procedures for the wake geometry component
and for the components using the wake geometry. Specifically, the time and time step used to evaluate the wake
geometry as component input must be the same as the time and time step used to calculate it as component output (to
an input/output interface). With direct access, the response for rW must not have a nominal or reference. Note that
if the wake geometry is constant in a frame moving with the wing, and only the relative positions of the collocation
points and wake geometry are important, then the constant option can be used even if the geometry is not actually
constant in inertial axes. Evaluating the wake geometry at an arbitrary time is the standard way for components
to communicate, but is an approximation (unless it is actually equivalent to direct access). Interpolating the wake
geometry for fixed wake element is least efficient, but is necessary if the wake geometry is to be used at times that
are independent of the times at which it was calculated.
The standard procedures for evaluating the system response interpolate the wake geometry vector in time,
hence for fixed wake age. Wake geometry interpolation should however be accomplished by following the path
of a specific wake element in time, hence by considering the geometry for a fixed value of the time that a wake
element was created, δ = t − τ . Such interpolation proceeds as follows:
rW (t , τk−1 ) and rW (t , τk ) interpolated to rW (t , τk − t + t )

rW (t +1 , τk ) and rW (t +1 , τk+1 ) interpolated to rW (t +1 , τk −t+t +1 )
then rW (t , τk − t + t ) and rW (t +1 , τk −t+t +1 ) interpolated to rW (t, τk )
(assuming that time t and age τk have the same increment Δt). The standard procedures for evaluating the
response interpolate in time t:
rW (t , τk ) and rW (t +1 , τk ) interpolated to rW (t, τk )
The two results are the same if the four points involved are coplanar, hence if adjacent nodes of the wake geometry
are convected with the same mean velocity direction. Specifically, consider interpolating r(t , τk ) to arbitrary t.
Find such that t ≤ t ≤ t +1 ; and calculate the weights w and w +1 for linear interpolation of the geometry in
time. The weights wk and wk+1 interpolate in age to τk . Note that 0 ≤ t − t ≤ Δt, so τk−1 ≤ (τk − t + t ) ≤ τk ;
and τk − (τk − t + t ) = t − t so wk = w +1 . The interpolated geometry is:
r(t, τk ) = w r(t , τk − t + t ) + w +1 r(t +1 , τk − t + t +1 )

= w wk r(t , τk−1 ) + wk+1 r(t , τk ) + w +1 wk r(t +1 , τk ) + wk+1 r(t +1 , τk+1 )
= w r(t , τk ) + w +1 r(t +1 , τk )

+ w w +1 r(t +1 , τk+1 ) − r(t +1 , τk ) + r(t , τk−1 ) − r(t , τk )
using w +1= 1 − w . Interpolating the wake geometry in time for fixed τk instead is equivalent to neglecting the
second line in the last result. That term is zero however if the four points involved lie in a plane.
For the option to interpolate the wake geometry for fixed wake element, the more general case in which the time
and age have difference increments must be considered. The wake geometry is evaluated at times t = t0 + Δt,
with the wake age values τk = k Δτ . First , m, and n are found, such that
t ≤t≤t +1
τm−1 ≤ τk − t + t ≤ tm
τn ≤ τk − t + t +1 ≤ tn+1
Note that if the time and age have the same increment, Δt = Δτ , then m = n = k. Thus
% &
t − t0
=
Δt
t − t0
w +1 = − =1−w
Δt
% &
τk − t + t
m−1=
Δτ
τk − t + t
wm = − (m − 1) = 1 − wm−1
Δτ
% &
τk − t + t +1
n=
Δτ
τk − t + t +1
wn+1 = − n = 1 − wn
Δτ
(brackets indicating integer truncation), and the interpolated value of the wake geometry is:

r(t, τk ) = w wm−1 r(t , τm−1 ) + wm r(t , τm ) + w +1 wn r(t +1 , τn ) + wn+1 r(t +1 , τn+1 )
For τk = 0, or if any τ value is out of range, then
r(t, τk ) = w r(t , τk ) + w +1 r(t +1 , τk )
is used to calculate the wake geometry instead.
These special procedures to interpolate the wake geometry in time could be avoided by defining the vector
rW k to consist instead of the positions of points on the wake, so k identifies a particular node at fixed δ = t − τ
rather than the age τ . However, the position of a given point would not exist for t < δ, with δ varying within the
vector. Thus it would still not be possible to rely on the standard procedures for evaluating the response at arbitrary
time.
26–3 Theory
26-3.1 Description of Geometry
The undistorted geometry is calculated from the position in the air at which the wake element was created,
plus convection by the wind. Then the distortion produced by the self-induced velocities of the wing set is added.
Let t be the current time, and τ the age of the element in the wake. Thus δ = t − τ is the time when the vorticity
was created. Then
I
rW I
(t, τ ) = rQ (t − τ ) + τ vW
I
+ DI (t, τ )
is the wake geometry. Convection by the wind gives the distortion term
$ t
I I
ΔrW = vW dτ
t−τ
I I
If the wind velocity vW is constant (no ground boundary layer), the integrated convection is just τ vW . With a
I
ground boundary layer, the term τ vW is calculated from the reference (constant) wind speed W , and the convection
produced by variation of the air velocity relative W is included in DI (or neglected). Note that the wing position
rQ is evaluated at past times. The wake geometry distortion DI (t, τ ) is the perturbation of the position from the
undistorted geometry, in I axes. By definition the wake geometry connects to the wing at τ = 0, so D(t, 0) = 0.
The distortion is produced by the self-induced velocities in the wing set. In general the distortion is required for
all nodes of the wake geometry:
⎧
⎪ DT Lk left tip vortex
⎪
⎪
⎨ DT Rk right tip vortex
Dk = DSLk left edge inboard sheet
⎪
⎪
⎪
⎩ DSRk right edge inboard sheet
DAik wing span stations, i = 0 to M + 1
Note that the tip vortex and sheet edges have the same undistorted geometry, but different distortion. Because
of the dominant role of the tip vortices, the most important information in the wake geometry is the tip vortex
position, and a less accurate definition of the inboard sheet geometry is often acceptable. For example, the free
wake geometry might be obtained only for the tip vortices. Also, the distorted geometry of the inboard sheet might
be described by just the sheet edges, with linear interpolation
rT R − rAi
w=
r T R − rT L
D = wDSL + (1 − w)DSR
giving the distortion at the aerodynamic span stations rAi .
The wing position rQ is required at the wing tips, and perhaps at the boundaries between the far wake panels.
The positions at the wing tips (span stations rT L and rT R ) are available as component input. If the tip vortex forms
at a span station rT V inboard of the wing tip, then rQ is obtained by interpolation from the component input. The
index i is found such that rAi ≤ rT V ≤ rA(i+1) ; then linear interpolation gives
rA(i+1) − rT V
w=
rA(i+1) − rAi
(rQ )T V = w(rQ )Ai + (1 − w)(rQ )A(i+1)
for the wing position at rT V L and rT V R . The conventions rA0 = rT L and rA(M +1) = rT R is used here. At the
boundary between far wake panels, the index i is found such that rAi ≤ rBj ≤ rA(i+1) ; and linear interpolation
gives
rA(i+1) − rBj
w=
rA(i+1) − rAi
(rQ )Bj = w(rQ )Ai + (1 − w)(rQ )A(i+1)
gives the wing position at rBj . The distortion at these boundaries can be interpolated using the same factors.
26-3.2 Betz Rollup
In the undistorted wake geometry, the initial span station of the tip vortex rT V can be obtained from Betz
rollup: calculated assuming that the centroid of the rolled up trailed vorticity is conserved. Alternatively, rT V can
be an input parameter. Consider the bound vorticity from rA to rB , rolling up into a trailed line. The centroid of
the trailed vorticity is at rC :
$ rB $ rB
∂Γ
rC (ΓA − ΓB ) = − r dr = ΓA rA − ΓB rB + Γ dr = ΓA rA − ΓB rB − (rA − rB )ΓM
rA ∂r rA
in terms of the mean bound circulation ΓM ; hence
rC = (1 − w)rA + wrB
with w = (ΓM − ΓB )/(ΓA − ΓB ). The spanwise displacement calculated by the Betz rollup model is multiplied
by an input factor f ; and for the wing tips, added to the input displacement. At the wing tips, it is assumed that the
vorticity that rolls up is between the nearest bound circulation peak and the tip (Γ = 0 and r = rT L or rT R at the
tip). At inboard stations, it is assumed that the vorticity that rolls up is between the nearest peak in the panel to the
left and the nearest peak in the panel to the right. With the single-peak wake model, this peak is at the maximum
bound circulation; with dual peaks in the circulation distribution, the left or right peak is used. The circulation
peaks are found at the aerodynamic span stations, so the limits of the rolled up vorticity (rA and rB ) are also at the
aerodynamic span stations; except that rollup at a wing tip includes from the last aerodynamic panel to the tip as
well. The mean bound circulation ΓM is evaluated assuming piecewise-linear variation of the bound circulation
over the span.
26-3.3 Interpolation and Extrapolation of Distortion
The undistorted part of the wake geometry can be evaluated at any time t and age τ as required. The distortion
may be calculated at a different set of times and ages however. For example, efficiency or convergence can lead
to using larger time and age increments in the free wake calculation, and the distortion might be calculated for
only the wake nearest the wing. Thus it can be necessary to interpolate or extrapolate the distortion. This is
accomplished by following the path of a specific wake element in time, hence by considering the geometry for a
fixed value of the time that a wake element was created, δ = t − τ . The distortion may be required at an age τ
beyond which it has been calculated. Let τ last be the maximum age of the available distortion. The distortion is
extrapolated by assuming that the vortex element is convected for time (τ − τ last ) by a constant velocity:

DI (t, τ ) = DI t − (τ − τ last ), τ last + (τ − τ last ) vconv
I
Note that the distortion is used at a constant value of δ. The convection velocity vconv is obtained from the
corresponding rigid, prescribed, or free geometry model. The free distortion is calculated for wake ages up to
τmax . The accuracy of the wake geometry may be improved by not using the last part of the calculated distortion,
hence starting the extrapolation at τlast less than τmax .
The distortion calculated at discrete points must be interpolated to arbitrary t and δ = t − τ . Assume that the
time and age at these discrete points have the same increment Δt. The distortion is available at:
D( , j) = D(t , δj )
where t = Δt and δj = jΔt (the distortion might actually be stored as a function of t and τ ). Note that with
= 1 to L and k = 0 to K, it follows that j = − k has the range j = to ( − K); and the time when a wake
element was created is always less than the current time, δ ≤ t . Then with square brackets indicating integer
truncation: ⎧% &
⎪ t
⎪
⎨ Δt for t ≥ 0
= % &
⎪
⎩ − −t − 1 for t < 0
⎪
Δt
⎧% &
⎪ δ
⎪
⎨ Δt for δ ≥ 0
j= % &
⎪
⎩ − −δ − 1 for δ < 0
⎪
Δt
gives
t ≤t≤t +1
δj ≤ δ ≤ δj+1
The interpolation factors are
t−t t
wt = = −
t +1 − t Δt
δ − δj δ
wδ = = −j
δj+1 − δj Δt
So linear interpolation in t and δ:

D = (1 − wt ) (1 − wδ )D( , j) + wδ D( , j + 1)

+ wt (1 − wδ )D( + 1, j) + wδ D( + 1, j + 1)
gives the required distortion. Alternatively, the calculated distortion can be stored as a function of t and τ :
D(t , τk ), k = − j. Then

D = (1 − wt ) (1 − wδ )D( , − j) + wδ D( , − j − 1)

+ wt (1 − wδ )D( + 1, − j + 1) + wδ D( + 1, − j)
is the required distortion. If = j, then the interpolation becomes D = D( + 1, j)τ /Δt = D( + 1, k = 1)τ /Δt.
The values of τ in this expression (k = − j, − j + 1, − j − 1) need not be in the calculated range of wake
age, k = 0 to klast . If k ≤ 0, then D = 0 is used. If k > klast , then the distortion is extrapolated:

D( , k) = D − (k − klast ), klast + (k − klast ) Δt vconv
I
as described above. If the calculated distortion starts at nonzero time tz , then t − tz replaces t in this procedure.
The distortion at τ = 0 is always D = 0, for all times.
26-3.4 Lifting Line Wing
The collocation points are at the quarter chord or three-quarter chord of the aerodynamic span stations. The
quarter chord point is on the lifting line. The three-quarter chord point is a half-chord aft of the lifting line, in the
direction of the local flow. The first panel of the wake geometry is used to get the local flow direction. Let r0 and
r1 be the positions of the leading edge (k = 0, τ = 0) and trailing edge (k = 1, τ = Δτ ) of the wake panel, at the
i-th span station:
I I
r 0 = rW (t, 0) = rQ (t)Ai
I
r1 = rW I
(t, Δτ ) = rQ (t − Δτ )Ai + Δτ vW
I
(ignoring the distortion and any variation of the wind velocity in a ground boundary layer). Then the three-quarter
chord collocation point is at
r1 − r0
r P = r0 + di
|r1 − r0 |
where di = ci /2 is the semichord. The quarter-chord collocation point is simply
I
rP = r 0 = r Q (t)Ai
(di = 0). Optionally a straight lifting-line can be used to generate the wake geometry. In this case, rQ is used
only at the points designated as the tips of the straight wing. Linear interpolation
rT R − rAi
w=
rT R − r T L
(rQ )Ai = w(rQ )T L + (1 − w)(rQ )T R
gives the position vector at the aerodynamic span stations rAi .
26-3.5 Display Geometry
It is conventional to draw the wake geometry in appropriate axes moving with the wing set. Hence a sensor
vector (component output) is available that is the calculated wake geometry in display axes. The display geometry
is the calculated geometry rotated to wing frame coordinates, and/or shifted to the wing set origin. Let Q be the
wing set frame. The analysis can use either wing set axes S or wing tip axes T for this frame. Then the display
geometry rQ consists of all wake geometry position vectors rI , transformed as follows:
ΔrQ = C QI (rI − rQI/I )
Optionally only the rotation or only the origin shift can be implemented (or neither). Note that rQI/I contains the
basic convection of the wake by the velocity of the air relative to the wing, caused by the wing motion itself. The
rotation matrix can be constructed from the basis vectors of the Q frame in I axes:
I
C IQ = i j k Q
The origin of the Q frame (rQI/I ) and the frame normal and longitudinal vectors (k and i) can be obtained from
a reference plane component, for Q=S or Q=T as required. Often these quantities should be filtered to eliminate
any high frequency motion of the reference. Note that the filtered normal and longitudinal vectors must be made
unit and orthogonal.
Chapter 27
WING WAKE GEOMETRY COMPONENT
27–1 Description
A Wing wake geometry component calculates the wake geometry for a wing set. This component incorporates
the standard features of wake geometry components. The wake geometry model can be rigid, prescribed, or free.
The prescribed geometry is the same as for the rotor wake geometry component. The models implemented have
the following features.
a) The rigid model calculates the wake geometry distortion from the mean interference velocity at the wing set.
b) The prescribed wake geometry is obtained from an empirical model for a hovering rotor.
c) The free wake geometry is obtained by calculating the distortion simultaneously for all wings in the wing set.
The wake model of the wing wake component is used in this free geometry calculation.
The rigid and prescribed models described here are both considered “rigid wake geometry.” For the trim task, it
is assumed that the wake geometry distortion is constant or periodic (in an appropriate frame). The free wake
geometry can include the influence of an airframe flow field, using the simple model of the airframe flow field
component; and can include the influence of the gust field.
Figure 1 illustrates the functionality of the wing wake geometry component.
Component Input
I
a) Aerodynamic interfaces at lifting line: location rQ at left wing tip, right wing tip, and collocation points (quarter
chord).
b) Wake geometry (from this component).
c) Wing frame coordinates: wing set plane normal k I , and longitudinal unit vector iI (S or T frame).
d) Position wing set origin: rSI/I or rT I/I .
e) Component input for rigid geometry:
I
1) Average interference velocity at wing set: vmean .
f) Component input for prescribed geometry:
1) Wing set force and moment (S or T axes).
2) Reference wing bound circulation peaks (mean): magnitude (Γmax , ΓL , ΓR ) and position (rmax , rL , rR )
for entire wing and multiple far wake panels; and centroid data (Γ, rC , rG , f , δ, set) for far wake panel edges.
g) Component input for free geometry:
1) Wing bound circulation peaks: magnitude (Γmax , ΓL , ΓR ) and position (rmax , rL , rR ) for entire wing
390 WING WAKE GEOMETRY COMPONENT
aero interfaces at lifting line aero interfaces at lifting line

location wing tips and QC location coll points
(QC or 3QC)
wing frame normal
wing frame long vector
position wing set origin wake geometry
average int velocity
mean bound circ peaks
bound circulation display geometry

bound circulation peaks
sensors
location of wings, bodies
airframe velocity relative air
airframe lift/q
gust amplitude
location of origin of gust field
Figure 27-1 Functionality of wing wake geometry component.

