Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Canadian Aeronautics and Space Institute 50th AGM and Conference

16th Aerospace Structures and Materials Symposium


Montréal 28-30 April 2003

Equivalent Finite Element Wing Structural Models used for


Aerodynamics-Structures Interaction
M. Abdo, R. L’Heureux, F. Pépin, F. Kafyeke
Bombardier Aerospace
Tel: (514) 855-5000
Mohammed.Abdo@canadair.ca

Abstract
Equivalent beam finite element models, also known as stick models, are often used for static or dynamic aeroelastic
analysis and optimization. These stick models are usually generated when the wing structure has already been
defined and sized. Two methodologies were developed to create wing stick models for known wing structures. The
end product of both methods is a NASTRAN finite element beam model, which can be coupled to the aerodynamic
CFD model. The first method relies on a numerical procedure that is applied to determine the torsional and flexural
properties of multi-cellular cross sections. The second method extracts the stiffness properties of the beam model
from the complete finite element model of the wing. Both methods provide stiffnesses that are accurate. However,
these methods cannot be used at the conceptual design stage when the wing structure has yet to be defined. In that
case, one has to rely on historical data. The stiffness distributions of four of Bombardier’s existing aircrafts was
normalized and fitted with linear relations. These empirical relations can then be used to create a stick model for a
new aircraft early in its development phase. The accuracy of the empirical relations was checked and was found to
be acceptable.

1. Introduction The difficulty is to find or develop aeroelastic models


that are sufficiently simple to be called thousands of
Deformations of an aircraft during flight may have times during optimization, but are sophisticated enough
significant consequences on the aerodynamic to accurately predict wing deformations (both bending
performance and handling qualities. Predicting and twisting) [1,2]. Simplified beam finite element
accurate values of the bending and twisting of the models of aircraft wing structures, also known as stick
wing in flight depends on the fidelity of the stiffness models, are often used for aerodynamic-structure
properties (bending and torsional moments of inertia) interaction. Such models can be used for both static or
of the finite element model of the wing as well as the dynamic aeroelastic analysis [3,4,5].
aerodynamic loads. The finite element (FE) analysis
model is an idealization of a given structural system Two methodologies were developed to build wing stick
which is used to determine the system’s response models with acceptable stiffness accuracy. Both methods
(displacements, stresses, internal loads, etc.) to given are equivalent in terms of stiffness accuracy. The only
sets of applied loads. Validation of an F.E. model difference is that the first technique does not require
means making sure that the structural response of the building the full finite element model of the wing. The
model reproduces the structural response of the real second method requires the existence of a full finite
structure, within an acceptable accuracy. The element model. Once a model is available, the method is
complete finite element (FE) model of an aircraft robust and quick.
wing is usually prepared once the wing’s layout and
structural details are obtained. Different levels of The first method depends on a numerical procedure that
wing structural models have been used for wing static is applied to determine the torsional and flexural
aeroelastic analysis and optimization, ranging from properties of multi-cellular cross sections, which are
simple models based on analytic or empirical used frequently in modern wing structures. The shear
expressions, to complex finite element structural flow (used to find the shear center) in multi-cellular wing
models. sections (or thin-walled single-cell sections) becomes

1
indeterminate, thus requiring as many compatibility section, shear factors in two planes, area moment of
equations as the total number of cells. An algorithm inertia, area product of inertia, shear center location,
to evaluate these indeterminate shear flows has been torsional stiffness parameter and warping coefficient)
developed and incorporated into a computer involves various integrations. The numerical evaluation
program [3]. of these integrations is presented in the following
sections.
The second methodology depends on extracting the
stiffness properties of the beam model from the 3. Numerical Procedure for analyzing thin-
complete finite element model of the wing. The walled wing-box sections (method 1)
method requires some processing in PATRAN and
NASTRAN [6]. The numerical procedure that is developed for the
structural analysis of wing-box sections is presented in
2. Conceptual Structural Idealization this section. The theoretical basis of the present method
is given in [7-10]. The determination of the cross-
Thin-walled wing structural elements may be sectional properties can readily be obtained by
extremely complex. To enable such structures to be considering the section to be composed of a series of
analyzed, simplifying assumptions must be made, the interconnected prismatic straight thin-walled segments.
number and nature of which determine the accuracy
and degree of complexity of the analysis. Generally 3.1 General Sectional Properties
the more complex the analysis, the greater the
accuracy obtained. However, the amount of Consider the open section shown in Figure 2.
simplification introduced is governed by the
particular situation surrounding the problem; for an y
initial estimate or appraisal speed and simplicity are xc
often of greater importance than extreme accuracy Dx 0
while a final solution must be as exact as x