WING WAKE GEOMETRY COMPONENT 391
and multiple far wake panels; and centroid data (Γ, rC , rG , f , δ, set) for far wake panel edges; for each wing.
h) Component input for Betz rollup:
1) Wing bound circulation peaks.
2) Wing bound circulation: Γ at all span stations, for each wing.
i) Component input for airframe flow field influence:
1) Airframe velocity relative air.
2) Location of wings (left tip, middle, and right tip) and bodies (center and nose).
3) Airframe lift/q (L/q at wing-body, horizontal tail, and vertical tail).
j) Component input for gust influence:
1) Gust amplitude (velocity, angular velocity, gradients).
2) Location of origin of gust field.
Component Output
a) Aerodynamic interfaces at lifting line: location rPI of collocation points (quarter chord and three-quarter chord).
I
b) Wake geometry: rW .
c) Display geometry: rW , in S or T axes.
d) Sensor: wake geometry parameters.
27-3.1 Character of Wing Wake Geometry
was created. This geometry is distorted by the wake self-induced velocity. Figure 2 illustrates the calculated free
wake geometry of a wing.
The component uses quantities in wing set axes S, or wing tip axes T, depending on the definition of the
input variables. Wing tip axes should be used if possible, but wing set axes are probably acceptable. The general
convention is that the origin of the axes is at the center of the wing set, with the x–y plane the wing set “plane”.
The z-axis is up (positive lift direction), the x-axis is aft (positive velocity direction), and the y-axis is to the right.
These conventions affect the calculated wake geometry in several ways, depending on the model used:
a) Rigid geometry: definition of normal and inplane (if different empirical factors are used).
b) Prescribed geometry: definition of vertical convection and thrust.
c) Display geometry: selection of axes for transformed output.
It is assumed that the normal is in the z-axis direction of the axes used (S or T).
For the trim task, it is assumed that the wake geometry distortion is constant or periodic in frame F, which is
specified by the component frame. Often the distortion (but not the total position) is periodic in the inertial frame
I. An exception is the case of an aircraft in turning flight, for which the wing wake geometry component should
use a frame that includes the turn angle.
Figure 27-2 Calculated wing free wake geometry.

a) Wing frame coordinates: wing set plane normal and longitudinal unit vector (k I and iI ), from a reference frame
component.
b) Position wing set origin: rSI/I or rT I/I , from a frame component. This quantity is the reference position for
the display geometry.
I
c) Average interference velocity at wing set: perturbation velocity of the air vmean , in inertial axes; obtained by
average over the wing surface, summation over all wings, and then summation of all contributions. The average
over the wing surface can be performed by a reference frame component, and the summation over all wings by a
differential equations component. Other components require that this velocity be divided into induced (from this
I
wing set) and interference (from all other sources) terms. Finally vmean can be calculated by using a reference
frame component to sum all contributions and transform to inertial axes. What contributions are included defines
the interaction between wing sets that is accounted for in the rigid geometry. A filter component can be used to
get the mean velocities.
d) Wing set force and moment: the total aerodynamic load (force F and moment M ) acting on the entire wing set,
e) Reference wing bound circulation peaks: peak bound circulation values along the span, and the corresponding
span station of the peaks (left, right, and maximum peaks). The prescribed geometry requires the mean circulation;
the average over all wings in the wing set could be used, instead of reference wing values. The model requires the
circulation peaks over the span of the entire wing.
f) Wing bound circulation peaks: magnitude (Γmax , ΓL , ΓR ) and position (rmax , rL , rR ) for entire wing and
multiple far wake panels; and centroid data (Γ, rC , rG , f , δ, set) for far wake panel edges; for each wing.
g) Wing bound circulation: Γ at all span stations (midpoints of the aerodynamic panels), for each wing.
h) For the airframe flow field influence: airframe velocity relative air; airframe lift/q; and location of wings (left
tip, middle, and right tip) and bodies (center and nose). The locations rI must be in inertial axes.
i) For the gust influence: gust amplitude (in wind/gust axes); and the location of the origin of the gust field (in
inertial axes).
I
A filter component can be used to get the mean input for the wake geometry model. The locations rQ of the
I I
wing should not be filtered. A filtered k or i will not be a unit vector.
27-3.4 Wake Geometry Models
The rigid wake geometry is calculated by assuming that the wake elements are all convected by the average
I
interference velocity vmean . The variation of the wind velocity in a ground boundary layer is neglected in this
model.
The prescribed wake geometry combines several models in a single functional form: rigid (convection by the
normal interference velocity); empirical for a hovering rotor (from the rotor loading); and input. The distortion is
calculated by assuming that the geometry is described by two-stage vertical convection and exponential spanwise
contraction. The effects of turn rate or transient motion of the rotor are included in the basic undistorted geometry,
but not in the prescribed distortion. Interference from other aerodynamic components is not considered. The
variation of the wind velocity in a ground boundary layer is neglected in this model.
The free wake geometry is calculated using the model of the wing wake component. Ground effect can be
included in the free geometry calculation. Only the far wake region of the wing wake component is considered
here (note that the terms “near wake” and “far wake” have different meanings for the wing wake component and
the free geometry calculation). The far wake trailed vorticity can be divided into several spanwise panels. The
wake geometry is described by the positions of the vortex elements at the boundaries of these panels, separately
for the tip vortices and the inboard sheet edges if the trailed wake rolls up at a boundary. The wake geometry
distortion can be calculated optionally for the tip vortices; for all the rolled-up trailed vortices; and for the inboard
sheet edges as well. With multiple far wake trailed vorticity panels, the trailed lines at the panel edges can be
consolidated into a single rolled-up line using the trailed vorticity moment to scale the rate of rollup. The rigid
or prescribed distortion is used where the free distortion is not calculated. The possibility that the distortion is
identical for different wings (perhaps with a time shift) is also considered.
The wake geometry distortion is constant or periodic in the trim task. The distortion is calculated for the
entire wing set simultaneously; for the entire period in the trim task, or at the next time step in the transient task;
using the time increment Δt and wake age increment Δτ = Δt of the wake geometry component. Then from
this saved distortion, the component evaluates as required the wake geometry for a specified wing; interpolated or
extrapolated to the times of the solution procedure, and the wake ages of the wake component.
27–4 Theory
27-4.1 Rigid Distortion
The rigid distortion is calculated by assuming that the wake elements are all convected by the average
I
interference velocity vmean . This velocity is the sum and average of all sources of induced and interference
velocities acting on this wing set. This uniform convection gives
DI (t, τ ) = vmean
I
τ
Empirical corrections factors for the normal (fn ) and inplane (fi ) convection are introduced, with k I (S or T frame)
defining the normal direction. Thus
I
DI (t, τ ) = fn kk T + fi (I − kk T ) vmean I
τ = vconv τ
I
is the distortion, from the convection velocity vconv .
27-4.2 Prescribed Distortion
The prescribed distortion is calculated by assuming that the geometry is described by two-stage vertical
convection and exponential spanwise contraction. The model is the same as that of the rotor wake geometry
component. It is intended for application of the wing wake geometry component to rotors, particularly in hover.
If the free distortion is only calculated for the tip vortices, the prescribed model can be used for the inboard wake.
27-4.3 Free Distortion
27-4.3.1 Background
The free distortion is calculated by integrating in time the self-induced velocities at all collocation points
on all wakes of the wing set. Special techniques are required to keep the computational effort reasonable. The
following are the key aspects of the approach, both from previous work and as developed for this component.
a) When calculating the velocity at a collocation point, the wake is divided into “near wake” and “far wake” regions.
The velocity contribution of the far wake is small, so when the velocity is required again in the algorithm, only the
contribution of the near wake need be recalculated (ref. 1).
b) The near wake regions are defined by transition points between far wake and near wake. It is assumed that the
relative age of these transition points depends on the time the wake element was created, δ = t − τ , but not on the
wake age τ (ref. 2).
c) During the basic time step, the existing wake is convected by the wind and the induced velocity with little
relative distortion, while the wings move and generate new wake behind the trailing edges (ref. 2).
d) Implicit time integration methods (with relaxed iteration) are used for stability. In the predictor-corrector form,
these methods require only one evaluation of the velocity per time step.
e) Integration of the wing-induced velocity in time is approximately equal to the average velocity from the bound
vortex (a vortex line segment), which is equivalent to the velocity from a vortex sheet element (ref. 2).
f) For hover, the initial convection velocities are defined such that the wake leaves the wing tangent to the wing
surface.
g) The effect of the core vorticity distribution on the self-induced velocity of a vortex arc is obtained using an
appropriate cutoff distance (ref. 3).
h) The wake model of the wing wake component is used for the free distortion calculation, including features such
as the dual-peak model; multiple rollup of the trailed vorticity; models for entrainment and stretching in the rollup
process; and the vorticity distribution and core radius growth of the tip vortices.
i) The calculation of the distortion is performed with time as the outer loop, for both the trim and transient tasks.
The trim distortion is constant or periodic (in an appropriate frame). The calculation includes an iteration between
revolutions with distortion increment relaxation and propagation; and an outer iteration with distortion relation.
j) The velocity at a collocation point is calculated with a full update (including determination of the near wake
regions), a near wake update, or no update. The update frequency can be chosen to balance efficiency and accuracy.
27-4.3.2 Distortion Calculation
The undistorted geometry is calculated from the position in the air at which the wake element was created,
plus convection by the wind. Then the distortion produced by the self-induced velocities of the wing set is added.
Let t be the current time, and τ the age of the element in the wake. Thus δ = t − τ is the time when the vorticity
was created. Then
I
rW I
(t, τ ) = rQ (t − τ ) + τ vW
I
+ DI (t, τ )
is the wake geometry. Here vW is the constant or reference wind velocity; and rQ is the wing position, evaluated
at past times. Note that the time at which the vorticity was created (t − τ ) identifies a particular element in the
wake. The wake geometry distortion D(t, τ ) is the perturbation of the position from the undistorted geometry, in
I axes. By definition the wake geometry connects to the wing at τ = 0, so D(t, 0) = 0. The distortion is produced
by the self-induced velocities in the wing set. The free distortion is calculated for wake ages up to τmax .
In general the distortion is required for all structures of the wake geometry: for the left and right tip vortices,
and the left and right edges of the inboard sheet; for each trailed wake panel; for each wing. The wake geometry
distortion can be calculated optionally for the tip vortices; for all the rolled-up trailed vortices; and for the inboard
sheet edges as well. The free distortion is initialized to the rigid or prescribed geometry for all structures; thus the
rigid or prescribed distortion is used where the free distortion is not calculated. If the wake does not roll up at a
wing tip or panel boundary, then the tip vortex geometry is not required and the free distortion is not calculated
(the tip vortex geometry still exists in the component output, set to the rigid or prescribed distortion). If the wake
is rolled up at a boundary, then the right tip vortex geometry of the left panel is identical to the left tip vortex
geometry of the right panel. If the wake is not rolled up at a boundary, then the right sheet edge geometry of the
left panel is identical to the left sheet edge geometry of the right panel. The distortion at the wing span stations
(for the near wake) is obtained by interpolation from the distortion of the sheet edges.
The distortion may be required at an age τ beyond which it has been calculated. Let τ last be the maximum
age of the available distortion (perhaps less than τmax ). The distortion is extrapolated by assuming that the vortex
element is convected for time (τ − τ last ) by a constant velocity:

DI (t, τ ) = DI t − (τ − τ last ), τ last + (τ − τ last ) vconv
I
Note that the distortion is used at a constant value of δ. Extrapolation uses the convection velocity vconv , obtained
from the average distortion increment at τlast :
1 1
vconv = D(t, τlast ) − D(t − Δt, τlast − Δt)
T t
Δt
(in inertial axes, for each distorted structure). This convection velocity is recalculated at the end of each revolution of
the wake geometry algorithm, and the final value is saved for use when the wake geometry is evaluated. Optionally,
the convection velocity from the rigid or prescribed model can be used to extrapolate the free distortion. For the
transient task, the velocity is filtered over the latest period; or the trim convection velocity can be used. The
distortion is calculated and used in the inertial frame I. For the trim task, it is assumed that the wake geometry
distortion is constant or periodic, in an appropriate frame F (specified by the component frame). The trim distortion
is calculated over a reference revolution tZ to tZ + T , where T is the period, and the start of the revolution tZ is
defined by the solution procedure. Thus the distortion at time t outside this reference revolution is:
DI (t, τ ) = C IF (t)C F I (t0 )DI (t0 , τ )
where t0 is t reduced to the reference revolution by adding or subtracting a multiple of the period T as required.
A subset of the wings can have identical trim distortion (in an appropriate frame, with a time shift). Then it is only
necessary to calculate the distortion for one wing of the subset (the parent wing). The distortion of the child wing
is evaluated from the parent:
I I
Dchild (tchild , τ ) = Dparent (tparent , τ )
where tchild = Δt, tparent = ( + shift )Δt, and shift is an input time shift. For example, with N blades
uniformly spaced over the period T = J Δt, the time shift for the m-th blade is shift = (m/N )J (which is an
integer if J is a multiple of N ). If tparent is outside the reference revolution, it is reduced to the reference revolution
in order to evaluate the distortion; with the transformation between the F and I frames as necessary. If shift and
hence parent are not integers, then the parent distortion is interpolated over t and δ = t − τ to obtain the child
distortion.
The distortion is calculated by integrating in time the self-induced velocity q acting on the wake element
created at time δ = t − τ : $ t
D(t, τ ) = q(t = σ, τ = σ − δ) dσ
t−τ
The integration is performed for fixed δ. The velocity q does not include the reference wind velocity; with a ground
boundary layer, it does include the variation of the wind velocity relative the reference value. The induced velocity
is evaluated by integrating over all vorticity in the wake. For incompressible flow, the Biot-Savart law gives the
velocity as an integral of the wake strength times an influence coefficient:
$$
q(t) = C(t, τ ; r)G(t, τ ; r) dr dτ
where τ is the wake age and r the wing span variable, so the integral is over the wake surface. The influence
coefficient depends on the wake geometry. The wake strength G depends on the wing bound circulation at past
times, Γ(t − τ ; r).
The wake age is discretized by using the geometry and strength only at a set of ages τk = k Δτ , k = 0 to
K, for fixed wake age increment Δτ . Time is discretized in the distortion calculation, with fixed increment Δt.
The time increment and the wake age increment must be equal, Δt = Δτ , in order to implement the integration
algorithm. The number of steps J in one revolution is defined. Then if the trim distortion is periodic, the time
increment Δt = T /J is calculated from the period T . If the trim distortion is constant, Δt must be input directly.
The distortion calculation in the transient task uses the time increment from the trim task.
For the trim task, the distortion is calculated over one revolution: t = tZ to t = tZ + JΔt = tZ + T . The
start of the revolution tZ is defined by the solution procedure. At the first time step of the solution procedure, the
distortion is calculated for the entire wing set simultaneously, for the entire revolution, using the time increment
Δt and wake age increment Δτ = Δt of the wake geometry component. Then from this saved distortion, the
component evaluates as required the wake geometry for a specified wing; interpolated or extrapolated to the times
of the solution procedure, and the wake ages of the wake component.
For the transient task, the distortion is calculated over the time range tB to tE = tB + N Δt, where tB and
tE are defined by the solution procedure. The transient distortion is initialized at tB using the trim distortion at
tB = tZ + B Δt (which need not be in the trim reference revolution). At each step of the solution procedure, the
distortion is calculated for the entire wing set simultaneously, for times tB + Δt up to the current time, using
the time and wake age increment Δt of the wake geometry component. Then from this saved distortion, the
component evaluates as required the wake geometry for a specified wing; interpolated or extrapolated to the times
of the solution procedure, and the wake ages of the wake component.
Figure 3 is an outline of the process for calculating the wake geometry distortion. Figure 3a shows the process
for the trim task, which is executed at the first time step of the solution procedure. The process for the transient task
is similar, but omits initialization, iteration, both relaxations and saves, revolution and time loops, propagation,
and debug print. The process for the transient task is executed at the next time needed in the transient range. In
figure 3a, the term “structure” refers to the wake structures for which the distortion is calculated: left and right
rolled-up vortices and left and right sheet edges (depending on the rollup options and the calculation options), for
each trailed wake panel, for each wing.
Since the trim distortion is periodic or constant, the solution process includes an outer iteration. This is a
successive substitution iteration, with relaxation on the distortion to improve convergence. A specified number of
iterations are performed; there is no test for convergence. Often relaxation is not required, and with the relaxation
factor λI = 1 the iteration is equivalent to simply extending the time integration for more revolutions; hence the
iteration can be omitted in such a case. The basic loop of the process is the integration in time, which is performed
for M revolutions, each revolution covering the period T . The trim distortion is stored only for the reference
revolution. The number of revolutions in the time integration should be greater than the maximum wake age, so
the new wake generated is at least equal to the amount of distortion needed. Typically M T = 2τmax is used. For
each time step, the velocity q at a collocation point is integrated to get the change in distortion there, during time
t − Δt to t. At time t, the wake strength is known for all wake ages τ > 0; and during the trim task, for τ = 0 as
well, since the circulation is periodic.
The distortion is calculated by integrating the self-induced velocity q acting on the wake element created at
time δ. The discretized integral at time t for δ = t − τ is:
$ t
D(t, τ ) = D(t − Δt, τ − Δt) + q(t = σ, τ = σ − δ) dσ = D(t − Δt, τ − Δt) + ΔD
t−Δt
where ΔD is the distortion increment, evaluated from the velocity at the current and past times, and constant δ.
The general integration algorithm consists of a predictor followed by an iterated corrector with relaxation:
qjold = 0
ΔDj = Δt P (qj−1 , . . .) predictor
iteration
evaluate qj from Dj
qj = λR qj + (1 − λR )qjold relaxation
qjold = qj
ΔDj = Δt C(qj , . . .) corrector
where subscript j indicates q or D at time t = jΔt. The following integration algorithms are implemented.
a) Euler (first order): ΔDj = Δt qj . Using only the predictor gives explicit Euler integration (similar to ref. 4).
b) Trapezoidal (second order): ΔDj = Δt 2 (qj + qj−1 ). A single pass (no iteration) gives the trapezoidal predictor-
corrector algorithm. The result with one iteration and λ = 12 is similar to the PIPC algorithm of Bagai (ref. 5).
c) Backward difference (second order). The result with one iteration and λ = 12 is similar to the PC2B algorithm
of Bhagwat (ref. 6).
method predictor corrector

Euler ΔDj = Δt qj−1 ΔDj = Δt qj
Trapezoidal ΔDj = Δt qj−1 ΔDj = Δt
2 (qj + qj−1 )
Backward difference ΔDj = Δt qj−1 ΔDj = − 23 Dj−1 + Dj−2 − 13 Dj−3 + 23 Δt(qj + qj−1 )
evaluate wake strength, circulation peak positions, undistorted geometry

initialize D, Dold , ΔDold (rigid or prescribed geometry), vconv
iteration
relax distortion: D = λI D + (1 − λI )Dold ; Dold = D
revolution loop: M revs
time loop: t = tZ + Δt to tZ + T (j = 1 to J)
qold = 0
predictor: ΔD = Δt P (wake age τ = Δt to τmax )
D(t, τ ) = D(t − Δt, τ − Δt) + ΔD
integration iteration
velocity calculation, each structure: at t, wake age τ = 0 to τmax
total velocity qT
every nF steps in t
full update for all τ (set tfar = t)
else
at wing (τ = 0): full update always
every nN steps in t: near wake update
else no update
bound vortex velocity qB
wind and gust velocity qW
airframe flow field velocity qA
initial convection velocity qK
sum velocity: q = qT (τ ) + qB (τ ) + qW (τ ) + qA (τ )
initial convection: q = (1 − τ /τK )qK + (τ /τK )q if τ < τK
integration, each structure: wake age τ = Δt to τmax
relax velocity: q = λV q + (1 − λV )qold ; qold = q
corrector: ΔD = Δt C
D(t, τ ) = D(t − Δt, τ − Δt) + ΔD
relax ΔD: ΔD = λR ΔD + (1 − λR )ΔDold
D(t, τ ) = D(t − Δt, τ − Δt) + ΔD
distortion increment: δD(τ ) = ΔD − ΔDold ; ΔDold = ΔD
propagation, each structure: wake age τ = Δt to τmax
future time loop: (t + iΔt) i ≥ 1 while (τ + iΔt) ≤ τmax
propagate distortion: D(t + iΔt, τ + iΔt) = D + δD(τ )
consolidation, each structure
calculate vconv
debug print: rms(D − Dold )
Figure 27-3a Outline of process for calculating wake geometry distortion.

full update: total velocity qT at t and τ

δ1 = τmax + t − tfar − τ + Δt
evaluate collocation point position: rP (τ ) on structure
qN = qF = 0
wing loop
1 = 0, set current region = neither
nT (δ)
wake age loop: τle = 0 to τ + τbelow ; τte = τle + Δt
evaluate wake geometry: rle at τle , rte at τte
calculate velocity increment Δq
near wake: |Δq| > ΔqN W (or collocation point on end of line segment)
qN = qN + Δq
if current region is not near wake
1 = nT (δ)
nT (δ) 1 +1
1 nT ) = τle
τT B (δ,
set current region = near wake
else far wake
qF = qF + Δq
if current region is near wake
1 nT ) = τle − Δt
τT E (δ,
set current region = far wake
if current region is near wake
1 nT ) = τ + τbelow
τT E (δ,
1 = qF
qfar (δ)
qT = qN + qF
Figure 27-3b Outline of process for calculating wake geometry distortion.

near wake update: total velocity qT at t and τ

δ1 = τmax + t − tfar − τ + Δt
qN = 0
wing loop
wake age loop: τle = near wake; τte = τle + Δt
qN = qN + Δq
1
qT = qN + qfar (δ)
no update: total velocity qT at t and τ

q1 = 0
wing loop
wake age: τle = 0, τte = Δt
q1 = q1 + Δq
qT (t, τ ) = qT (t − Δt, τ − Δt) + q1
bound vortex: velocity qB for t − Δt to t, and τ

qB = 0
wing loop
wake age: τle = 0, τte = Δt
calculate bound vortex velocity increment ΔqB (using sheets if |ΔqB | > ΔqBV )
qB = qB + ΔqB
wind, gust, and airframe flow field: velocity qW + qA at t and τ

evaluate wind and gust velocity qW at rP
evaluate airframe flow field velocity qA at rP
Figure 27-3c Outline of process for calculating wake geometry distortion.