circumstances allow. Centroid yc

Dy 0
The present structural analysis method uses thin-
walled single-cell sections to represent the wing-box yi
[7-10]. Each wing-box section is modeled with a set yi Shear Center

of skin-stringer panels [11], front and rear spar webs ri


and upper and lower spar caps (see Figure 1). bi

Dsi
ti
Airfoil xi ri
Geometry

Wing -Box xi
Figure 2 Properties of an open section

The area of the section A is


n
Skin
Web A= å [ (D s i )(t i ) ] (1)
Pad-up i =1

Upper -Stringer Rear -Spar


where t i is the thickness of the element “i” of the
section, D si is the length of the element “i” of the
section and i = 1, 2, ...n is the current number of an
element of the section. The coordinates of the centroid
Boom Equivalent Wing-Box
( x c , y c ) are
1 n
Figure 1 Equivalent wing box extracted from the xc = å [ (D si )(ti )(xi ) ] (2)
actual structure A i =1
1 n
Essentially the determination of cross-sectional yc = å [ (D si )(ti )( yi ) ] (3)
A i =1
properties (such as, section area, centroid of the

2
where xi is the horizontal distance of the element “i” The warping of the cross-section due to unitary twist
to the origin and yi is the vertical distance of the angle per unit length may be calculated as follows
element “i” to the origin. The torsional stiffness
Wi = Wi - D (12)
parameter J is given by
where
1 n
J = å (D si )(ti )3
3 i =1
[ ]
(4) n
( )
å Wi (D si )(ti )
i
The area moment of inertia I1 (plane 1) is Wi = å [(D si )(ri )] and D= i =1
n
n é 1 ù
1
å (D si )(ti )
I1 = å ê (D si )(ti )( yi )2 + (ti )3 (cos b i )2 (D si )ú (5) i =1
i =1 ë 12 û
where ri is the distance from the center line of the
where b i is the angle between the central line of
element “i” to the shear center. The warping coefficient
element “i” and the “x” axis. The area moment of can be calculated from
inertia I 2 (plane 2) is
n é

i =1 ë
1
12
ù
I 2 = å ê(D si )(ti )( xi )2 + (ti )3 (sin bi )2 (D si )ú
û
(6)
n
[
C w = å (Wi )2 (ti )(D si )
i =1
] (13)

The area product of inertia I12 is


n 3.2 Closed Section Properties
I12 = å [ (D si )(ti )( yi )( xi ) ] (7)
i =1
Consider the closed section shown in Figure 3. The twice
of the torsional area is computed from
The shear flow q1i due to unitary vertical shear load nu
in an element “i” is computed from Au = å (D si )(ri ) (14)
i =1
é i i ù where Au is the twice torsional area of the cell “ u ” and
ê I12 å ( xi )(t i )(D si ) - I 2 å ( y i )(t i )(D s i )ú
ë 1 1 û (8) D si is along the boundary of cell “ u ”.
q1i =
[
(I1 )(I12 ) - (I12 )2 ] y

The shear flow q 2i due to unitary horizontal shear


load in an element “i” is computed from
x1 x2
é i i ù
ê I1 å (xi )(t i )(D s i ) - I12 å ( y i )(t i )(D s i )ú
ë 1 1 û (9)
q 2i =
[
(I12 )2 - (I1 )(I 2 ) ]
yc x
The shear factors K1 in plane “1” and K 2 in plane D si
“2” can be evaluated from the rule of equal external
and internal energy of a segment of the thin-walled xc
beam. The formulas of the shear factors will be as ri
follows
1 1
K1 = and K 2 =
n Ds n Ds
i (
Aå i
(q1i )2
Aå q2i )2 ti
i =1 ti i =1 ti
The horizontal component of the distance from the Figure 3 A typical closed section
centroid to the shear center is computed from
n The diagonal factor “ Bu ” for cell “ u ” is computed from
D x = å [(q1i )(ri )(D si )] (10)
i =1 nu D si
The vertical component of the distance from the Bu = å (15)
centroid to the shear center is computed from i =1 ti
n
D y = - å [(q2i )(ri )(D si )] (11)
i =1
where Bu is the torsional shear factor of the cell “ u ”
where ri is the distance from the center line of the and u = 1, 2, ...nu is the current number of a cell in a
element “i” to the centroid. closed section.

3
The off-diagonal factors for cells “ v ” and “ u ” are where q 2i is defined in the previous section for an open
computed from section.
nv u The matrix equation is given by
D si
Bv u = - å (16)
i =1 ti
[B][Z ] = [D] (23)
where D si is along the common boundary of cells
“ v ” and “ u ”. The redundant unitary shear flow The formulas for the shear factors are similar to those for
“ X u ” in every cell “ u ” is calculated from the matrix the open section but redundant shear flows “ Yi ” and
equation as follow “ Z i ” are added

[B][X ] = [A ] (17) K1 =
1
(24)
nD si
where Aå (q1i - Yi )2
t
[X ] = [X 1 X 2 .. X u ..X nu ]T , [A ] = [A1 A2 ... Au ... Anu ]T i =1 i
1
and X u is the redundant unitary shear flow due to K2 = (25)
n D si
torque in cell “ u ”. The matrix B is Aå (q 2i - Z i )2
t
i =1 i
é B1 + B1, 2 + ... + B1,u + ... + B1,nu ù
êB + B2 + ... + B2,u + ... + B2,nu úú If the element “i” is on the boundary of cell “ u ” then the
ê 1, 2
ê .....
[B ] = ê B
..... ..... ..... ..... ..... ú
ú
redundant shear flows are Yi = Yu and Z i = Z u . If the
ê 1,u + B 2,u + ... + Bu + ... + Bu ,nu ú
ê .....
element “i” is on the common boundary of cells “ u ” and
..... ..... ..... ..... ..... ú
ê
+ B2,nu + ... + Bv,nu + ... + B nu ûú
ú “ v ” then the redundant shear flows are Yi = Yu - Yv and
ëê B1,nu
Z i = Z u - Z v . If the element “i” does not belong to any
For a “1-cell” closed section closed circuit then Yi = 0 and Z i = 0 . The formulas for
A the shear center are similar to those for the open section
X1 = 1 (18)
B1 but the redundant shear flows “ Yi ” and “ Z i ” are added
The torsional stiffness parameter for closed sections
should be calculated from n
nu D x = å [(q1i - Yi )(ri )(D si )] (26)
( )
J = å Au ( X u ) (19) i =1
n
D y = - å [(q2i - Z i )(ri )(D si )]
u =1
(27)
The vertical shear flow factors “ Cu ” is given by i =1
nu D si
Cu = å (q1i ) (20) 3.3 Testing the Numerical Procedure
i =1 ti
where q1i is defined in section 3.1 for an open
The method developed in the previous section is used to
section. The redundant shear flow “ Yu ” in every cell calculate the cross-sectional properties of several open
“ u ” is derived from the matrix equation given below and closed sections [3,5]. Some of the results obtained
using the present method are presented in this section
[B ][Y ] = [C ] (21) along with additional test cases. A two-cell asymmetric
test section is shown in Figure 4 (All dimensions are in
inches). The thicknesses of the elements are shown in
The matrix [B ] is defined in the previous section but
Figure 4 and the cross-sectional properties are shown in
vectors [Y ] and [C ] contain the redundant shear flow Table 1.
“ Yu ” due to a unitary shear load and factors “ Cu ” in y
36 36 36 36
every cell “ u ”.
x