The velocity at time t is evaluated from the sum
q(τ ) = qT (τ ) + qB (τ ) + qW (τ ) + qA (τ )
where qT is the total self-induced wake velocity; qB is the effect of the wing-induced velocities (bound vortices);
qW is the wind velocity relative the reference value (zero if there is no ground boundary layer) plus the gust
velocity; and qA is the airframe flow field velocity.
Note that if the relative distortion of the wake from t−Δt to t is ignored, then during this time the only change is
the addition of the new wake of age τ = 0 to Δt directly behind the wings. Then q(t, τ ) ∼ = q(t−Δt, τ −Δt)+q1 (τ ),
where q1 is the velocity at of the new wake generated behind the wing during t − Δt to t. With this approximation,

trapezoidal integration gives ΔD = Δt q(t − Δt, τ − Δt) + 12 q1 (τ ) . Johnson (ref. 4) used this integration

method. Similarly, Scully (ref. 2) used q(t − Δt, τ − Δt) ∼= q(t, τ ) − q1 (τ ), hence ΔD = Δt q(t, τ ) − 1 q1 (τ ) ;
2
with an inner loop for constant wake age.
Since the velocity and distortion are periodic in time, relaxation of the distortion increment can be introduced
to improve convergence. In addition, propagation of the distortion information can be included in the trim solution.
The distortion increment ΔD = q Δt is not the same as that calculated during the last revolution; the difference
is δD = ΔD − ΔDold . The difference δD affects all future values of D at this t − τ . Since the trim distortion is
periodic, δD also affects values of D at past times and larger age. Thus the propagation procedure adds δD to all
values of D(t, τ ) at future time t and fixed t − τ (subtracting the period T from t whenever the time exceeds the
reference revolution, while τ is less than the maximum wake age τmax ). Figure 3a outlines the process described
above. Figure 4 illustrates the integration, extrapolation, and propagation of the distortion.
27-4.3.3 Velocity Calculation
The velocity contributions qT , qB , qW , and qA are required at time t, for wake ages τ = 0 to τmax . These
velocities are calculated at a collocation point that is on the wake (as specified by the wake structure and age). As
defined above, qT is the total wake-induced velocity (excluding the bound vortices) at time t; qB is the effect of
the wing-induced velocities (bound vortices) during t − Δt to t; qW is the wind velocity (relative the reference
value) plus the gust velocity at time t; and qA is the airframe flow field velocity at time t. In order to minimize the
computational effort in calculating qT , for each collocation point the wake is divided into “near wake” and “far
wake” regions. The velocity contribution of the far wake is small, so when the velocity is required again in the
algorithm, only the contribution of the near wake need be recalculated. Thus the following update strategy is used
in calculating qT .
a) Full update every nF steps in time t: Calculate qT by summing the velocity contributions
from all elements in the wake. The near wake and far wake regions are determined, in terms
of the wake age relative the collocation point. The wake of age τle behind the i-th wing is in
the near wake if |Δq| > ΔqN W , where ΔqN W is an input criterion and Δq is the sum of the
velocities from all vortex elements at τle . The contribution to qT from the far wake is stored
as qfar .
b) Else near wake update every nN steps in time t: Calculate qT by summing the velocity
contributions from all elements in the near wake, and adding the contribution of the far wake.
wake age τ = τ max τ=0
propagation
time t
time integration
extrapolation
one period
time wake created δ = t – τ
extrapolate store
distortion distortion
wing
Figure 27-4 Distortion integration, extrapolation, and propagation.

c) Else no update: Calculate qT at time t from the contribution of the new wake (created during
last time step), and qT at the previous time step t − Δt.
Figure 3 shows the complete algorithm. The full and near wake update frequencies are chosen to balance efficiency
and accuracy. During the basic time step, the existing wake is convected by the wind and the induced velocity
with little relative distortion, while the wings move and generate new wake behind the trailing edges. Thus with
no update, only the velocity contribution from this new wake is calculated. As the wake distorts, it is necessary to
recalculate the velocity from at least the near wake in order to maintain accurate integration. A full update involves
the most computation, but also gives the most accurate value for the total velocity.
The near wake regions are defined by transition points between far wake and near wake. The near wake is
determined during a full update calculation of the velocity for a collocation point at age τ on a wake structure; it
consists of regions τle = τT B to τle = τT E behind the i-th wing. These transition points are used during a near
wake update. In order to minimize storage, it is assumed that the relative age of these transition points (τT B − τ
to τT E − τ ) depends on the time the wake element was created, δ = t − τ , but not on the wake age τ . Figure 3b
outlines the determination of the near wake regions during a full update. The near wake and far wake are defined
by nT sets of transition points τT B and τT E , and the velocity contribution qfar of the far wake. These quantities
are stored as a function of δ = t − τ (and wing number i for the transition points). Actually the data are stored as
a function of
δ1 = t − tfar − τ + τmax + Δt
where t is the current time; tfar is the time of the latest full update; and τmax is the maximum extent of the free
wake geometry. Figure 3c outlines the use of the near wake regions during a near wake update. The near wake for
the wake age loop in figure 3c consists of the wake created since the last full update (which calculated qfar ), and
1 transition regions:
the wake defined by nT (δ)
6
0 to τshift − Δt
τle =
1 + τshift
τT B (δ) to 1 + τshift
τT E (δ)
where the time since the last full update at this δ is τshift = min(τ, t − tfar ); τshift is positive, so as time increases
after a full update the near wake regions move down the wake (older τle ). The following table summarizes the
values of δ1 encountered. Note that δ1 is positive, with a maximum value of δ1 = τmax + nF + 1. Figure 5 illustrates
the velocity update algorithm.
full update use during

every nF time step new wake near wake update
t= tfar tfar + i Δt tfar + i Δt
τ= from 0 0 Δt
to τmax τmax
δ =t−τ = from tfar − τmax tfar + i Δt tfar − τmax + i Δt
to tfar tfar + (i − 1)Δt
δ1 = from Δt τmax + (i + 1)Δt (i + 1)Δt

to τmax + Δt τmax + i Δt
τ=0
δ =t –τ
multiple n N
multiple n F
set t far
multiple n N
multiple n F
set t far
full update
find near wake, using Δ q NW
store transition points and q far at δ
near wake update

use tfar to find transition point age shift
and wake created since q far calculated (dotted line)
Figure 27-5 Velocity update algorithm.

The wake model of the wing wake component is used for the free distortion calculation. In the rolled-up
wake (the far wake region of the wing wake component), the trailed vorticity can be divided into several spanwise
panels. For each panel, the wake strength depends only on the peak bound circulation values (left and right peaks,
or maximum). The self-induced velocity at a collocation point rP must be evaluated at time t, from the wake
geometry at t and the bound circulation at t and all past time. The wake is discretized in both span and age. Thus
the integral equation for the wake-induced velocity becomes:
KW
q(t) = CL (t, τk )GL (t, τk ) + CR (t, τk )GR (t, τk )
wings panels k=0
including a summation over all wings of this component, and perhaps a summation over spanwise panels as well.
In the wing wake component, the influence coefficients and wake strength are evaluated as separate component
output, which are then used to calculate the induced velocity. Here the induced velocity is evaluated directly. The
wake vorticity strength is
G(t, τ ) = Γ(t − τ )
For a single-peak circulation distribution (defined by ΓL = 0), there is only one term, from either the maximum
circulation Γmax or the outboard peak. The collocation point rP is located at some age less than τmax on the
wake. The wake used to calculate the induced velocity there extends from the wing to Mbelow revolutions below
the collocation point (k = 0 to KW ).
The induced velocity is calculated as in the wing wake component for points off the wing. The near wake and
extended far wake models of the wing wake component are not considered here. Since the collocation points are
now on the wake surface, special treatment is required for the induced velocity from vortex elements adjacent to
the collocation point. For a collocation point on a rolled-up trailed vortex line, the induced velocity from the two
adjacent line segments is calculated by replacing them by circular-arc vortex line segments. The effect of the core
vorticity distribution on the self-induced velocity of a vortex arc is obtained using an appropriate cutoff distance
(ref. 3). For a collocation point on an inboard vortex sheet, no change is considered appropriate for the typical
case where the sheet is approximated by line segments with large cores. If vortex sheet elements are used, a large
sheet thickness dvs can be specified if necessary.
The term qB is the velocity produced by the wing-induced velocities (bound vortices) during t − Δt to t.
Integration of the wing-induced velocities during this time interval gives the distortion:
$ t
ΔD = qwing dt = qB Δt
t−Δt
so qB is the average wing-induced velocity. For an instantaneous value of the wing-induced velocity, a good model
of the flow about the wing would be required. For the average value, representing the wing by a bound vortex
appears to be adequate. The average velocity from the bound vortex (a line vortex) as the wing moves from time
t − Δt to time t is equivalent to the velocity produced by a vortex sheet defined by the (convected) wing positions.
The strength of the vortex sheet should vary linearly from Γ(t − Δt) at the trailing edge to Γ(t) at the leading
edge. The vortex sheet element being used assumes constant strength of the shed vorticity in this direction, so the
element must have zero strength at the leading edge and strength
1
Γ(t) + Γ(t − Δt)
2
at the trailing edge. For efficiency, a rectangular-planar approximation of the wake sheet geometry is used; and
the velocity is calculated from a line segment unless its magnitude is greater than an input criterion ΔqBV .
The term qW is the velocity at a collocation point that is produced by the wind variation in a ground boundary
layer, and by the gust velocity. Thus qW is the wind velocity at position rW (t, τ ), minus the reference wind
velocity vW ; plus the gust velocity. The wake position rW is measured from the origin of the inertial frame, in
inertial axes, as required to obtain the velocity in the ground boundary layer and in the gust field. The gust velocity
is calculated as described for the gust component.
The term qA is the velocity at a collocation point that is produced by the airframe flow field. This interference
velocity is calculated using the simple model of the airframe flow field component, consisting of a set of wings
and bodies. A wing is modelled as a horseshoe vortex (lift) and a doublet line (thickness). A body is modelled by
the potential flow about a nonlifting body of revolution. These wings and bodies produce velocity perturbations
that distort the wake geometry, but the mutual interference between the wake elements and the airframe flow field
elements is not included.
With a three-dimensional wing, the Kutta condition requires that the wake leave the trailing-edge tangent to
the wing surface (ref. 7). In the absence of a calculation of the detailed flow field near the wing, this requirement
can be satisfied by using an initial convection velocity qK = Γ/πc, where Γ and c are the section bound circulation
and chord. Note that with the bound vortex at the quarter chord, this is its induced velocity at the three-quarter
chord. This result is obtained using the zero-lift chord line for the trailing-edge bisector, and a lift-curve slope of
2π. The velocity direction can be obtained from the geometry of the bound vortex and the collocation point. Let
rL and rR be the end points of the bound vortex at time t; and rW the wake collocation point at t and τ = Δτ .
Then
Γ(t) + Γ(t − Δt) (rW − rL ) × (rW − rR )
qK =
2πcmean |(rW − rL ) × (rW − rR )|
is the initial convection velocity. With the single-peak model, Γ is the maximum bound circulation. With the
dual-peak model, Γ is the left or right circulation peak, for the collocation point respectively on the left or right of
the trailed wake panel. The initial velocity qK is used at τ = 0, and the wake-induced velocity q at age τ ≥ τK , so
(1 − τ /τK ) qK + (τ /τK ) q
replaces q for τ < τK (figure 3a). The age τK must be selected based on correlation with measured wake
geometry and performance. Optionally this initial convection velocity can be ignored (qK never replacing q).
Hover performance calculations require τK > 0, but in forward flight best results are obtained with a much smaller
value of τK .
27-4.3.4 Consolidation
With multiple far wake trailed vorticity panels, the trailed lines at the panel edges can be consolidated into
a single rolled-up line, using the trailed vorticity moment to scale the rate of rollup. Figure 6 illustrates the
wake model options. The rollup is not well calculated even with many trailed vortex lines, because of the coarse
discretization and the neglect of viscosity. Hence it can be useful to impose consolidation in the wake geometry
calculation. The consolidation model is intended for use with a far wake trailed vorticity panel for each wing
aerodynamic panel, so there is a trailed vortex line at each wing panel edge. The trailed vorticity is partitioned
into sets of adjacent lines that have the same sign (bound circulation increasing or decreasing). It is assumed that
a) rolled-up wake model
b) multiple-trailer wake model
c) multiple-trailer wake model, with consolidation (entrainment form)
d) multiple-trailer wake model, with consolidation (compression form)
Figure 27-6 Illustration of wake models.

all the vorticity in a set eventually rolls up into a single vortex, located at the centroid of the original vorticity
distribution (refs. 8 and 9). For each set, the total strength Γ, centroid rC , and moment (radius of gyration) rG of
2
the trailed vorticity in the set are calculated. Then the characteristic time rG /Γ is taken as a measure of the rate of
consolidation (refs. 10 and 11). The consolidation is implemented after the integration and propagation steps in
the wake geometry calculation (figure 3). Propagation does not maintain the consolidated geometry, so should be
turned off when consolidation is used.
For trailed vortex elements at wake age τ , the total strength and moment are evaluated at the time the vorticity
was created (t − τ ), and used to calculate the time constant τcns = kcns rG
2
/Γ. Then the fraction of rollup, fcns , is
evaluated using an exponential (ref. 10), linear, or power dependence on wake age:
⎧
⎪ 1 − exp(−(τ − τBcns )/τcns ) exponential
⎨
fcns = (τ − τBcns )/τcns linear
⎪
⎩
((τ − τBcns )/τcns )n
power
The consolidation starts at wake age τBcns . In addition, a maximum consolidation fraction fEcns ≤ 1 can be
specified. The consolidation can be accomplished by entrainment or by compression.
With the entrainment form, vortex lines are consolidated into a single line of strength fcns Γ. For the k-th
trailed vorticity line, originating from the panel edge at rEk , fk is the fraction of the vorticity in the set that
is within the distance |rEk − rC | from the centroid. Then the vortex lines in the set that are consolidated at
age τ are those for which fk ≤ fcns (all lines within the minimum |rEk − rC | such that the strength is fcns Γ).
The consolidation is implemented by replacing the position of each consolidated line with the position of the

centroid: rcns = δk rW k / δk ; where δk is the strength of the k-th trailed vortex line, and the sums are over
the consolidated lines in the set. The position rW is the sum of the undistorted position and the distortion D(t, τ );
it is actually the distortion D that is replaced. In the velocity calculation, circular-arc vortex line segments are
used if the collocation point is on any of the consolidated lines.
With the compression form, fcns is the fraction of consolidation. If fcns ≥ 1, the position of each line is

replaced with the position of the centroid: rcns = δk rW k / δk ; where δk is the strength of the k-th trailed
vortex line, and the sums are over all the lines in the set. If fcns < 1, the position of each line is replaced with
(1 − fcns )rW k + fcns rcns .
For a set of trailed lines at the edge of the wake, the rollup may be dominated by the flow over the wing tip,
so the final position of the consolidated lines is at the edge rather than at the centroid. For such cases the centroid
rcns is replaced by wE rtip + (1 − wE )rcns . The input factor wE = 1 for tip rollup at the edge of the wake, and
wE = 0 for tip rollup at the centroid.
27–5 References
1) Landgrebe, A.J. “An Analytical Method for Predicting Rotor Wake Geometry.” Journal of the American
Helicopter Society, Volume 14, Number 4, October 1969.
2) Scully, M.P. “Computation of Helicopter Rotor Wake Geometry and Its Influence on Rotor Harmonic Airloads.”
3) Bliss, D.B.; Teske, M.E.; and Quackenbush, T.R. “A New Methodology for Free Wake Analysis Using Curved
Vortex Elements.” NASA CR 3958, December 1987.
4) Johnson, W. “A General Free Wake Geometry Calculation for Wings and Rotors.” American Helicopter Society
51st Annual Forum, Ft. Worth, TX, May 1995.
5) Bagai, A., and Leishman, J.G. “Rotor Free-Wake Modeling Using a Pseudo-Implicit Relaxation Algorithm.”
Journal of Aircraft, Volume 32, Number 6, November-December 1995.
6) Bhagwat, M.J., and Leishman, J.G. “Stability, Consistency and Convergence of Time-Marching Free-Vortex
Rotor Wake Algorithms.” Journal of the American Helicopter Society, Volume 46, Number 1, January 2001.
7) Hess, J.L. “The Problem of Three-Dimensional Lifting Potential Flow and Its Solution by Means of Surface
Singularity Distribution.” Computer Methods in Applied Mechanics and Engineering, Volume 4, Number 3,
November 1974.
8) Betz, A. “Behavior of Vortex Systems.” Zeitschrift fuer Angewandte, Mathematik und Mechanik, Bd. XII, Nr.
3, 1932; also NACA TM 713.
9) Rossow, V.J. “On the Inviscid Rolled-Up Structure of Lift-Generated vortices.” Journal of Aircraft, Volume 10,
Number 11, November 1973.
10) Bilanin, A.J., and Donaldson, C.DuP. “Estimation of Velocities and Roll-Up in Aircraft Vortex Wakes.”
Journal of Aircraft, Volume 12, Number 7, July 1975.
11) Quackenbush, T.R.; Lam, C.-M.G.; Wachspress, D.A.; and Bliss, D.B. “Computational Analysis of High
Resolution Unsteady Airloads for Rotor Aeroacoustics.” NASA CR 194894, May 1994.
wing k I , iI wing set I

vmean wing tip wing/panel
I
wake origin rQ collocation
geometry point
wing wake X yes no yes yes (a) yes (a) (c)

geometry
wing display yes (a) yes yes no no no

geometry
sensor X yes no yes no no
wing/panel X no no no no yes (b) (c)

point
frame force and mean circulation airframe gust

moment circulation circ peaks flow field
wing wake yes no (d) yes yes yes yes (e)

geometry
wing display X X X X X X
geometry
sensor no yes yes yes no no
wing/panel X X X X no no
point
Notes:
b) yes only if wing and panel of output and input are same
c) if straight lifting line, then quarter chord position from wing tips
d) yes if prescribed geometry
e) gust amplitude and origin: see gust component
Figure 27-6 Functionality of wing wake geometry component.

Chapter 28
ROTOR WAKE GEOMETRY COMPONENT
28–1 Description
A rotor wake geometry component calculates the wake geometry for a wing set. The wing set consists of all
blades of a rotor. This component incorporates the standard features of wake geometry components. The wake
geometry model can be rigid, prescribed, or free. The models implemented have the following features.
a) The rigid model calculates the wake geometry distortion from the mean interference velocity at the wing set.
b) The prescribed wake geometry is obtained from an empirical model for a hovering rotor.
c) The free wake geometry is obtained by calculation for a helicopter rotor. The free wake geometry is generated
only for the tip vortex of a single rotor (the wing set is the blades of one rotor); and it is only available for the trim
task.
The rigid and prescribed models described here are both considered “rigid wake geometry.” Two free wake methods
are implemented, from Scully and from Johnson. For all the models of this component, the variation of the wind
velocity in a ground boundary layer is neglected.
Figure 1 illustrates the functionality of the rotor wake geometry component.
Component Input
I
a) Aerodynamic interfaces at lifting line: location rQ at left wing tip, right wing tip, and collocation points (quarter
chord).
b) Wake geometry (from this component).
c) Wing frame coordinates: wing set plane normal k I , and longitudinal unit vector iI (S or T frame).
d) Position wing set origin: rSI/I or rT I/I .
e) Component input for rigid geometry:
I
1) Average interference velocity at wing set: vmean .
f) Component input for prescribed geometry:
1) Wing set force and moment (S or T axes).
2) Reference wing bound circulation peaks (mean): magnitude (Γmax , ΓL , ΓR ) and position (rmax , rL , rR )
g) Component input for free geometry:
1) Wing set velocity (S or T axes).
2) Reference wing bound circulation peaks: magnitude (Γmax , ΓL , ΓR ) and position (rmax , rL , rR )
414 ROTOR WAKE GEOMETRY COMPONENT
aero interfaces at lifting line aero interfaces at lifting line

location wing tips and QC location coll points
(QC or 3QC)
wing frame normal

wing frame long vector
wake geometry
position wing set origin
average int velocity
mean bound circ peaks
display geometry
wing set velocity
bound circulation peaks sensors
coning angle
normal to ground
height origin above ground
Figure 28-1 Functionality of rotor wake geometry component.