The corresponding formulas for the horizontal shear 0.5” 30


1”
flow factors “ Du ” and the redundant shear flow
“ Z u ” due to unitary shear load are as follows 0.75” 1” 0.75” 30

nu D si
Du = å (q2i ) (22)
i =1 ti Figure 4 Asymmetric two-cell test section

4
Present The stringers are attached to the wing-box skin using
Section Property Units Yoo & Acra
Method open segments (or rivets) with zero thickness and 1”
offset from the skin. This will make sure that the open
A In 2 390 390
segments will not contribute to the section’s properties
such as the area but it will offset the booms from the
I1 In 4 256818 256820
skin, which for example may affect the shear center
location and the wing-box overall stiffness. This type of
I2 In 4 657344 657350
wing-box section modeling will increase the fidelity of
I12 In 4 -67591 -67591 the analysis since it is more representative of the actual
structure. The properties of all the elements (skin
Centroid XC In 83.354 83.4 thickness and boom area) are given in Tables 2 and 3.
The cross-sectional properties of the wing-box section
Centroid YC In -26.077 -26.1 are shown in Table 4. Dimensions are in inches.
Shear Center To
In 7.435 7.435 Skin Skin Skin Skin
Centroid DX
Shear Center to (From- To) thickness (From-To) thickness
In -3.471 -3.510
Centroid DY (1-2) 0.11 (12-13) 0.08
(2-3) 0.11 (13-14) 0.16
Torsional Stiffness J In 4 450480 450570
(3-4) 0.11 (14-15) 0.16
(4-5) 0.11 (15-16) 0.16
Table 1 Comparison for Asymmetric two-cell section (5-6) 0.11 (16-17) 0.16
(6-7) 0.11 (17-18) 0.057
A wing-box test section with the booms (or stringers) (7-8) 0.11 (18-19) 0.16
offset from the skin is shown in Figure 5. (8-9) 0.11 (19-20) 0.16
255
18 17
(9-10) 0.11 (20-21) 0.16
21

250
22 14
13 (10-11) 0.11 (21-22) 0.16
(11-12) 0.11 (22-1) 0.103
245

240
Table 3 Skin thickness used in a wing-box test section
y

235
Description Units Present method
230
A In 2 28.805
225 8 10 12

1
2 5
I1 In 4 3.283311E+03
220
380 390 400 x 410 420 430 440
I2 In 4 6.879926E+03
Figure 5 Wing-box test section with booms offset
from the skin by zero thickness segments I12 -5.836029E+01
In 4
Boom # Area, In2 Boom # Area, In2 J 3.3243E+03
In 4
1 1.31 12 1.196
2 0.88 13 1.196 K1, K2 0.101, 0.2834
3 0.88 14 0.584
Twice Cells Area In 2 2.0949722E+03
4 0.86 15 0.584
5 0.77 16 0.584 Distance from origin In XC=411.2,
6 0.61 17 0.561
7 0.77 18 0.667 to Centroid YC=237.4
8 0.61 19 0.65 Distance from origin In XS=408.8,
9 0.77 20 0.65
10 0.88 21 0.85 to Shear Center YS=237.9
11 0.88 22 1.56 CW 0.54464E+08
In 6
Table 2 Area of the various booms used in a wing-
box test section Table 4 Properties of a wing-box test section

5
A skin-stringer test section is shown in Figure 6. This 3.4 Equivalent Wing Stick Models from
test panel shows a different method for modeling the Wing-Box Sections
wing-box cross-sections. In this test section, the three
stringers are modeled using its original shape (instead The stick model of a wing structure is basically a series
of a lumped concentration of area attached to the of tapered or uniform beam or bar elements (see
skin-boom) and offset from the skin using zero Figure 7). The properties of elementary beam end points
thickness segments. are extracted from the wing box rib’s structural
characteristics at the element’s extremities. The
Wing-box models of this kind increase the degree of numerical procedure developed in this section is used to
complexity of the analysis and simulate the actual calculate the wing-box cross sectional properties at a
behavior of the original structure. The cross-sectional given wing station.
properties of this panel are given in Table 5.
4.5 Rib station 2