ROTOR WAKE GEOMETRY COMPONENT 415
3) Reference wing coning angle.
4) Normal to ground, kg (S or T axes).
5) Height origin of wing set above ground level.
Component Output
a) Aerodynamic interfaces at lifting line: location rPI of collocation points (quarter chord and three-quarter chord).
I
b) Wake geometry: rW .
c) Display geometry: rW , in S or T axes.
d) Sensor: wake geometry parameters.
28-3.1 Character of Rotor Wake Geometry
was created. This geometry is distorted by the wake self-induced velocity. Relative the rotor disk, the wake is
convected downward (normal to the disk plane) by the mean induced velocity and free stream, and aft in forward
flight by the inplane component of the free stream. The self-induced velocity produces substantial distortion of
the vortex filaments as they are convected. The wake geometry thus consists of distorted, interlocking helices, one
behind each blade, skewed aft in forward flight.
The distorted wake geometry exhibits an overall pattern in which the edges of the wake arising from the
rotor disk roll up to form vortices, as behind a circular wing (figure 2). In fact this wake consists of the helical
tip vortices from individual blades. The consequence of this pattern is that near the rotor the tip vortices tend to
move upward on the sides of the disk, and tend to move downward in the middle of the disk. Generally then the
self-induced distortion moves the tip vortices closer to the blades on the advancing and retreating sides (compared
to a rigid geometry), thereby increasing the blade-vortex interaction loads.
The free wake geometry has a large influence on the blade airloading at low speed (advance ratios below
about 0.20–0.25). At higher speeds, the propelling rotor has a large tip-path-plane angle of attack, typically several
degrees forward to provide the propulsive force. In such flight conditions the wake is convected downward relative
to the disk by the normal component of the free stream, and the distorted geometry is less important.
For a rotor, only the vortex from the blade tip rolls up significantly. A rolled up vortex from the root of the
blade is seldom observed. The wake components can optionally analyze a configuration with rollup of only the
right tip vortex, which is appropriate for a counter-clockwise rotating rotor. Hence the tip vortex geometry is only
required for the right tip; the left tip vortex position is set equal to the position of the left edge of the inboard sheet
(and not used). For clockwise rotation of the rotor, the convention used is that the span stations are still ordered
from root to tip (now from right to left), and the “right” and “left” rolled up vortices are still at the tip and root. So
the definition of “right” and “left” have been reversed.
The component uses velocity and forces in wing set (shaft) axes S, or wing tip (tip-plane) axes T, depending
Figure 28-2 Calculated rotor free wake geometry (CT /σ = 0.087 and μ = 0.18).
on the definition of the input variables. Wing tip axes should be used if possible, but wing set axes are probably
acceptable. The general convention is that the origin of the axes is at the center of the wing set, with the x–y plane
the wing set “plane”. The z-axis is up (positive lift direction), the x-axis is aft (positive velocity direction), and
the y-axis is to the right. These conventions affect the calculated wake geometry in several ways, depending on
the model used:
a) Rigid geometry: definition of normal and inplane (if different empirical factors are used).
b) Prescribed geometry: definition of vertical convection and thrust.
c) Free geometry: definition of axes of distortion.
d) Display geometry: selection of axes for transformed output.
It is assumed that the normal and thrust are in the z-axis direction of the axes used (S or T). The orientation of the
axes is the same for both clockwise and counter-clockwise rotation of the blades. The direction of rotation must be
specified however for the free geometry model. Dimensionless quantities are used in the analysis for the prescribed
and free models, based on the air density ρ, rotor radius R, and rotor rotational speed Ω (from a specified period;
the absolute value is used if the rotational speed is negative). The tip speed is ΩR, and the disk area is A = πR2 .
a) Wing frame coordinates: wing set plane normal and longitudinal unit vectors (k I and iI ), from a reference
frame component.
b) Position wing set origin: rSI/I or rT I/I , from a frame component. This quantity is the reference position for
the display geometry.
I
c) Average interference velocity at wing set: perturbation velocity of the air vmean , in inertial axes; obtained by
average over the wing surface, summation over all wings, and then summation of all contributions. The average
over the wing surface can be performed by a reference frame component, and the summation over all wings by a
differential equations component. Other components require that this velocity be divided into induced (from this
I
wing set) and interference (from all other sources) terms. Finally vmean can be calculated by using a reference
frame component to sum all contributions and transform to inertial axes. What contributions are included defines
the interaction between wing sets that is accounted for in the rigid geometry. A filter component can be used to
get the mean velocities.
d) Wing set force and moment: the total aerodynamic load (force F and moment M ) acting on the entire wing set,
e) Wing set velocity: the wing set velocity relative the air v, in S or T axes. Typically this is the velocity of the
origin of the wing set frame, from a reference frame component.
f) Reference wing bound circulation peaks: peak bound circulation values along the span, and the corresponding
span station of the peaks (left, right, and maximum peaks). The prescribed geometry requires the mean circulation;
the average over all wings in the wing set could be used, instead of reference wing values. The free geometry
requires the time varying circulation; the analysis assumes that all wings have the same loading. These models
require the circulation peaks over the span of the entire wing.
g) Reference wing coning angle: quantity β0 , obtained from the appropriate sensor of a structural dynamic
component.
h) Normal to ground: kg , in S or T axes, for ground effect in the Johnson free wake method; from a reference
frame component.
i) Height origin of wing set above ground level: zg , for ground effect in the Johnson free wake method; from a
reference frame component.
I
A filter component can be used to get the mean input for the wake geometry model. The locations rQ of the
I I
wing, and the bound circulation peaks for the free geometry should not be filtered. A filtered k , i , or kg will not
be a unit vector.
28-3.4 Wake Geometry Models
For all the models of this component, the variation of the wind velocity in a ground boundary layer is
neglected. The rigid wake geometry is calculated by assuming that the wake elements are all convected by the
I
average interference velocity vmean .
The prescribed wake geometry combines several models in a single functional form: rigid (convection by the
normal interference velocity); empirical for a hovering rotor (from the rotor loading); and input. The distortion is
calculated by assuming that the geometry is described by two-stage vertical convection and exponential spanwise
contraction. The effects of turn rate or transient motion of the rotor are included in the basic undistorted geometry,
but not in the prescribed distortion. Interference from other aerodynamic components is not considered.
The free wake geometry is calculated only for the tip vortex of a single rotor. The rigid or prescribed geometry
model is used for the distortion of the inboard sheet (at the sheet edges, and at the wing span stations). The tip
vortex core radius is constant, with the Scully vorticity distribution. The effects of turn rate of the rotor are included
in the basic undistorted geometry, but not in the free distortion. Interference from other aerodynamic components
is not considered. For efficiency, the distortion calculation assumes periodicity of the wake geometry, hence this
model can only be used in the trim task. The free wake geometry distortion is calculated for the entire rotor period
at the first time step. Then the saved distortion is used at subsequent time steps in the period. The distortion
is calculated using an internally generated time and wake age increment (typically 15 degrees in azimuth). The
distortion is interpolated or extrapolated as required to the times of the solution procedure, and the wake ages of
the component. Two free distortion methods are implemented, from Scully (ref. 3) and from Johnson (a simplified
version of the wing wake geometry component). The Scully method is applicable only in forward flight; the
Johnson method can be used in both forward flight and hover.
28–4 Theory
28-4.1 Rigid Distortion
The rigid distortion is calculated by assuming that the wake elements are all convected by the average
I
interference velocity vmean . This velocity is the sum and average of all sources of induced and interference
velocities acting on this wing set. This uniform convection gives
DI (t, τ ) = vmean
I
τ
Empirical corrections factors for the normal (fn ) and inplane (fi ) convection are introduced, with k I (S or T frame)
defining the normal direction. Thus
I
DI (t, τ ) = fn kk T + fi (I − kk T ) vmean I
τ = vconv τ
I
is the distortion, from the convection velocity vconv .
28-4.2 Prescribed Distortion
The prescribed distortion is calculated by assuming that the geometry is described by two-stage vertical
convection and exponential spanwise contraction. In terms of the dimensionless age φ = |Ω|τ , the vertical and
spanwise distortion (scaled with radius R) are:
0
K1 φ φ < φ1
Dz =
K1 φ1 + K2 (φ − φ1 ) φ > φ1
Dr = (1 − e−K3 φ )(1 − K4 )
where φ1 = 2π/N is the age at the encounter with the following blade (N is the number of blades). For the inside
edge of the inboard sheet, Dr is multiplied by the root cutout ratio rroot . This geometry is steady relative to the
moving wing, so Dz and Dr are functions of only wake age φ, not time. The vertical convection is defined by
K1 and K2 , the rates before and after encountering the first following blade. These are dimensionless velocities,
scaled with ΩR. The direction of the vertical convection is assumed to be the z-axis of the tip-plane axes: kTI
(although the component might use shaft axes S instead). The spanwise contraction is defined by K3 and K4 ,
the rate of contraction and the maximum contraction ratio respectively. The contraction time constant is 1/K3 Ω,
and K4 is the dimensionless distance, scaled with R. The direction of the contraction is assumed to be the wing
span direction, calculated from the locations of the blade root and tip (identified as the left and right wing tips,
regardless of the rotation direction):
I
rRT − rLT
I
jI =
|rRT
I − rLT
I |
evaluated at t − τ . Combining the vertical convection and spanwise contraction, the dimensional distortion is thus

DI (t, τ ) = R − Dz k I − Dr j I
The distortion is required for the tip vortex and the edges of the inboard sheet. The convection velocity is taken as
I
vconv = −(K2T ΩR)k I
for use in extrapolating the wake geometry. The following models are implemented for the prescribed geometry:
rigid, empirical, and input. The input model simply uses specified values of the constants K1 , K2 , K3 , and K4 .
28-4.2.1 Prescribed Model: Rigid
The rigid model assumes that the wake elements are all convected vertically by the normal component of the
average interference velocity:
−(k I )T vmean
I
λ=
ΩR
Including empirical factors, the vertical convection constants are thus:
K1 = f1 λ
K2 = f2 λ
Input values of the contraction constants K3 and K4 are used.
28-4.2.2 Prescribed Model: Empirical
The empirical models implemented were developed for hovering rotors from experimental model rotor flow-
visualization data. These geometries depend on the rotor loading, in terms of the thrust coefficient:
T
CT =
ρA(ΩR)2
where T is the z-axis component of the wing set force. Alternatively, an equivalent thrust coefficient can be
calculated from the peak bound circulation:
N Γmax
CT =
ΩR2 fΓ
where fΓ is an input factor. The models used were originally developed in terms of the rotor thrust, but variations
in planform can perhaps be better accommodated using the peak bound circulation. The empirical models also
depend on the number of blades N , the rotor solidity ratio σ, and the linear twist rate θtw (in degrees). Landgrebe
(ref. 1) developed a prescribed wake model of the above form. The vertical convection constants are, for the tip
vortex:
K1 = 0.25(CT /σ + 0.001θtw )

K2 = fK2 (1 + 0.01θtw ) CT
for the outside sheet edge:
K1 = 1.55 CT

K2 = 1.90 CT
and for the inside sheet edge:
K1 = 0

K2 = −(0.0025θtw
2
+ 0.099θtw ) CT
The spanwise contraction parameters are
K3 = 0.145 + 27CT
K4 = 0.78
for the tip vortex and the inboard sheet. Kocurek and Tangler (ref. 2) revised the tip vortex geometry based on
experimental data including low aspect-ratio, two-bladed rotors, obtaining:
m
CT
K1 = B + C
Nn
" 1/m #1/2
B
K2 = fK2 CT − N n −
C

K3 = 4.0 CT
where
B = 0.000729θtw
C = 2.3 − 0.206θtw
m = 1.0 − 0.25e0.04θtw
n = 0.5 − 0.0172θtw
The other constants are the same as for the Landgrebe model. These two prescribed models are based on flow-
visualization results, in which typically only four spirals of the wake are observed. Hence such experiments do not
determine the tip vortex vertical convection in the far wake, which has a significant influence on the calculated rotor
performance. It is therefore necessary to calibrate the wake model, in terms of measured and calculated power. For
this purpose, a factor fK2 has been introduced in the expressions for the tip vortex far-wake convection constant
K2 . Limited correlation suggests a values of fK2 = 0.9 at moderate to high thrust. These models generally define
well (compared with experiment) the radial and vertical position of the tip vortex when it first encounters the
following blade, hence greatly improve the prediction of the spanwise loading distribution. In addition, the factor
fΓ can be used to calibrate the model when the peak bound circulation is used. Note that a constant spanwise
bound circulation would give fΓ = 2π, and a bound circulation proportional to span station would give fΓ = 3π.
It is found that typically fΓ = 11–12 for 4 to 5 blades, and fΓ = 9–10 for two-bladed rotors.
28-4.3 Scully Free Distortion
The free distortion can be calculated using the method developed by Scully (ref. 3) for the wake geometry
of a single rotor in steady state flight. This method gives good results for performance and airloads at advance
ratios of 0.125 and above. The wake model of reference 3 consists of line segments for the tip vortices, and for the
inboard shed and trailed wake. It is assumed that the blade bound circulation has the same sign along the entire
span. The distortion of the tip vortex relative the basic helical geometry is calculated for a wake element of age τ ,
created by the wing at time δ = t − τ : DS (δ, τ ). A tip-path plane coordinate frame is used, with the x-axis to the
right (advancing side), the y-axis aft, and the z-axis down. The analysis is dimensionless, dealing with azimuth
angle rather than time.
The procedure for calculating the wake geometry consists of integrating the induced velocity at each wake
element. The outer loop in the calculation is an iteration on the wake age τ . The induced velocities q(δ) are
calculated at all wake elements for a given age φ, and all azimuth angles δ. Then the increment in the distortion
as the wake age increases by Δt is:
DS (δ, τ ) = DS (δ, τ − Δt) + Δt q(δ)
An efficient calculation of the wake geometry requires many variations on this basic procedure. Reference 3 adopted
the near-wake/far-wake scheme for reducing the computation. The first time the induced velocity is evaluated at
a point in the wake, the contributions from all wake elements must be found. For subsequent evaluations of the
induced velocity at that point, only the induced velocity produced by nearby wake elements are recalculated.
The other major consideration for minimizing the computation is the matter of updating the induced velocity
calculation. At a given point in the wake geometry calculation, there is a boundary in the wake between the
distorted geometry and the initial, rigid geometry. The distortion has been calculated between the rotor disk and
the boundary; downstream of the boundary the wake is undistorted. As time increases by Δt, the entire wake
is convected downstream, and the rotor blades move forward by Δt, adding new trailed and shed vorticity at the
beginning of the wake. If there were no distortion of the wake during the time Δt, then the induced velocity
at a given wake element would not change except for the contributions from the newly created wake vorticity
just behind the blade. Thus the normal calculation procedure consists of calculating the induced velocity at the
boundary, by just adding at each step the contribution from the new wake directly behind the blade. Of course,
the wake does distort as it is convected and as the estimate of the distortion improves, thus it is necessary to
update the calculation of the induced velocity in the wake. In boundary updating (induced velocity updating), the
induced velocity is calculated at the boundary still, by summing the contributions from all elements in the wake.
In general updating (distortion updating), the induced velocity is recalculated at all points in the wake upstream
of the boundary. Boundary updating is typically done every 90◦ on the front and rear portions of the helices, and
every 45◦ along the sides where the distortion is greater. General updating is typically done every 180◦ . General
updating can not be done often if the amount of computation is to be kept low, but it does improve the accuracy and
convergence. Numerous techniques for secondary improvements in the efficiency and accuracy are also included.
The distorted wake geometry is required for m revolutions, where m decreases with forward speed. A single
iteration of the free wake analysis consists of calculating the distortion DS (δ, τ ) for δ = 0 to 360◦ , and τ = 0 to
m360◦ . Usually two iterations are sufficient to obtain the converged solution for the wake geometry.
The analysis of reference 3 calculates the scaled distortion as a function of δ and τ , in tip-path plane axes:
6D positive right
Sx
DS (δ, τ ) = DSy positive aft
DSz positive down
For a clockwise rotating rotor, the sign of Dx must be changed. This distortion is stored as a function of t and τ ,
in wing tip axes T:
⎛ ⎞ ⎛ ⎞
DxT DSy
⎜ ⎟ ⎜ ⎟
DT (t, τ ) = ⎝ DyT ⎠ = ⎝ DSx ⎠
DzT −DSz
which is still scaled with R. The time and age increment in this solution (Δt = Δψ/Ω) is internally defined. The
distortion is calculated for the reference blade (m = N ), at discrete points:
D( , k) = D(t , τk )
where t = Δt ( = 1 to J) and τk = k Δt (k = 1 to K). The distortion is periodic in t, so
Dm ( m , k) = D( = m + mJ/N, k)
is the distortion for the m-th blade (m = 1 to N ). To calculate the wake geometry rW at times defined by the
solution procedure and wake ages defined by the component, the stored distortion is interpolated and extrapolated
to the required t and τ as required. Extrapolation uses the convection velocity vconv , obtained from the average
distortion increment at τlast :
vconv 1 1
= D(j, klast ) − D(j − 1, klast − 1)
ΩR J j
Δψ
(in tip-path plane axes). Optionally, the convection velocity from the rigid or prescribed model can be used to
extrapolate the free distortion. Then

DI (t, τ ) = R DxT iIT + DyT jTI + DzT kTI
is the free distortion in inertial axes.
To calculate the free wake geometry, the analysis requires the following quantities. From the velocity of the
wing set relative the air v, the dimensionless advance ratio components are:
⎛ ⎞
−μx
vT
= ⎝ μy ⎠
ΩR
μz
The analysis uses μx and μz . The inflow ratio λ = λi + μz is used to initialize the wake geometry, with
−(k I )T vmean
I
λi =
ΩR
The dimensionless bound circulation as a function of azimuth is:
Γmax (t)
Γ(ψ) =
ΩR2
where ψ = Ωt + ψ0 , and ψ = j Δψ (j = 0 to J). For a dual peak wake model, ΓR is used instead of Γmax . The
coning angle β0 is used to define the blade position relative to the tip-path plane. The number of azimuth steps per
revolution J is calculated from the number of blades N . It is required that J be a multiple of N , and the maximum
value is 30. The analysis was developed for an azimuth increment of 15◦ . Thus the following values for J and
Δψ are used.
N J Δψ (deg) N J Δψ (deg)
1 24 15 9 27 13.33
2 24 15 10 30 12
3 24 15 11 22 16.36
4 24 15 12 24 15
5 25 14.4 13 26 13.85
6 24 15 14 28 12.86
7 28 12.86 15 30 12
8 24 15 16–30 N 22.5–12
28-4.4 Johnson Free Distortion
The free distortion can be calculated using the method developed for the wing wake geometry component.
This method gives good results for performance and airloads in both forward flight and hover. For efficiency in the
rotor wake geometry component, the method is simplified to a case consistent with the limitations of the Scully
method: the distortion is calculated for just the tip vortices of a single rotor in steady state flight. The wake model
consists of line segments for the tip vortices, and for the inboard shed and trailed wake. It is assumed that the
blade bound circulation has the same sign along the entire span. The distortion of the tip vortex relative the basic
helical geometry is calculated for a wake element of age τ at time t: DJ (t, τ ). A tip-path plane coordinate frame is
used, with the x-axis aft, the y-axis to the right (advancing side), and the z-axis up. The analysis is dimensionless,
dealing with azimuth angle rather than time.
For efficiency, the near-wake/far-wake scheme is used, and the induced velocities are not completely updated
every time step. In full updating, the induced velocity is calculated at every collocation point along the tip
vortex (and the transitions between near-wake and far-wake are determined). In near wake updating, the near-
wake induced velocity is calculated at every collocation point, and added to the stored far-wake induced velocity.
Otherwise the induced velocity is calculated by just adding the contribution from the new wake created as the
blade moves during Δt. Typically full updating is done every 180◦ and near wake updating every 45◦ of azimuth
(dimensionless time). Updating every time step for ages up to the encounter with the next blade gives good results
for performance and airloads at advance ratios of 0.125 and above. Good performance results at lower advance
ratios can be obtained using appropriate parameters (and more computation time).
The advance ratio components μx and μz , inflow ratio λ, dimensionless bound circulation, and coning β0 are
calculated as for the Scully method. For ground effect, the dimensionless height above the ground is z = zg /R. The
rigid or prescribed model defines the geometry of the inboard sheet. The number of azimuth steps per revolution
J is calculated from the number of blades N , as for the Scully method. A single iteration of the free wake analysis
consists of a number of revolutions in time equal to the number of revolutions of distorted geometry calculated.
Usually two iterations are sufficient to obtain the converged solution for the wake geometry.
28-4.5 Comparison of Algorithms
The differences between the algorithms used by the Johnson and Scully methods affect both accuracy and
efficiency of the wake geometry calculation. The Scully method is faster. For comparable results, the Johnson
method requires about four times the computation time as the Scully method. Using appropriate parameters, the
Johnson method gives good results for both forward flight and hover; while the Scully method gives good results
for performance and airloads at advance ratios of 0.125 and above. Key aspects of the algorithm used by the
Johnson method are as follows.
a) Loops: Integrate velocity in time t (for several periods), for each vortex element of age
τ = 0 to m revolutions; calculate velocity at τ from all wake at t.
b) Velocity update for efficiency: full, near, or no update.
c) Features: distortion and distortion increment relaxation; propagation and extrapolation of
distortion.
d) Initial convection: wake leaves wing tangent to the wing surface.
Key aspects of the algorithm used by the Scully method are as follows.
a) Loops: Overall iteration; then for boundary between rigid and free distortion (from τM = 0
to m revolutions) and for each vortex element of age τ = 0 to τm ; integrate velocity in time t
(one period); calculate velocity at τ from all wake at t.
b) Velocity update for efficiency: general, boundary, or no update.
c) Features: velocity relaxation; propagation and extrapolation of distortion.
The fundamental difference is the order of the time integration and wake age loops in the two methods. Figure
3 illustrates the differences in the integration algorithms. The Johnson and Scully methods use the same guiding
concepts in the velocity update for efficiency, relaxation, and propagation and extrapolation of distortion. The
implementations of these features are different however, because the top level loops are different.
28–5 References
1) Landgrebe, A.J. “The Wake Geometry of a Hovering Helicopter Rotor and its Influence on Rotor Performance.”
Journal of the American Helicopter Society, Volume 17, Number 4, October 1972.
2) Kocurek, J.D., and Tangler, J.L. “A Prescribed Wake Lifting Surface Hover Performance Analysis.” Journal of
the American Helicopter Society, Volume 22, Number 1, January 1977.
3) Scully, M.P. “Computation of Helicopter Rotor Wake Geometry and its Influence on Rotor Harmonic Airloads.”
JOHNSON METHOD SCULLY METHOD
age
time wake
τM τ
wing
integration
in time
time wake element created
Figure 28-3 Integration algorithms of Johnson and Scully methods.

wing k I , iI wing set I

vmean wing tip wing/panel
I
wake origin rQ collocation
geometry point
wing wake X yes no yes yes (a) yes (a) (c)

geometry
wing display yes (a) yes yes no no no

geometry
sensor X yes no yes no no
wing/panel X no no no no (c) yes (b) (c)

point
force and mean velocity circulation coning ground

moment circulation kg , zg
wing wake no (d) no (d) yes yes yes yes

geometry
wing display X X X X X X
geometry
sensor yes yes yes no yes no
wing/panel X X X X X X
point
Notes:
b) yes only if wing and panel of output and input are same
c) if straight lifting line, then quarter chord position from wing tips
d) yes if prescribed geometry
Figure 28-3 Functionality of rotor wake geometry component.

Chapter 29
WING PERFORMANCE COMPONENT
29–1 Description
A wing performance component calculates performance quantities for a wing set. The component is intended
for the trim task, using the mean (filtered) values of the component input. Applications in the transient or flutter
tasks will require the appropriate filters of the input, and interpretation of the results.
Figure 1 illustrates the functionality of the wing performance component.
Component Input
a) Wing set force and moment.
b) Wing set velocity.
c) Total wing set power (Pind , Pint , Po ).
d) Average induced velocity at wing set (vind ).
e) Average interference velocity at wing set (vint ).
Component Output
a) Sensors.
The velocity and forces of the component can be in (a) wing set axes S; (b) wing tip axes T. The component
input can be identified as being in the S or T axes (the component frame is not used). The component output is
in the same axes as the component input. The general convention is that the origin of the axes is at the center of
the wing set, with the x–y plane the wing set “plane”. The z-axis is up (positive lift direction), the x-axis is aft
(positive velocity direction), and the y-axis is to the right. These conventions are used to name and label the output
quantities.
428 WING PERFORMANCE COMPONENT
wing set force and moment sensors

wing set velocity velocity
total wng set power force and moment
average ind and int velocity power
performance indices
Figure 29-1 Functionality of wing performance component.

WING PERFORMANCE COMPONENT 429
The summation and transformation can be performed by a reference frame component. A filter component can be
used to get the mean loads.
b) Wing set velocity: the wing set velocity relative the air v, in S or T axes. Typically this is the velocity of the
origin of the wing set frame, from a reference frame component. A filter component can be used to get the mean
velocity.
c) Total wing set power: induced, interference, and profile power terms (Pind , Pint , Po ), obtained by integration
over the wing surface and then summation over all wings. The wing model generates the power for one wing.
The summation over all wings can be performed by a differential equations component. A filter component can
be used to get the mean power.
d) Average induced and interference velocity at wing set: perturbation velocities of the air vind and vint , in S or T
axes; obtained by average over the wing surface and then summation over all wings. The average over the wing
surface can be performed by a reference frame component, and the summation over all wings by a differential
equations component. A filter component can be used to get the mean velocities.

29-3.4 Sensors
a) velocity;
b) force and moment;
c) power;
d) performance indices.
degrees, horsepower, or Mach number. Automatic scaling can also be applied to produce values in wing coefficient
form. The sensor scales used are as follows:
quantity coefficient
angle deg
velocity V
force qS
moment qSc (pitch) or qSb (yaw and roll)
power qSV
where the reference parameters are the air density ρ; and the wing area S, chord c, and span b. The reference
velocity is the magnitude of the wing set velocity, V = |v|, and q = 1/2ρV 2 .
29–4 Theory
29-4.1 Ideal Induced Velocity
The ideal induced velocity is calculated from the minimum induced drag, Di = L2 /πqb2 . The equivalent
inflow is
Di L
λideal = =
L πqb2
where b is the wing span. The velocity and dynamic pressure are obtained from the wing set velocity: q = 1/2ρ|v|2 .
The lift L is the component of the wing set force perpendicular to the wing set velocity. Then the ideal induced
velocity is videal = λideal |v|.
29-4.2 Sensors
The following sensor quantities can be defined as component output. Each quantity is a vector consisting of
a number of performance parameters.
29-4.2.1 Velocity
a) Velocities. The output is the vector of the velocity of the air relative the wing set. From the wing set velocity
v, the output is: ⎛ ⎞
−Vx
v = ⎝ Vy ⎠
Vz
Thus Vx is positive aft (forward flight), Vy is positive from the right, and Vz is positive down (climb).
b) Angles. The angle-of-attack of the x–y plane relative the air velocity, and the sideslip angle, are:
−Vz
α = tan−1
Vx
Vy
β = sin−1
|v|
(the angle of attack is positive for aft tilt of the plane).
c) Reference velocity. The magnitude of wing velocity is Vmag = |V |.
29-4.2.2 Force and Moment
a) Force and moment. The output is the vector of the aerodynamic force and moment acting on the wing set. From
the wing set force F and moment M , the output is:
⎛ ⎞
Fx
F = ⎝ Fy ⎠
Fz
⎛ ⎞
Mx
M = ⎝ My ⎠
Mz
b) Wind axis forces. Let D be the drag force of the wing set, defined as the component of the wing set force F
in the direction of the wing set velocity v. Let L be the lift force of the wing set, defined as the component of the
wing set force F perpendicular to the wing set velocity v. Thus:
D = −(v/|v|)T F

L = |F |2 − D2
c) Angles. The pitch and roll angles of the forces relative the z-axis are:
Fx
φX = tan−1
Fz
Fy
φY = tan−1
Fz
D
φD = tan−1
L
29-4.2.3 Power
a) Drag power. The work required by the wing drag force gives: Pdrag = |v|D = −v T F .
b) Profile and induced power. The total wing set power (component input) gives the quantities Pind , Pint , and Pos
(subscript “os” indicating wing section integration), which are included in this sensor vector. The profile power is
also calculated as
Po = Pdrag − Pind − Pint
c) Ideal power. The minimum induced power loss is calculated from the ideal induced velocity videal : Pm =
videal Fz .
29-4.2.4 Performance Indices
a) Induced and interference velocities. The output is the vectors of the average induced and interference velocity
at wing set, vind and vint (component input). The mean inflow ratios (velocity magnitudes) are
vm ind = |vind |
vm int = |vint |
The ideal induced velocity is videal . Equivalent inflow ratios are defined as
ve ind = Pind /L
ve int = Pint /L
The excess induced loss is measured by the ratios
1 Pind Dind
= = πb2
eind Pm (L/q)2
1 Pint Dint
= = πb2
eint Pm (L/q)2
where Pm is the minimum induced power loss.

b) Wing efficiency measures. The wing drag gives Pdrag = |v|D. So the contributions to the drag are defined as
follows:
Do = Po /|v|
Dind = Pind /|v|
Dint = Pint /|v|
The corresponding equivalent drag areas are Do /q, Dind /q, and Dint /q, where q = 1/2ρ|v|2 . Finally,
L
(L/D)w =
D
is the wing lift-to-drag ratio.

velocity force power vind , vint
velocity yes X X X
force yes yes X X
power yes yes yes X
performance yes yes yes yes

indices
Figure 29-2 Functionality of wing performance component.

Chapter 30
ROTOR PERFORMANCE COMPONENT
30–1 Description
A rotor performance component calculates performance quantities for a wing set. The wing set consists of all
blades of a rotor. The component is intended for the trim task, using the mean (filtered) values of the component
input. Applications in the transient or flutter tasks will require the appropriate filters of the input, and interpretation
of the results. Performance measures appropriate for a ducted fan, autogyro, wind turbine, or fixed wing are also
considered.
Figure 1 illustrates the functionality of the rotor performance component.
Component Input
a) Wing set force and moment (S or T axes).
b) Duct force and moment (S or T axes).
c) Wing set velocity (S or T axes).
d) Total wing set power (Pind , Pint , Po ).
e) Average induced velocity at wing set (vind , S or T axes).
f) Average interference velocity at wing set (vint , S or T axes).
g) Shaft torque.
h) Power into rotor.
i) Coning angle.
j) Tip plane rotation vector.
k) Controls.
Component Output
a) Sensors.
The velocity and forces of the component can be in the following axes: (a) wing set (shaft) axes S; (b) wing
tip (tip-plane) axes T; (c) control plane axes C. The component input can be identified as being in the S or T axes
(the component frame is not used). The component output can be requested in the S, T, or C axes. The general
convention is that the origin of the axes is at the center of the wing set, with the x–y plane the wing set “plane”.
436 ROTOR PERFORMANCE COMPONENT

duct force and moment sensors
wing set velocity
total wng set power velocity
average ind and int velocity force and moment
shaft torque power
power into rotor performance indices
coning angle trim quantities
tip plane rotation vector
controls
(for coll and cyclic)
Figure 30-1 Functionality of rotor performance component.

ROTOR PERFORMANCE COMPONENT 437
The z-axis is up (positive lift direction), the x-axis is aft (positive velocity direction), and the y-axis is to the right.
These conventions are used to name and label the output quantities. Note that the orientation of the axes is the
same for both clockwise and counter-clockwise rotation of the blades.
The orientation of the T axes relative S is defined by the tip plane rotation vector β:
⎛ ⎞
βx
β T S = ⎝ βy ⎠
βz
(component input). The elements of β are rotations about the S axes, which do not depend on the direction of
rotation of the blades. If the x–y plane is truly the wing set plane, then βz is probably zero. General relations
(including βz ) are used to transform the forces and velocities. However, the sensors that measure the tip-plane
orientation ignore βz , and do account for the direction of rotation. The orientation of the C axes relative S is
defined by the control plane rotation vector δ:
⎛ ⎞
δx
δ CS = ⎝ δy ⎠
0
The rotor pitch control is obtained from a gain times a specified element of a control vector:
a) Collective pitch: δ0 = G0 c0
b) Lateral cyclic pitch: δx = Gx cx
c) Longitudinal cyclic pitch: δy = Gy cy
(component input). The elements of δ are rotations about the S axes, which do not depend on the direction of
rotation of the blades. It is assumed that the x–y plane is the wing set plane, so δz is zero. However, the sensors
that measure the control plane orientation do account for the direction of rotation.
The summation and transformation can be performed by a reference frame component. A filter component can be
used to get the mean loads.
b) Duct force and moment: the aerodynamic load (force F and moment M ) acting on the duct, with the same
axes and moment origin as the wing set force and moment. The duct loads can be calculated by an aerodynamic
component, and the transformation to the required axes can be performed by a reference frame component. A
filter component can be used to get the mean loads.
c) Wing set velocity: the wing set velocity relative the air v, in S or T axes. Typically this is the velocity of the
origin of the wing set frame, from a reference frame component. A filter component can be used to get the mean
velocity.
d) Total wing set power: induced, interference, and profile power terms (Pind , Pint , Po ), obtained by integration
over the wing surface and then summation over all wings. The wing model generates the power for one wing.
The summation over all wings can be performed by a differential equations component. A filter component can
be used to get the mean power.
e) Average induced and interference velocity at wing set: perturbation velocities of the air vind and vint , in S or T
axes; obtained by average over the wing surface and then summation over all wings. The average over the wing
surface can be performed by a reference frame component, and the summation over all wings by a differential
equations component. A filter component can be used to get the mean velocities.
f) Shaft torque, power to rotor, coning angle: quantities Q, P , and β0 , obtained from the appropriate sensors of a
structural dynamic component. A filter component can be used to get the mean values.

30-3.4 Controls
A set of controls is available to the component. These vectors are component input, for connection to a system
input piece or to an input/output interface. The component can use the controls for rotor collective and cyclic
pitch. A particular control vector and element can be used once, more than once, or not at all.
30-3.5 Sensors
a) velocity;
b) force and moment;
c) power;
d) performance indices;
e) trim quantities.
The velocities and loads can be obtained in wing set, wing tip, and control plane axes. The value of any sensor
can be multiplied by a scale factor. Automatic scaling produces as appropriate a value in degrees, horsepower,
or Mach number. Automatic scaling can also be applied to produce values in rotor coefficient form or rotor
coefficient/solidity form. The sensor scales used are as follows:
quantity coefficient coefficient/solidity

angle deg deg
velocity ΩR ΩR
force ρA(ΩR)2 ρA(ΩR)2 σ
moment ρA(ΩR)2 R ρA(ΩR)2 Rσ
power ρA(ΩR)3 ρA(ΩR)3 σ
where the reference parameters are the air density ρ; the rotor radius R (A = πR2 ); the rotor solidity ratio σ; and
the rotor rotation speed Ω, obtained from a specified period.
30–4 Theory
30-4.1 Definition of Axes T and C
Wing tip axes T are defined by a rotation from the wing set axes S. The T axes have the same origin as the S
axes. The orientation of the T axes relative S is defined by the tip plane rotation vector β:
⎛ ⎞
βx
β T S = ⎝ βy ⎠
βz
(component input). The elements of the rotation vector are interpreted in one of the following ways.
a) Aircraft Euler angles: C T S = Xβx Yβy Zβz

b) Rodrigues parameters: C T S = C(p), p = β
c) Arbitrary Euler angles: C T S = Uβx Vβy Wβz , for rotations U V W = ±XY Z
d) Small angles: C T S = I − β
Control plane axes C are defined by a rotation from the wing set axes S. The C axes have the same origin as the S
axes. The orientation of the C axes relative S is defined by the control rotation vector δ:
⎛ ⎞
δx
δ CS = ⎝ δy ⎠
0
The elements of the rotation vector are interpreted in one of the following ways.
a) Aircraft Euler angles: C CS = Xδx Yδy Zδz

b) Rodrigues parameters: C CS = C(p), p = δ
c) Arbitrary Euler angles: C CS = Uδx Vδy Wδz , for rotations U V W = ±XY Z
d) Small angles: C CS = I − δ
30-4.2 Ideal Induced Velocity
The ideal induced velocity is calculated from the momentum theory expression:
CT
λideal =
2 λ2 + μ2
using the wing set force F and velocity v (in the input axes). The advance ratio is:
⎛ ⎞
−μx
v
= ⎝ μy ⎠
ΩR
μz

and then μ = μ2x + μ2y , λ = λideal + μz . The thrust coefficient is
T kT F
CT = 2
=
ρA(ΩR) ρA(ΩR)2
where k is the unit vector along the z-axis. If μ is zero, the equation for λideal can be solved analytically. Otherwise,
for non-axial flow, the equation is written as follows:
sλ2h
λ= + μz
λ2 + μ2
where λ2h = |CT |/2 (λh is always positive) and s = sign CT . A Newton-Raphson solution for λ gives:
2
1in = sλh
λ
λ2n + μ2
λn − μz − λ1in
λn+1 = λn − f
1in λn /(λ2 + μ2 )
1+λ n
A relaxation factor of f = 0.5 is used to improve convergence. To eliminate the singularity of the momentum
theory result at ideal autorotation, the expression
% &
.373μ2z + .598μ2
λ = μz − .991
λ2h
is used when
1.5μ2 + (2sμz + 3λh )2 < λ2h
Then the ideal induced velocity is videal = λideal ΩR, which is assumed to be in the direction of the z-axis.
30-4.3 Sensors
30-4.3.1 Velocity
a) Velocities. The output is the vector of the velocity of the air relative the wing set. From the wing set velocity v
(in S or T axes), the output is: ⎛ ⎞
−Vx
v = ⎝ Vy ⎠
Vz
Thus Vx is positive aft (forward flight), Vy is positive from the right, and Vz is positive down (climb). The velocity
in S, T, or C axes is obtained using the appropriate transformation.
b) Angles. The angle-of-attack of the x–y plane relative the air velocity is:
−Vz
α = tan−1
Vx2 + Vy2
(positive for aft tilt of the plane). The yaw angle is:
Vy
ψ = tan−1
Vx
The angles relative the S, T, or C axes are obtained by using the velocity in the specified axes.
c) Reference velocities. The rotor tip speed is obtained from the sensor reference parameters: Vtip = ΩR.
Assuming that the blades are rotating about the z-axis, the velocity seen by the blade tip is:
⎛ ⎞
Vx + Vtip sin ψ
⎝ Vy + Vtip cos ψ ⎠
Vz
which has a maximum magnitude (advancing tip velocity) of