3.5 Panel shear


center Skin-stringer Skin-stringer
Skin thickness Beam element 1
model 1 model 2
3 t = 0.1 in.
Rivet
2.5
t = 0 in.
y 2 Rib station 1

1.5

Stringer thickness
Figure 7 A typical stick FEM beam element
1
Panel Centroid t = 0.375 in.
0.5
Origin
The structural properties are extracted from the actual
0
structural details of the wing including skin thickness,
stringers areas, locations of front and rear spars, details
-0.5
of the front and rear spars, etc. The stations at the ribs are
-2 0 2 4 x 6 8 10 12
modeled by upper and lower surface effective
thicknesses and stringers areas. The stringers areas are
Figure 6 Skin-stringer test panel with realistic computed from the actual structure. The rear and front
stringer modeling spars, as well as the spar caps are also used in the
modeling. The effective width of the skin at the front and
Description Units Present method rear spar caps is used in computing the corresponding
A 5.50 areas. An algorithm is developed to generate
In 2 automatically the property cards required as an input in
I1 4.0172 NASTRAN such as PBEAM, RBAR, GRID, FORCE,
In 4 etc [3]. The stick model of the wing is then generated by
I2 57.31425 a series of NASTRAN property cards that depend
In 4 completely on the wing-box.
I12 In 4 -1.575
4. Wing Stick Models Extracted from the
J In 4 0.21427 Full Finite Element Model (method 2)
K1, K2 0.2319, 0.1431 The second methodology (Figure 8) extracts the stiffness
Distance from origin In XC=5.409, properties of the beam model from the complete finite
element model of the wing. The method requires some
to Centroid YC=1.2 processing in PATRAN and NASTRAN and it requires
Distance from origin In XS=4.747, that the user be knowledgeable of both software
packages. The stiffness extraction process begins by the
to Shear Center YS=3.029 definition of the elastic axis at each wing station. At
CW 41.3211 every elastic axis “kink” or “break”, a local coordinate
In 6 system is created along the elastic axis as well as the axis
of the principle inertias.
Table 5 Cross-sectional properties of skin-stringer
test section

6
At each elastic axis node, a specific analysis moment along the x-axis to calculate the vertical bending
coordinate system is assigned to the proper moment of inertia. The second set is the moment along
coordinate system (created in the previous step). In the y-axis to calculate the horizontal bending moment of
the present convention, the “x” and “y” axis are along inertia. The third set is the moment in the z-axis direction
the axis of the principle inertias while the “z” axis is to calculate the torsional stiffness rigidity. The wing
along the elastic axis. rotation in “x”, “y” and “z” directions due to the
corresponding applied moments are computed using
Rigid Bar NASTRAN and the equivalent EI’s and GJ’s are then
Rib Elements computed from
Full Finite Position
Element M x DL
Model of (EI1 ) i ®i +1 = (28)
the Wing q xi +1 - q xi