4
2
Vmax = Vtip + Vx2 + Vy2 + Vz2
at azimuth tan ψ = Vx /Vy . The magnitude of the rotor velocity is Vmag = |V |. The quantities relative the S, T,
or C axes are obtained by using the velocity in the specified axes.
d) Tip-path plane and control plane tilt. The tip-path plane orientation relative some reference plane is defined by
the harmonics of a positive Fourier series representation of the tip motion:
β = β0 + βc cos ψ + βs sin ψ
where βc is the longitudinal tip-path plane tilt, and βs is the lateral tip-path plane tilt. The control plane orientation
relative some reference plane is defined by the harmonics of a positive Fourier series representation of the pitch
motion:
δ = δ0 + δc cos ψ + δs sin ψ
where δs is the longitudinal control plane tilt, and δc is the lateral control plane tilt. The orientation of the T and C
axes relative the S axes has been defined by the appropriate component input. Hence the tilt of the tip-path plane
and control plane relative the shaft axes S (hub plane) is:
βc = −βyT S
βs = βxT S
δs = δyCS
δc = δxCS
Figure 2 illustrates the geometry. The tilt relative the other frames follows from the relation δ CT = δ CS − β T S .
Hence the tilt of the tip-path plane relative the control axes C is:
βcC = (βc + δs )S
βsC = (βs − δc )S
and the tilt of the control plane relative the tip plane axes T is:
δsT = (βc + δs )S
δcT = −(βs − δc )S
Since the blade motion is being described by a Fourier series in the azimuth angle ψ, the signs of βs and δc must
be changed for a clockwise-rotating rotor. The coning angle β0 and collective pitch δ0 are obtained from the
appropriately defined component input.
C axes
βc θs
x S axes
y
y axis out
of paper T axes
T axes
-θ c βs
y S axes
x
x axis out
of paper C axes
Figure 30-2 Geometry of tip-path plane, control plane, and hub plane.
30-4.3.2 Force and Moment
a) Force and moment. The output is the vectors of the aerodynamic force and moment acting on the wing set.
From the wing set force F and moment M (in S or T axes), the output is:
⎛ ⎞
H
F =⎝Y ⎠
T
⎛ ⎞
Mx
M = ⎝ My ⎠
Mz
The loads in S, T, or C axes are obtained using the appropriate transformation.
b) Wind axis forces. Let X be the drag force of the wing set (negative for a propulsive force), defined as the
component of the wing set force F in the direction of the wing set velocity v. Let L be the lift force of the wing
set, defined as the component of the wing set force F perpendicular to the wing set velocity v. Thus:
X = −(v/|v|)T F

L = |F |2 − X 2
using F and v in the same axes.
c) Angles. The pitch and roll angles of the forces relative the z-axis are:
H
φH = tan−1
T
Y
φY = tan−1
T
X
φX = tan−1
L
The angles relative the S, T, or C axes are obtained by using the force in the specified axes.
d) Shaft torque. The shaft torque Q is obtained from the appropriately defined component input.
30-4.3.3 Power
Following an energy-balance analysis of rotor performance, the rotor power can be split according the type
of loss. The output consists of the following quantities.
a) Power absorbed. The power into the rotor P = Pin is obtained from the appropriately defined component input.
b) Climb plus parasite power. The work performed by the rotor propulsive force gives (Pc +Pp ) = −|v|X = v T F .
It is not possible to separate the climb and parasite losses until the entire system is considered.
c) Profile and induced power. The losses attributed to rotor lift and drag forces are:
(Po + Pi ) = P − (Pc + Pp )
where Pi includes both induced and interference losses. Note that if P = 0:

(Po + Pi ) = −(Pc + Pp ) = |v|X
The total wing set power (component input) gives the quantities Pind , Pint , and Pos (subscript “os” indicating
wing section integration), which are included in this sensor vector. The profile power is also calculated as
Po = (Po + Pi ) − Pind − Pint
d) Ideal power. The minimum induced power loss is calculated from the ideal induced velocity videal : Pm =
videal T . Then the ideal power is defined as
Pideal = minimum induced + interference + climb + parasite
= Pm + Pint + (Pc + Pp )
and
Pn = P − Pideal
= (Po + Pi ) − Pm − Pint
= profile + excess induced
is the nonideal power.
a) Induced and interference velocities. The output is the vectors of the average induced and interference velocity
at wing set, vind and vint (component input; in the input axes). The mean inflow ratios (velocity magnitudes) are
vm ind = |vind | vm int = |vint |
The ideal induced velocity is videal . Equivalent inflow ratios are defined as
ve ind = Pind /T ve int = Pint /T
The excess induced loss is measured by the ratios
κind = Pind /Pm κint = Pint /Pm
where Pm is the minimum induced power loss.
b) Profile power. The profile power can be expressed in terms of equivalent blade section drag coefficients:
CP o 1 8Po 1
cdo = 8 = 3
σ f ρA(ΩR) σ f
CP n 1 8Pn 1
cdn = 8 =
σ f ρA(ΩR)3 σ f
where Pn is the nonideal power. The factor f accounts for the increase of the blade section velocity with rotor
3
inplane and axial speed: CP o = 12 σcd U 3 dr, U 2 = u2T + u2R + u2P ; so (from Harris)
$ 2π $ 1
1 3/2
f =4 (r + μ sin ψ)2 + (μ cos ψ)2 + μ2z dr dψ
2π 0 0

∼ 5 2 3 2 4 + 7V 2 + 4V 4 9 μ4
= 1+V 2 1+ V + μ −
2 8 (1 + V 2 )2 16 1 + V 2
" √ #
3 4 3 2 2 9 4 1+V2+1
+ μ + μ μ + μ ln
2 z 2 z 16 V
with V 2 = μ2 + μ2z . This expression is exact when μ = 0; and f ∼ 4V 3 for large V .
c) Axial flow efficiency measures. The rotor figure of merit and propeller propulsive efficiency are:
Pm + Pint T (videal + ve int )
M= =
Po + P i Po + P i
P c + Pp −|v|X
η= =
P P
For hover and no interference loss, this figure of merit has the usual form:
3/2 √
T videal C / 2
Mhover = = T
P CP
Since −X is the propeller thrust, η is the usual definition of propulsive efficiency. A generalized figure of merit
for axial flow is:
Pideal Pn T (videal + ve int ) − X|v|
Ma = =1− =
P P P
which equals M at low speed and η at high speed.
d) Edgewise flow efficiency measures. The rotor equivalent drag is defined as

D = P/|v| + X = P − (Pc + Pp ) /|v| = (Po + Pi )/|v|
The contributions to D are thus defined as follows:

Do = Po /|v|
Dind = Pind /|v|
Dint = Pint /|v|
The corresponding equivalent drag areas are D/q, Do /q, Dind /q, and Dint /q, where q = 1/2ρ|v|2 . Finally,
L L|v|
(L/D)r = =
D Po + Pi
is the rotor lift-to-drag ratio.
e) Efficiency measures for autogyros and fixed wings. The lift and drag are the wind axis forces L and X. The
coefficients based on rotor disk area and the wind or flight speed are then CL = L/qA and CX = X/qA, where
q = 1/2ρ|v|2 . Wing coefficients based on the wing area S = Aσ are CLw = L/qS and CXw = X/qS. The wing
lift-to-drag ratio is L/X; which for zero rotor shaft power equals the rotor lift-to-drag ratio (L/D)r . The wing
induced efficiency e can be obtained from the lift and induced drag:
1 (L/q)2
e=
πb2 Di /q
using an appropriate wing span b.
f) Efficiency measures for wind turbines. The power coefficient based on the wind or flight speed is CP =
−P/qA|v|. A measure of wind turbine efficiency is then
CP CP
Mw = =
CP max 16/27
where CP max = 16/27 is the Betz limit on the power that can be extracted from the wind.
30-4.3.5 Trim Quantities
The trim quantities are a selected subset of the parameters available in the other sensor vectors. These
parameters are typically used as targets in the trim solution. The following quantities are selected:
a) Tip-path plane tilt: βc and βs (relative shaft axes S)

b) Aerodynamic forces acting on the rotor: H, Y , and T (shaft axes S)
c) Aerodynamic forces acting on the rotor: H, Y , and T (wing tip axes T)
d) Wind axis lift and drag: L and X
e) Power into rotor: Pin
30-4.4 Ducted Fan
For a ducted fan, the ideal induced velocity is calculated from the following momentum theory expression,
using the thrust coefficient calculated from the total load (rotor plus duct):

1 CT 1
λideal = + − fV z μz
fT 2 (fV z μz + λi )2 + (fV x μ)2 fT
where fT is the ratio of the rotor thrust to the total thrust. In the form
" #
1 CT sλ2h /fT μz
fV z μz + λi = + μz = +
fT 2 (fV z μz + λi )2 + (fV x μ)2 (fV z μz + λi )2 + (fV x μ)2 fT
it is possible to solve for λ using the free-rotor expressions given above, replacing λ2h , μz , μ, λ with respectively
λ2h /fT , μz /fT , fV x μ, fV z μz + λi . The sensor quantities are calculated using the total loads (rotor plus duct),
except for the minimum induced power loss and the equivalent inflow ratios:
Pm = videal Trotor
ve ind = Pind /Trotor

ve int = Pint /Trotor
which use just the rotor thrust.

shaft power coning tip plane controls

torque into rotor rotation
velocity X X yes yes yes
force yes X X no (a) no (b)
power X yes X X X
performance X yes X X X
indices
trim X yes X yes X
velocity force power vind

duct force vint
velocity yes X X X
force yes yes X X
power yes yes yes X
performance yes yes yes yes

indices
trim yes yes X X
Notes:
a) yes if need transformation between S and T axes
b) yes if need C axes
Figure 30-3 Functionality of rotor performance component.

Chapter 31
ROTORCRAFT PERFORMANCE COMPONENT
31–1 Description
A rotorcraft performance component calculates performance quantities for a system. The system is a rotorcraft.
The component is intended for the trim task, using the mean (filtered) values of the component input. Applications
in the transient or flutter tasks will require the appropriate filters of the input, and interpretation of the results.
Several of the quantities calculated depend on the power absorbed by the rotors, P . If P does not account for all
the power sources, then these parameters will not be meaningful. Typically, only the rotor shaft power is included
in P .
Figure 1 illustrates the functionality of the rotorcraft performance component.
Component Input
a) Total rotorcraft power.
Component Output
a) Sensors.
31-3.1 Operating Condition
The system performance quantities are calculated using information from a specified operating condition:
a) System velocity relative the air: V

b) Climb velocity: Vc
A system gross weight W must also be specified. If these parameters are not correctly defined, some of the
performance quantities will not be meaningful.
The total rotorcraft power is obtained from a power sensor of a rotor performance component (with no scale
factors), summed over all rotors of the system. The elements in the sensor vector are as follows:
a) Power absorbed: P = Pin

b) Climb plus parasite power: (Pc + Pp )
450 ROTORCRAFT PERFORMANCE COMPONENT
total rotorcraft power sensors

power
performance indices
trim quantities
Figure 31-1 Functionality of rotorcraft performance component.

ROTORCRAFT PERFORMANCE COMPONENT 451
c) Profile and induced power: (Po + Pi ), Pind , Pint , Pos , Po

d) Minimum induced, ideal, nonideal power: Pm , Pideal , Pn
The summation over all rotors can be performed by a differential equations component.

31-3.4 Sensors
a) power;
b) performance indices;
c) trim quantities.
The value of any sensor can be multiplied by a scale factor. Automatic scaling produces as appropriate a value
in horsepower. Automatic scaling can also be applied to produce values in rotor coefficient form or rotor coeffi-
cient/solidity form. The sensor scales used are as follows:
quantity coefficient coefficient/solidity

3
power ρA(ΩR) ρA(ΩR)3 σ
where the reference parameters are the air density ρ; the rotor radius R (A = πR2 ); the rotor solidity ratio σ; and
the rotor rotation speed Ω, obtained from a specified period.
31–4 Theory
31-4.1 Sensors
31-4.1.1 Power
Following an energy-balance analysis of rotor performance, the rotorcraft power can be split according the
type of loss. The output consists of quantities (a-d) obtained from the total rotorcraft power (component input,
which has been summed over all rotors in the system), and derived quantities (e-g), as follows.
a) Power absorbed: the power into the rotorcraft is P = Pin .

b) Climb plus parasite power: the work performed by the rotor propulsive force is (Pc + Pp ).
c) Profile and induced power: the terms obtained by integration over the wing section are Pind , Pint , and Pos .
d) Ideal power: the minimum induced power loss is Pm .
e) Profile and induced power. The losses attributed to rotor lift and drag forces are:
(Po + Pi ) = P − (Pc + Pp )
where Pi includes both induced and interference losses. The profile power is also calculated as
Po = (Po + Pi ) − Pind − Pint
f) Climb and parasite power. The climb power is defined as Pc = Vc W , using the specified operating condition
(for Vc ) and gross weight. Then
Pp = (Pc + Pp ) − Pc
is the parasite power.
g) Ideal power. The ideal rotorcraft power is defined as
Pideal = minimum induced + climb + parasite

= Pm + (Pc + Pp )
= (Pideal )rotors − Pint
since the ideal power of the rotors includes the interference losses; and
Pn = P − Pideal
= (Po + Pi ) − Pm
= profile + excess induced + interference
is the nonideal power.
a) Axial flow efficiency measures. The rotorcraft figure of merit is:
Pm
M=
Po + P i
A generalized figure of merit for axial flow is:
Pideal Pn
Ma = =1−
P P
which equals M at low speed.

ROTORCRAFT PERFORMANCE COMPONENT 453
b) Edgewise flow efficiency measures. The rotor equivalent drag is defined as D = (Po + Pi )/|V |, using the
system velocity from the specified operating condition. The corresponding equivalent drag area is D/q, where
q = 1/2ρV 2 . Finally the rotor lift-to-drag ratio is defined as
W
(L/D)r =
D
using the specified gross weight W .
c) Forward flight efficiency measure. The rotorcraft lift-to-drag ratio is defined as L/D = W V /P , using the
specified gross weight W .
31-4.1.3 Trim Quantities
The trim quantities are a selected subset of the parameters available in the other sensor vectors. These
parameters are typically used as targets in the trim solution. The following quantities are selected:
a) Power into rotorcraft: Pin

power
sensor yes
Figure 31-2 Functionality of rotorcraft performance component.

Chapter 32
RIGID WING COMPONENT
32–1 Description
A rigid wing component combines a rigid body component and a lifting line wing component. Figure 1
illustrates the combination. The intent of this component is to trade configuration flexibility for computational
efficiency. A rigid wing component is a structural dynamic component. It has no elastic motion. The component
has an internal aerodynamic model. Hence the calculations of velocity, position, and force on the wing are internal
to the component, rather than being implemented as interfaces.
The rigid body component combines the functionality of the rigid body component and the lifting line wing
component. The following variables are from the rigid body component:
Degrees of Freedom
The component can have up to two degree-of-freedom vectors: rigid motion, and joint motion.
Component Input
a) Structural dynamic interface: the input is force and moment for a complete interface; or torque for a torque
interface.
b) Aerodynamic interface: the input can be force, moment, interference velocity, or gust velocity.
c) Controls: for use by joints, actuators, or applied loads.
Component Output
a) Structural dynamic interface: the output is axes motion for a complete interface; or rotation for a torque interface.
b) Aerodynamic interface: the output can be the velocity relative to the air, or the position relative the origin of
the axes.
c) Sensors.
The following variables are from the lifting line wing component:
Degrees of Freedom
The component can have one degree of freedom vector for the unsteady loads model, and one vector for the
dynamic stall model.
Component Input
F F
a) Aerodynamic interfaces at collocation points: induced velocity (vind ) and interference velocity (vint ).
456 RIGID WING COMPONENT
velocity relative air

RIGID position relative wing frame LIFTING
BODY LINE
WING
aerodynamic force
Figure 32-1 Rigid wing as combination of rigid body and lifting line wing.
RIGID WING COMPONENT 457
b) Controls: for use by trailing edge flap.

Component Output
b) Aerodynamic interfaces at collocation points: aerodynamic force and moment, at trailing-edge flap hinge.
g) Sensors.
The component frame is also the wing frame. Using the standard conventions of a structural dynamic
component, connections must be defined at the quarter chord and three-quarter chord collocation points (panel
midpoints) along the wing span. These connections are defined by a constant displacement relative the location
(z CJ ; not at a joint), and constant displacement of the location relative the origin of the rigid body axes (z EB ). They
are defined with no rotation relative the body axes (C CJ = I and C EB = I). The lifting line wing component
uses the following aerodynamic interfaces at the collocation points:
a) Velocity of air, at quarter chord and three-quarter chord (from a structural dynamic compo-
nent).
b) Position relative origin of wing frame, at quarter chord and three-quarter chord (from a
structural dynamic component).
c) Aerodynamic force, at quarter chord and three-quarter chord (to a structural dynamic com-
ponent); or aerodynamic force and moment, at quarter chord.
The rigid wing component does not need these interfaces, since it calculates and uses the quantities internally as
required.
The component can efficiently calculate the velocity and position of the wing, since the motion of the body
relative the frame or the frame relative inertial space (BF or FI) is the same for all collocation points; and the
position of the connections relative the body axes (CB = CJ + EB) is constant. Moreover, the motion and resulting
forces are evaluated simultaneously at all collocation points, so common calculations need not be repeated. In
order to internally calculate the velocity of the air at the collocation points, the rigid wing component must have the
appropriate interference velocities and gust velocities available as standard aerodynamic interfaces (component
input). These velocities are those that would be used by some structural dynamic component to provide the velocity
of the air for a lifting line wing component. Note that as a wing component, the same interference velocities are
F
required as separate input (to calculate the power): summed as induced velocity (vind ) and interference velocity
F
(vint ), and transformed to the wing frame F.
32–4 Theory
Following the conventions of the aerodynamic interface of a standard structural dynamic component, the rigid
wing component must calculate following quantities at the collocation points (connections C):
a) velocity relative air (v R , q, v̇ R );

b) position relative origin of axes (rR and ṙR ).
The axes R are the component or wing frame axes (R=F). The velocity relative to the air is calculated using all the
interference and gust velocities (component input) defined at this connection. The calculations are simplified by
the restrictions placed on the definition of the connections at the collocation points. The component rigid motion
(BF, FI, BI) is the same for all collocation points. The position of the connection C, relative the origin of the axes
F, is calculated as follows:
rF = xCF/F = C F B z CB/B + xBF/F

BF/B z CB/B )
ṙF = C F C v CF/C = C F B (v BF/B + ω
from the rigid motion BF, and the constant position
z CB/B = z CJ/J + z EB/B
of the connection relative the origin of the body axes (C CB = I).
The velocity of the connection C relative to the air, in axes F, is calculated as follows:
v F = vB
F
− vW
F
− F
vG − F
vA
v̇ F = v̇B
F
− v̇W
F
q = 1/2ρ(v F )2
where the first term in each equation is the motion of the connection relative to the inertial frame, and remaining
terms are the motion of the air relative to the inertial frame (wind, gust, and aerodynamic interference). Note that
time derivatives of the interference and gust velocities are not considered. The required terms are calculated as
follows. The motion of the connection C relative to the inertial frame I, in F axes is:
F
vB BI/B z CB/B )
= C F C v CI/C = C F B (v BI/B + ω
F
v̇B = C F C (v̇ CI/C + ω BI/B z CB/B + ω
CF/C v CI/C ) = C F B (v̇ BI/B + ω̇ BF/B v CI/C )
from the rigid motion BF and BI, and the constant position CB. The wind at connection C relative to the inertial
frame I, in F axes is:
F
vW = C F I vWI
F
v̇W = −
ω F I/F C F I vW
I
I
from the rigid motion FI, and the wind velocity vW . With a ground boundary layer in the wind model, the velocity
I CI/I IB CB/B
vW is obtained at the position x =C z + xBI/I (and the resulting time variation of vWI
is neglected
R
in calculating v̇W ). This wind term is the same for all collocation points if there is no ground boundary layer. A
gust at connection C (perturbation of the air velocity relative to the inertial frame) in F axes is:
F Q
vG = C F Q vG
RIGID WING COMPONENT 459
from the rigid motion FQ, and the gust velocity in Q axes (component input). An aerodynamic interference at
connection C (perturbation of the air velocity relative to the inertial frame) in F axes is:
F Q
vA = C F Q vA
from the rigid motion FQ, and the interference velocity in Q axes (component input). For best efficiency, the gust
and interference velocities should be input in the wing frame axes (Q=F)
The standard structural dynamic component evaluates the reactions (total force and moment) for the equations
of motion and for sensors. The rigid wing component must add to this reaction the internally calculated aerodynamic
forces at all the quarter chord and three-quarter chord collocation points (or force and moment at the quarter chord).
These forces are in the F axes, acting at connections. The equations of motion require the force and moment on
the body, in B axes. Thus the aerodynamic forces must be multiplied by C BF . The sensor quantity is the total
load acting on Q, about a reference R, in R axes. Q can be a connection, a joint, a location, or the body origin
(Q=CJEB). The reference R is a connection (R=C), or the origin of Q. Thus the aerodynamic forces must be
multiplied by C BF ; and the moment arm of the aerodynamic force about R is:
xCC/C = 0 R = connection
xCE/E = z CJ/J R = location
xCB/B = z CJ/J + z EB/B R = body
since the C, E, and B axes are identical for the connection of the aerodynamic force; and the aerodynamic force is
not at a joint.
β, ξ SD torque aero control aero panel aero

ξ vind , vint loads
motion (a) yes (e) yes (g) yes yes X X
SD 0 (b) no (f) no (i) (f) no (f) X X
torque no (c) no no (i) no X X
aero 0–2 no no (h) no no X X

(d) (j)
sensor 2 (j) (a) (a) (a) yes X X
aero 2 X no (k) yes 1 (l) no no

motion
aero 2 X no (k) yes yes no X

loads
wing 2 X no (k) yes yes yes X

sensor
Γ X X X X X X yes
Γpeak X X X X X X yes
force 0 X X X X X yes
power X X X X X yes yes
panel X X X X X X yes
Fh
“SD” means a complete structural dynamic interface; “torque” means a torque

structural dynamic interface; “aero” means an aerodynamic interface.
Notes:
a) as for rigid body component
b) no if joint degree of freedom and interface not at joint
c) yes if joint degree of freedom; no frame dependence
d) depending on interface kind
e) no if joint equation and interface not at joint
f) yes if full residual
g) no if joint equation
h) yes if vA or vG for v, and at same connection
i) yes if at controlled joint, or joint degree of freedom with actuator
j) frame dependence may actually be no (not implemented)
k) yes if vA or vG
l) second order for some dynamic stall models
Figure 32-2 Functionality of rigid wing component.

Chapter 33
HELICOPTER TAIL BOOM COMPONENT
33–1 Description
A helicopter tail boom component calculates the aerodynamic forces acting on a subsystem consisting of a
circulation-control boom and a reaction jet. This subsystem is intended to provide torque reaction and yaw control
for a single main-rotor helicopter. The circulation-control boom operates in the wake of the main rotor. The
aerodynamic model of the boom is lifting-line theory. Lifting-line theory assumes that the boom has a high aspect
ratio, which allows the problem to be split into separate wing and wake models. This component solves the wing
problem, and includes an approximate wake model based on ideal-wing theory. The reaction jet is modelled as a
force with specified direction and magnitude. Component degrees of freedom may be needed to account for the
dynamics of the internal aerodynamics of the circulation-control boom and reaction jet. The component equations
can be static, or first-order differential equations.
The boom is defined by collocation points on a line. The total number of collocation points equals the
number of aerodynamic panels on the boom, plus one for the reaction jet. The component interfaces are described
in terms of the velocity and force on the boom line (discretized). Note that the geometry is actually calculated
on the structural dynamic side of the interfaces with the component. Lifting-line theory deals with the section
aerodynamic environment. The required section geometry is obtained from the geometry of the collocation points.
The component analyzes the entire boom.
The component implements user-defined calculations. The basic version of this component is a simple model
of the tail boom. Other versions are constructed by modifying subroutines called by the component.
Figure 1 illustrates the functionality of the helicopter tail boom component.
Degrees of Freedom
The component can have one degree of freedom vector, for the first order differential equation model.
Component Input
1) Velocity of air.
2) Position relative origin of boom frame.
b) Control vectors.
462 HELICOPTER TAIL BOOM COMPONENT
degrees of
freedom
velocity of air aerodynamic force
position rel boom frame
total boom force and moment

control power required
sensors
Figure 33-1 Functionality of helicopter tail boom component.

HELICOPTER TAIL BOOM COMPONENT 463
Component Output
a) Aerodynamic interfaces at collocation points: aerodynamic force.
b) Total boom force and moment (in boom frame axes, about origin of boom frame).
c) Power required.
d) Sensors.
33-3.1 Geometry
The analysis is performed in the axes of the boom frame F. The boom frame should be the frame or parent
frame of all structural dynamic components composing the boom. The boom consists of a set of aerodynamic
panels, with a collocation point at each panel. In addition, there is one collocation point for the reaction jet. It is
assumed that the boom position and velocity relative the air (component input) are in the F axes, and the calculated
forces (component output) are thus also in the F axes. The velocity relative the air (v F ) includes interference
terms from other aerodynamic components, and gust terms. The location of the boom (rF ) is measured from the
origin of the boom frame F. The aerodynamic forces F F are obtained at each collocation point. These discretized
section loads are summed to obtain the total force and moment of the tail boom: FtF and MtF (about the origin of
F, using the position vectors rF ).
The boom geometry is described in terms of a span station r, which is a variable (often dimensionless)
that identifies the aerodynamic panels. This variable must be used consistently by the component, including its
definition in tables. There is be a constant scale factor R such that Rr is a true measure of distance along the
boom. In order to discretize the loads, the boom area is needed. It is assumed that the chord is defined such that
the panel width (including the factor R) times the panel chord gives the true area of the panel. This panel chord is
used to define the aerodynamic coefficients. The boom consists of K aerodynamic panels, aligned spanwise. The
panels are defined by the span stations of the edges: rEk , for k = 1 to K + 1. The collocation points are at the
panel midpoints:
1
rAk = (rE(k+1) + rEk )
2
for k = 1 to K. The corresponding panel width (dimensional) is
ΔrAk = (rE(k+1) − rEk ) R
If rAk were specified directly, it could not be assured that there is a set of panel edges with rAk as the midpoints.
The panel width is used to discretize the loads:
(panel load)k = (section load)k ΔrAk
In addition, collocation point (K + 1) is the point of action of the reaction jet.
The aerodynamic properties of the boom are defined parametrically, in terms of values at an arbitrary set of
span stations. The properties are evaluated at the panel midpoints (and wherever else they are required) assuming
linear variation between the specified points.
33-3.2 Airfoil Table
The component typically obtains the aerodynamic coefficient data from tables. The basic version uses a two
dimensional table, containing lift and drag coefficient data as a function of angle of attack and blowing coefficient.
User-defined calculations can make use of any table class and type.
The aerodynamic coefficients are defined in terms of some reference chord in the section, which defines how
angle of attack is measured; and some reference axis, which defines where the forces are applied. The boom section
geometry is defined by the positions of the collocation points along the boom line, and a pitch angle measured
relative the boom frame. It is assumed that this pitch angle corresponds to the reference chord as used by the
aerodynamic coefficients; and that the section forces are acting at the boom line.
33-3.3 Controls
A set of control vectors is available to the component. These vectors are component input, for connection to
a system input piece or to an input/output interface. The tail boom model can use a vector δ of input variables,
typically to determine the internal aerodynamics of the circulation-control boom and reaction jet. Each element
of δ is identified as an element of a control vector. A particular control vector and element can be used once, more
than once, or not at all.
33-3.4 Sensors
to a system output piece or to an input/output interface. The sensors are obtained from user-defined calculations.
Let ξ be the degree of freedom vector, x a component output vector, and v the tail boom input vector. A set
of control vectors f is available to the component; v is obtained from f . The static equations have the form
x = UB (v, f, t)
where UB is a user-defined function. The first order equations have the form
˙ ξ, v, f, t)
0 = UA (ξ,
˙ ξ, v, f, t)
x = UB (ξ,
where UA and UB are user-defined functions. All parameters and input (including the position and velocity at
collocation points) are available to evaluate UA and UB . User-defined routines are also required to calculate the
aerodynamic forces of the circulation-control boom and reaction jet.
An input parameter specifies the kind of user-defined calculations to be performed. Hence more than one
version can be implemented simultaneously. The basic version of the component is a simple model of the tail
boom. User-defined calculations are implemented by modifying subroutines that are called by the helicopter tail
boom component. The following subroutines are available to the user:
UHTBRD read parameters
UHTBWT write parameters
UHTBTB tables required
UHTBIN initialize
UHTBEN element names
UHTBA1 function UA (first order)
UHTBBP function UB (power)
UHTBBS function UB (sensor)
UHTBFC circulation-control boom force
UHTBFJ reaction jet force
Subroutine names and common names beginning with the letter “U” are reserved for the user. The exception is
that common UNITCM is not available to the user.
Arrays of component parameters are available. Input and print of these parameters can be in user-defined
form, but the data must be stored in these arrays for proper handling by the analysis. The user-defined calculations
can use tables. An input parameter specifies the number of tables required. An array of saved quantities can
be defined, for use during the component calculations. The analysis subroutines have an argument that signals
when the component is being executed in order to obtain equation matrices (by numerical perturbation). Thus
if necessary a special model can be implemented to produce the appropriate linearized matrices of a nonlinear
system. Utilities provide access to all component input vectors, and to all component parameters. Subroutines
implement calculations for section aerodynamics.
33–4 Theory
33-4.1 Boom Geometry
The boom geometry is defined by the position vectors xk at the collocation points of the boom line, for
aerodynamics panels at span stations rk , k = 1 to K. The position is measured from the origin of the boom frame
F, in F axes. Figure 2 illustrates the geometry. Positions on the boom line define a tangent vector tk . The section
chord vector ck is the portion of a nominal chord vector that is perpendicular to this tangent. The nominal chord
vector is in the y–z plane of the F frame, at a pitch angle θ relative the y-axis.
The aerodynamic analysis is performed in section axes X, defined by the chord, normal, and tangent vectors
shown in figure 2 (c, n, and t). The section plane is thus c–n. The tangent vector is
tk = xk+1 − xk−1
(the position at k is used for panels at the boom ends, where k − 1 or k + 1 does not exist). The chord vector is
⎛ ⎞
0
ck = (I − t Tt ) ⎝ − cos θ ⎠
sin θ
force n
force
t
reaction
k-1 k k+1 jet
c
boom line
y
x
boom frame
Figure 33-2 Boom geometry.

The normal vector is then nk =

ck tk . Thus the rotation matrix from the boom frame F to the section axes X is:
⎡ T⎤
c
XF ⎢ T⎥
C = ⎣ t ⎦
Tn
where is the appropriate unit vector. With this definition, the X axes have x chordwise (positive aft), y spanwise,
and z normal (positive up).
33-4.2 Section Motion and Loads
Figure 3 shows the section aerodynamic environment in section axes. The velocity of the wing relative the
air, at the boom line, is v F . So the velocity of the air relative the section, in X axes, is:
⎛ ⎞
ux
uX = ⎝ uy ⎠ = −C XF v F
uz
Then the angle of attack, resultant velocity, and Mach number are:
uz
α = tan−1
ux

U = u2x + u2z
M = U/cs
These quantities are used to obtain the quasistatic load, typically by using using airfoil tables. The loads on the
circulation-control boom also depend on the value of the blowing coefficient.
The aerodynamic analysis obtains the lift and drag forces acting on the section. The lift L and drag D act
in the x–z plane of the X axes, respectively normal and parallel to the resultant velocity U . The forces act at the
collocation point on the boom line. The total force acting on the section, in X axes, is:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
D D cos α − L sin α (Dux − Luz )/U
F X = Yα ⎝ 0 ⎠ = ⎝ 0 ⎠=⎝ 0 ⎠
L L cos α + D sin α (Lux + Duz )/U
Note that the components of F X are the chord and normal forces (cX and cN ). The lift and drag forces are
evaluated from the corresponding coefficients, defined as follows:
1 2
L= ρU c c
2
1
D = ρU 2 c cd
2
where ρ is the air density, c the wing chord, and U the resultant velocity at the section (always positive). The section
coefficients of the circulation-control boom depend on angle of attack, Mach number, and blowing coefficient:
c = c (α, M, Cμ )
cd = cd (α, M, Cμ )
The blowing coefficient Cμ is defined as follows:

1 2
μ= ρU c Cμ
2
α
D
x
U
Figure 33-3 Section aerodynamic environment (section axes).

where μ = ṁV is the section momentum through the boom slots. In general Cμ depends on the component
degrees of freedom and controls. The sign conventions in reverse flow must be consistent with the definition of
angle of attack in the coefficient data. Reverse flow occurs when ux is negative. Then positive pitch or negative
uz gives an angle of attack near −180 degrees. The convention is that the lift coefficient is then positive (as for
testing in a wind tunnel). The drag coefficient is always positive. The transformation to X axes above then gives
a negative z-axis force from the lift, and a negative x-axis force from the drag.
The aerodynamic analysis provides the section force vector F X (load per unit length), acting on the boom
line. The discretized load acting at the collocation point, in the boom frame axes, is thus:
F F = C F X F X ΔrA
where ΔrA is the panel width.
33-4.3 Induced Angle of Attack
The component can use an approximate wake model based on ideal-wing theory. The ideal-wing result for
the induced angle of attack in terms of the total lift is:
CL
αi =
πeAR
where AR = b2 /S is the wing aspect-ratio. The empirical correction factor e < 1 is included to account for
non-ideal losses. Considering the tail boom as a wing, this result is used here in perturbation form, giving a section
angle of attack change from the section lift coefficient. Thus the section loads are evaluated using the effective
angle of attack αe = α − αi in place of α (for both normal and reverse flow). Since the induced angle of attack
depends on the unknown lift, an iterative solution is required:
αiold = αi
c
α1i =
πeAR
αi = λ1αi + (1 − λ)αiold
c = c (α − αi )
starting with αi = 0 and c = 0. The convergence criterion is based on the change in angle of attack:
|αi − αiold | < max(0.01, α)
where is a tolerance. A relaxation factor λ < 1 is introduced to ensure convergence of the iteration for small
aspect ratio. The iteration involves a fixed-point operator αi = G(αi ) where
c (α − αi )
G=λ + (1 − λ)αi
πeAR
cα
G = 1 − λ 1 +
πeAR
Without the relaxation factor, the iteration will thus diverge for AR < c α /πe. The ideal value of the relaxation
factor (for G = 0) is
AR
λ=
AR + c α /πe
Finally, αe = α − αi is used in place of α to transform the wind axis lift and drag to the section axes force F X .
33-4.4 Reaction Jet Load
The reaction jet produces a force F F , acting on the (K + 1) collocation point at rF . A reference direction
in the boom frame is defined by the following rotations:
a) rotation about the positive z-axis by azimuth angle ψj ;

b) rotation about the positive y-axis by elevation angle θj ;
c) rotation about the positive x-axis by cant angle φj .
Then the positive y-axis gives the reference direction of the reaction jet:
C JF = Xφ Yθ Zψ
⎛ ⎞ ⎛ ⎞
0 cos θ sin ψ
FJ ⎝ ⎠
uFj =C 1 = ⎝ cos φ cos ψ + sin φ sin θ sin ψ ⎠
0 − sin φ cos ψ + cos φ sin θ sin ψ
In general the magnitude and direction of this force depend on the component degrees of freedom and controls.
33-4.5 Total Loads
The total force and moment on the helicopter tail boom are obtained by summing the panel and jet reaction
loads. The loads about the origin of the boom frame F, in F axes, are thus:
FtF = FF
MtF = rF F F
where rF is the position of the collocation point relative the origin of the boom frame. The summation is over all
collocation points, both boom panels and jet reaction.
33-4.6 Simple Aerodynamic Model
The basic version of this component is a simple model of the tail boom aerodynamics. Other versions are
constructed by modifying subroutines called by the component, to implement user-defined calculations. The
aerodynamic analysis must calculate the section lift and drag coefficients, as a function of angle of attack, Mach
number, and blowing coefficient. In general, the blowing coefficient is obtained from the momentum μ:
1 2
μ= ρU c Cμ
2
This model obtains the blowing coefficient from a control variable gtb :
Cμ = Ktb gtb
where Ktb is a constant defined for each span station. Then the lift and drag coefficients are evaluated, either
from tables or from equations. A two dimensional table gives the coefficients as a function of angle of attack and
blowing coefficient: c (α, Cμ ) and cd (α, Cμ ). Alternatively, the following equations can be used:
c = c αα + c μ
Cμ
cd = cdo
where c α , c μ and cdo are constants defined for each span station. The reaction jet direction is assumed to be
given by uFj . The magnitude of the jet force is obtained from a control variable gtj . Then
F F = ftj uF F
j = Ktj gtj uj
where Ktj is a constant. The power required is calculated as a linear combination of the blowing and jet control
variables: P = Kpb gtb + Kpj gtj . The variables gtb and gtj determine the magnitudes of the boom and jet forces.
They can be obtained directly from δtb and δtb of the tail boom input vector δ:

gtb δtb
g= = =δ
gtj δtj
Alternatively, a first order differential equation can be used to calculate g = ξ from δ:
τ ξ˙ + ξ = δ
The time constant τ represents the dynamics of the internal aerodynamics of the circulation-control boom and
reaction jet.
ξ input
motion 1 yes
output 0 yes
Figure 33-4 Functionality of helicopter tail boom component.

Chapter 34
COMPUTATIONAL FLUID DYNAMICS COMPONENT
34–1 Description
A computational fluid dynamics component calculates the aerodynamics of a set of objects. The user provides
the CFD analysis, by modifying subroutines called by the component. A general interface with the CFD analysis is
implemented. The aerodynamic system consists of wings and other objects. The basic version of this component
is an internal model that can be used to test the component, and is also used if the component is being perturbed.
For the basic version, the wings are modelled as in the lifting line wing component (without states, dynamic stall,
or prescribed coefficient increments). The CFD analysis can be viscous or inviscid. For the inviscid case, the wing
drag is obtained from the lifting line model.
There may be other components corresponding to the same aerodynamic objects. Which component gives a
nonzero load then depends on the specified model, and perhaps on the loop solution level. The following models
can be used: zero, basic, or computational fluid dynamics. This component produces zero loads if the model is
“zero”, or in the flutter task; the other aerodynamic components must produce nonzero loads. For the trim task,
the model can be specified for each level of a designated wake loop.
Figure 1 illustrates the functionality of the computational fluid dynamics component.
Component Input
Wings:
1) Velocity of air, at quarter chord and three-quarter chord (v F , q, v̇ F ).
2) Velocity of air at quarter chord from three-quarter chord induced velocity (for second-order
lifting-line theory).
3) Position relative origin of wing frame, at quarter chord and three-quarter chord (rF , ṙF ).
F
4) Induced velocity (vind ).
F
5) Interference velocity (vint ).
6) Displacement and Euler angle rotation relative wing frame, at quarter chord.
b) Controls: for use by trailing-edge flaps.
Other objects:
d) Aerodynamic interfaces at collocation points:
474 COMPUTATIONAL FLUID DYNAMICS COMPONENT
wings wings
velocity of air (QC, 3QC) aerodynamic force (QC, 3QC)

position rel wing frame (QC, 3QC) force and moment (flap hinge)
induced and interference velocity
displacement and rotation (QC) aerodynamic loads
bound circulation
control total wing force and moment
bound circulation peaks
total wing power
wing sensors
other objects other objects

position rel object frame aerodynamic force and moment
displacement and rotation
aerodynamic loads
total object force and moment
flow field point flow field point

position relative inertial frame velocity
Figure 34-1 Functionality of computational fluid dynamics component.

COMPUTATIONAL FLUID DYNAMICS COMPONENT 475
1) Position relative origin of object frame (rF , ṙF ).

2) Displacement and Euler angle rotation relative object frame.
e) Aerodynamic loads (from this component).
Flow field points:

f) Location of point (rI ).
Component Output
Wings:
b) Aerodynamic interfaces at collocation points:
F F
1) Aerodynamic force, at quarter chord and three-quarter chord (FQC , F3QC ), or aerodynamic force and
F F
moment, at quarter chord (FQC , MQC ).
2) Aerodynamic force and moment, at trailing-edge flap hinge (FhF , MhF ).
g) Wing sensors.
Other objects:
h) Aerodynamic loads: force and moment (F , M ), for all collocation points.
i) Aerodynamic interfaces at collocation points:
1) Aerodynamic force and moment (F F , M F ).
j) Total object force and moment (in object frame axes, about origin of object frame).
Flow field points:

k) Velocity at point (v I ).
34-3.1 General CFD Interface
A general interface with the CFD analysis is implemented. The aerodynamic system consists of a set of
objects (wings and other objects). Each object is defined by a number of collocation points. The density of these
collocation points must be sufficient to define the motion and loading for the coupled aeromechanic solution. This
density may be much less than that of the CFD grid. For each aerodynamic object considered, there must be a
convention for the collocation points that determines how the interface data are used, although this convention may
not affect the form of the interface. For a wing, the collocation points consist of the quarter chord and three-quarter
chord points on a series of spanwise panels; and the flap hinge for wings with trailing edge flaps.
The following information is exchanged between the computational fluid dynamics component and the CFD
analysis, at each collocation point.
CFD Analysis Input

a) Position and velocity (r, ṙ): the motion of the structure, relative the origin of the frame.
b) Displacement and Euler angles of rotation of the structure, relative the wing frame.
c) Wind velocity (without ground boundary layer).
d) Flap angle: displacement, velocity, and acceleration; only for wings with trailing edge flaps.
e) Position at flow field point (rI ).
CFD Analysis Output

a) Force and moment (F , M ): the load acting on the structure. For wings, the force is provided at the quarter
chord and three-quarter chord collocation points; or the force and moment at the quarter chord. For wings with
trailing edge flaps, the force and moment at the flap hinge are provided.
b) Velocity at flow field point (v I ).
The above input quantities are available to the CFD analysis during its solution procedure. The above output
quantities are available to the component as sensors, after the CFD analysis has solved for the flow field. The loads
are provided for each object. In addition, the CFD analysis can provide the velocity at a specified flow field point.
Lifting line wing calculations of the component (for the basic version, or an inviscid CFD analysis) also require
the velocity of the air, and the induced and interference velocities; these quantities are not required by the CFD
analysis. The lifting line wing model always provides the bound circulation.
The position and velocity provided to the CFD analysis can be in wing/object frame axes, or in inertial axes.
The displacement and Euler angles of rotation are in wing/object frame. The displacement is relative the origin
of the frame. The rotation is the motion of axes initially parallel to the frame axes. The Euler angles describe the
rotation matrix from the rest position as C = Xφ Yθ Zψ (or using any of the other possible orders of X, Y , and Z).
The force and moment provided by the CFD analysis can be in wing/object frame axes, or in inertial axes. The
position and velocity at a flow field point are in inertial axes.
The operations performed by the CFD analysis include input and initialization; solving the flow field; and
evaluation of sensor quantities. The “solve” operation obtains the aerodynamic solution at time t, from the motion
at time t. The trim or transient solution procedure and the computational fluid dynamics component define the
times at which loads are required from the CFD analysis. The first time step of the task is identified, so the CFD
analysis can perform special operations at the beginning of the task. Whether it is the trim task or the transient
task being performed is also identified for the CFD analysis. The CFD analysis solves the entire system of the
component when the first aerodynamic load is evaluated by the component.
The trim part (harmonic or time finite element method) obtains the periodic solution for t0 to t0 + T , where T
is the period. The transient part (integration method) obtains the solution from tB to tE . The numerical integration
by the transient part may involve an iteration at each time step. So there may be more than one call of the component
by the transient part, and of the CFD analysis by the component, at the current time t. The CFD analysis must
recognize repeated calls at the same time.
The “sensor” operation obtains the aerodynamic load, flow field velocity, or other quantity at time t, at or
before the latest time for which the CFD solution has been executed. Sensor quantities are calculated from the
stored solution, in the trim or transient task. Sensor quantities can be required by the component immediately
after the “solve” operation; during an iterative solution; or at the end of the task, when the system solution has
converged.
An input parameter specifies the kind of user-defined CFD analysis to be executed. Hence more than one
version can be implemented simultaneously. The basic version of the component is an internal model that can
be used to test the component, and is also used if the component is being perturbed. The user provides the
CFD analysis, by modifying subroutines called by the computational fluid dynamics component. The following
subroutines are available to the user:
CFDRDP read parameters
CFDWTP write parameters
CFDTBL tables required
CFDINT initialize
CFDQQQ compare parameters
CFDSOL solution
CFDSEN sensor
CFDFLW flow field sensor
Figure 2 shows the operation of the interface routines. Subroutine names and common names beginning with the
letter “U” are reserved for the user. The exception is that common UNITCM is not available to the user. Subroutine
names and common names beginning or ending with the letters “CFD” are also reserved for the user.
An array of component parameters is available. Input and print of these parameters can be in user-defined
form, but the data must be stored in this array for proper handling by the analysis. The user-defined calculations
can use tables. An input parameter specifies the number of tables required. An array of saved quantities can be
defined, for use during the component calculations.
34–4 Theory
A frame F is specified for the component input and output of each aerodynamic object. The component input
is the position and velocity at the collocation points, in F axes relative the origin of F (rF and ṙF ). The CFD
analysis requires the position and velocity at the collocation points, perhaps in inertial axes relative the origin of
the inertial frame (rI and ṙI ). Then from the frame motion (FI):
rI = rCI/I = C IF xCF/F + xF I/I

= C IF rF + xF I/I
F I/F xCF/F + v F I/F )

ṙI = ṙCI/I = C IF ẋCF/F + Ċ IF xCF/F + ẋF I/I = C IF (ẋCF/F + ω
F I/F rF + v F I/F )
= C IF (ṙF + ω
The CFD analysis produces the force and moment at the collocation points, perhaps in inertial axes. The component
output is the force and moment at the collocation points, in F axes:
F F = CF I F I
M F = CF I M I
from the frame motion (FI).

Computational Fluid Dynamics Component
Component Interface CFD

Routines Routines Analysis
input and CFDRDP input and

initialization CFDWTP initialization
CFDTBL
CFDINT
CFDQQQ
trim or CFDSOL solve flow field

transient (aerodynamic
task CFDMTN solution of entire
CFDFRM CFD input system at time t
from motion at t)
CFDSEN evaluate sensor

CFDFLW CFD output quantities
(at time t)
control
data
Figure 34-2 Operation of interface routines.

The CFD analysis can be viscous or inviscid. For the inviscid case, the wings are modelled as in the lifting
line component in order to obtain the section drag coefficient, and its contribution to the force at the collocation
point.
If the flap loads are ignored in the interface with the structure, then the CFD loads at the quarter chord and
three-quarter chord points must include the flap force and moment.
other input aero loads
aero loads yes X
other output yes yes
Figure 34-3 Functionality of computational fluid dynamics component.

Chapter 35
PLUGIN COMPONENT
35–1 Description
A plugin component allows software developers and users to add content to CAMRAD II. Plugins can add to
the functionality of both the shell and components. Shell plugins can be used independently of plugin components,
to extend the capability of the shell to construct core input. Plugin components can be used with core input entirely,
but a corresponding shell plugin is recommended, to make it easy to use the new component technology. The
plugin component can not be a structural dynamics component.
A plugin based on the rotor inflow component is available, as a demonstration and test of plugin development.
Each plugin is identified by a character string, the variable “KIND”. It is recommended that each plugin
adopt a three-letter code XXX, and use subroutine and common names beginning with “YXXX”. Most compilers
accept names with more than six characters. It is recommended that plugin developers register the KIND and XXX
character strings with Johnson Aeronautics, in order to avoid conflicts with other plugins.
35-2.1 Plugin Calculations
Plugin calculations are implemented by modifying subroutines that are called by the system and by the plugin
component. The shell plugin and plugin components work through the following subroutines:
start
YPLGHD print headers
YPLGIC initialize for case
shell plugin
YPLGSH run shell plugins
YPLGSP print input of shell plugins
plugin component
YPLGRD read input
YPLGWT write input
YPLGTB identify tables
YPLGBR initialize
YPLGBQ component information
YPLGEQ equation evaluation
YPLGWD vector description
482 PLUGIN COMPONENT
Subroutine names and common names beginning with the letter “Y” are reserved for the plugins. Names beginning
with “YPLG” should not be used, to avoid possible conflicts with future releases. To allow more than one plugin
to be installed, the subroutines YPLGXX should contain only a test of the variable kind, and a subroutine call for
each plugin. The subroutines YPLGXX perform the following functions:
Start
a) Subroutine YPLGHD: After the header at the start of a job, print one or more lines identifying the plugin.
b) Subroutine YPLGIC: At the start of each case, initialize the plugin.
Shell Plugin
c) Subroutine YPLGSH: After the rotorcraft shell creates the core input, run shell plugins.
d) Subroutine YPLGSW: After print of the rotorcraft shell input, print the input of the shell plugins.
Plugin Component
e) Subroutine YPLGRD: During core input, read the plugin input.
f) Subroutine YPLGWT: During print of the core input, print the plugin component input.
g) Subroutine YPLGTB: During input and initialization, identify the CAMRAD II tables required.
h) Subroutine YPLGBR: During system initialization, initialize the plugin component.
i) Subroutine YPLGBQ: During system initialization (after YPLGBR), provide component information about the
input, output, and degree of freedom vectors.
j) Subroutine YPLGEQ: During the solution process, evaluate the plugin component equations at time t: an equation
of motion vector or an output vector, depending on component degree of freedom vectors and component input
vectors.
k) Subroutine YPLGWD: Before write of output from the plugin component, provide a description of the degree
of freedom vector or the output vector.
These subroutines are contained in the file YPLG. Their functions are described in more detail below.
35-2.2 Shell Plugin
Subroutine YPLGSH is called if the shell CASE input variable “PLUGIN” is not zero. Shell plugins will
probably read additional input. This input is read after the standard shell input (after the NLDEF namelist with
“action = end of shell”), before the core input.
Shell plugins can use the utility subroutines GMSHLL, GMCORE, and PLGFCG; and common blocks containing
the shell and core input variables. If access to shell input variables is required, the utility GMSHLL is called to
load these variables into a common block. To create core input, first the utility GMCORE is called to move the
core input variables from storage into a common block (ACTION = 1); the variables are set to the default values if
the core piece does not exist. Next the shell plugin can define values of the core input variables. Then the utility
GMCORE is called to move the core input variables from the common block into storage (ACTION = 0). The utility
GMCORE can also be used to delete existing core input blocks (ACTION = -1). The utility PLGFCB can be used to
determine if a core input block exists.
Access to the shell and core input variables is through common blocks. The file YPLG contains at the end
all shell commons, all core commons, and all core component commons. For convenience the commons are in
subroutines, but these subroutines do nothing and are never called. The commons can be used directly, or as
PLUGIN COMPONENT 483
include files. Future releases may change the contents of these commons.
35-2.3 Core Component Input
Core input for the plugin component begins by reading the standard NLDEF–NLVAL namelists, with the
variable “KIND” specifying the plugin. Then YPLGRD can read input (after the NLVAL namelist of the plugin
component). All variables in the NLVAL namelist except “KIND” can be read or defined by the plugin component
in YPLGRD (these variables are input to YPLGRD as read by NLVAL, and on return from YPLGRD their values are
stored in the component input block).
Input for the plugin component includes an identification number (the variable “IDENT”), so there can be
more than one component created for each kind.
The standard input for plugin components (managed by the analysis) includes arrays of real and integer
parameters (“PRMR” and “PRMI”), arrays of names and labels (character string arrays “PRMN” and “PRML”), the
number of CAMRAD II tables required, and the number of record and list indices for which storage is allocated.
The parameter arrays are available to store the input, but the values are best read by YPLGRD (not in the NLVAL
namelist), and must be printed by YPLGWT.
The plugin can use CAMRAD II tables. The input variable “NTABLE” specifies the number of tables required.
The subroutine YPLGTB supplies the class, type, and name for each table required; and perhaps the file name or
logical name (default = name) and unit number (default from NFTABL). CAMRAD II tables are identified by an
index. The indices are stored in an array “TABLE”, in the order specified by subroutine YPLGTB. Tables are used
during calculations by calling subroutine TIXXXX to get the interpolated values of dependent variables (TI2STD,
TI3STD, TIASTD, TIAC81, or TIACAM). Files of other formats can be read directly during initialization.
35-2.4 Initialization
Subroutine YPLGIC should set a counter for the variable “IDENT”, which serves to distinguish different
components of the same kind.
System initialization calls the subroutine YPLGBR and then the subroutine YPLGBQ. YPLGBQ provides
information about the input, output, and degree of freedom vectors. The order that the vectors are described
defines the vector number. For each vector, integers “KINDD” and “DEFSEQ” are specified; these variables can be
used to identify the vector during the solution process.
35-2.5 Solution Process
During the solution process, the system calls the subroutine YPLGEQ to evaluate the plugin component
equations at time t: an equation of motion vector or an output vector, depending on component degree of freedom
vectors and component input vectors. Write of output from the plugin component is handled by the analysis
routines. The subroutine YPLGWD can provide a description of the degree of freedom vector or output vector.
The format for each line printed should be (14X,107A).
Plugin components can use CAMRAD II records (real or integer arrays) and lists, which store information
in system memory. These data structures are created during initialization and used during the solution process.
Utilities that deal with records and lists are described below. During the solution process (YPLGEQ, YPLGWD), the
484 PLUGIN COMPONENT
plugin component must not delete or change the size of any record or list that exists at the end of the initialization
(YPLGBR, YPLGBQ).
35-2.6 Utilities
Plugin components can use utility subroutines to create and access records (real or integer arrays) and lists in
the system memory. The utility routines put a real value and/or integers into a list (VPUTXX) or get them from a
list (VGETXX); put a name and a label into a list (NPUTNL) or get them from a list (NPUTNL); establish a record
(VERECS); or get the address of a record (VGETAD, VGETAR). A list or record is identified by an index; the index
points to the address of the structure in memory. The utility is called initially with INDEX = 0 to create the structure;
the utility returns the index, which must be stored for later access to the structure. The plugin component input
variable “NREC” is the length of an integer array that is created to store record and list indices.
Plugin components can use the utility subroutines PLGGPR and PLGGPI to create real and integer records, put
data to the record, or get data from the record. Alternatively, the plugin component can use the utilities VERECS
and VGETAD to directly access records.
Plugin components can use the utility subroutine PLGINP to evaluate an input vector; and use the utility
subroutine PLGDOF to evaluate a degree of freedom vector (displacement, velocity, or acceleration). The vector
is identified by number or by name; it is most efficient to use the vector number. These routines return the vector
value at a current or past time, in an array (which must be supplied by the calling routine, but may be a record) or
in a record (returning the record address).
Plugin components can use the utility subroutines PLGENA, PLGENS, PLGENL, PLGENP, PLGOWP, and
PLGWND to get information. PLGENA returns the analysis environment: units and analysis tasks. PLGENS
returns the solution environment: the program, case number, and task; whether the equations are currently being
perturbed in order to calculate linearized matrices (so if necessary a special model can be implemented to produce
the appropriate linearized matrices of a nonlinear system); the time step of the current trim or transient part solution.
PLGENL returns the current level of a trim or transient loop. PLGENP returns the physical environment: fluid
properties, gravity, and ground. PLGOWP returns the current value of a quantity from an operating condition,
wind, or period piece. PLGWND returns the wind vector.
The input and output of all these utility subroutines are described in the YPLG file.

CAMRADIIvolume II

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CAMRADIIvolume II

Uploaded by

Copyright:

Available Formats

CAMRAD II

COMPREHENSIVE ANALYTICAL MODEL OF

Volume II: Components Theory

Wayne Johnson Distributed by

Johnson Aeronautics AMI Aero, LLC

3. STRUCTURAL DYNAMIC COMPONENTS 11

4. RIGID BODY COMPONENT 39

5. LINEAR NORMAL MODES COMPONENT 41

5-2.1 Linear Normal Modes 41

6. FINITE ELEMENT BEAM COMPONENT 45

8–2 Component Variables 89

9. REFERENCE FRAME COMPONENT 97

10. FILTER COMPONENT 105

11. REFERENCE PLANE COMPONENT 111

12. DIFFERENTIAL EQUATION COMPONENT 119

12–2 Component Variables 119

13. PROGRAMMABLE COMPONENT 125

14. TRANSFER FUNCTION COMPONENT 131

15. FOURIER SERIES COMPONENT 137

16. PRESCRIBED CONTROL COMPONENT 141

16-4.3 Random Time History 147

17. GUST COMPONENT 149

18. AERODYNAMIC COMPONENTS 157

19. RIGID AIRFRAME AERODYNAMICS COMPONENT 163

20. AIRFRAME FLOW FIELD COMPONENT 177

20–3 Implementation 179

21. LIFTING LINE WING COMPONENT 193

22. WING INFLOW COMPONENT 259

22-4.3 Dynamic Inﬂow Model 263

23. ROTOR INFLOW COMPONENT 267

24. ROTOR DYNAMIC WAKE COMPONENT 289

25. WING WAKE COMPONENT 307

25-3.5 Induced Velocity and Inﬂuence Coefﬁcients 341

26. WAKE GEOMETRY COMPONENTS 379

27. WING WAKE GEOMETRY COMPONENT 389

28. ROTOR WAKE GEOMETRY COMPONENT 413

28-3.2 Geometry and Frames 415

29. WING PERFORMANCE COMPONENT 427

30. ROTOR PERFORMANCE COMPONENT 435

31. ROTORCRAFT PERFORMANCE COMPONENT 449

31-3.2 Component Input 449

32. RIGID WING COMPONENT 455

33. HELICOPTER TAIL BOOM COMPONENT 461

34. COMPUTATIONAL FLUID DYNAMICS COMPONENT 473

35. PLUGIN COMPONENT 481

35-2.2 Shell Plugin 482

CAMRAD II is an analysis of aeromechanical systems that incorporates a combination of advanced tech-

CAMRAD II performs a nonlinear dynamic/static analysis of an aeromechanical system. Flexibility and

environmental physical logical

a) xA is a vector measured in axes A;

c) C BA transforms vectors from basis A to basis B, so xB = C BA xA .

solve equations (differential, integral, static, implicit) for motion of system

integrate in time RESPONSE

Figure 1-1 CAMRAD II tasks.

angular velocity and the time derivative of a rotation matrix:

a) linear motion: xBA/A and v BA/B = C BA ẋBA/A and v̇ BA/B

1–4 General Warnings and Limitations

rigid body Fourier series wing wake geometry

component input component output

degree of freedom ξ motion equation

Figure 2-1 Component functionality.

frame β degree of freedom ξ component input f

equation of motion A yes (0,1,2) or no yes or no