Elastic Axis Node M y DL


(EI 2 ) i ®i +1 = (29)
Stick FEM q yi +1 - q yi
Elastic Axis of the Wing M z DL
Definition (GJ ) i ®i +1= (30)
q zi +1 - q zi
Figure 8 Extracting the stick finite element model of
the wing from the full FEM
where M x , M y and M z are the moments applied at the
Rigid bar elements are then used to link the finite wing extremity (usually the tip) along the “x”, “y” and
element model (of the wing) nodes with the “z” axis while DL is the distance between two
corresponding planar elastic axis node. The proper consecutive elastic axis nodes. The corresponding twist
Bulk Data File of the elements is then created in (obtained from NASTRAN) due to the applied moment
PATRAN. All of the previous pre-processing steps for an elastic axis node (for example ( xi , yi , zi ) where
are done manually in PATRAN.
“i” is the node number) in “x”, “y” and “z” directions are
denoted by q xi , q yi and q zi . The modulus of elasticity
4.1 Evaluation of the Equivalent EI’s and
and rigidity are denoted by E and G , respectively.
GJ’s
To evaluate the equivalent moments of inertia and the 4.2 Evaluation of the Equivalent EA’s and
torsional rigidity of the stick model, two nodes are GK’s
created at the extremities of the structure (the root
and the tip). Those nodes are attached to the wing The process of evaluating the equivalent area and shear
structure by rigid bodies whose independent degrees factors spanwise distribution is different from that of the
of freedom are specified at a single grid point (elastic EI’s and GJ’s. This process requires executing
axis wing root and tip) and whose dependant degrees NASTRAN for three load subcases per each wing bay. In
of freedom are specified at an arbitrary number of all of these subcases, the three degrees-of-freedom
grid points (surface of the wing full FEM). One of the related to the rotation of the frames at the ribs are frozen
two extremity nodes (usually the root) is constrained and the other three degrees-of-freedom related to the
from displacement and rotation while the other displacements are freed. The elastic axis nodes are
extremity (usually the tip) is the point where the attached to the wing structure by rigid bodies whose
moments are applied. The elastic axis nodes between independent degrees of freedom are specified at a single
the two extremities are attached to the wing structure grid point and whose dependant degrees of freedom are
by rigid bar elements which defines the motion at a specified at an arbitrary number of grid points.
reference grid point (elastic axis node) as the
weighted average of the motions at a set of other grid In the first subcase, a known force in the “z” direction is
points (full FEM grid points). applied in order to calculate the equivalent area. In the
second subcase, the structure in the “y” direction is
The next step towards extracting the equivalent EI’s loaded to evaluate the vertical shear factor. In the last
and GJ’s of the wing structure is to load the wing subcase, the structure is loaded in the “x” direction to
finite element model with three sets of moments per calculate the horizontal shear factor. The equivalent area,
each segment of the elastic axis. The first set is the vertical and horizontal shear factors are calculated from

7
Fz DL
(EA)i ®i +1 = (31)
Dzi +1 - Dzi
Fy DL
(GK1 )i ®i +1 = (32) (EI )
Challenger 300
CRJ 200
D yi +1 - D yi
2 FEM
CRJ 700
Global Express

Fx DL
(GK 2 )i ®i +1 = (33)
D xi +1 - D xi
where Fx , Fy and Fz are the applied forces along
0 0.2 0.4 h 0.6 0.8 1

the “x”, “y” and “z” axis, respectively. The Figure 10 Spanwise variation of the in-plane bending
corresponding displacements in x, y and z directions stiffness
(obtained from NASTRAN) due to the applied forces
for an elastic axis node “i” are denoted
by Dxi , D yi and Dzi .

(GJ )
4.3 Stick Models of Bombardier’s Wings
Challenger 300
FEM
CRJ 200
CRJ 700
Global Express

The stiffness extraction method described above has


been used to generate stick models of four of
Bombardier’s wings namely, the Global Express, the
CRJ 200, the Challenger 300 and the CRJ 700. The
spanwise variation of the vertical bending stiffness 0 0.2 0.4 h 0.6 0.8 1

(EI1 )FEM , the in-plane bending stiffness (EI 2 )FEM Figure 11 Spanwise variation of the torsional stiffness
and the torsional stiffness (GJ )FEM are shown in
4.4 Empirical Relations
Figures 9-11 for the Global Express, CRJ 200,
Challenger 300 and CRJ 700.
The objective of this study is to obtain empirical
relations that could be used to estimate the stiffness
The curves in Figures 9-11 give a very good
distribution of a new wing using the stiffness distribution
indication of the structural stiffness behavior of the
of Bombardier’s existing wings. These relations could
wings. However, it could be misleading to compare
also be used to compare the stiffness behavior of a newly
the wing structures using these curves since most of
designed wing with the existing ones or check the
the differences observed on Figures 9 to 11 results
stiffness predicted by another method. As shown in the
from different planform geometry rather than from
previous section, the stiffness properties of different
different design philosophies. For example, the
wings can vary widely. In order to obtain empirical
Global Express has the highest stiffness values while
relations that apply to all these wings, one has to
the Challenger 300 has the lowest ones. This
normalize the data in order to collapse the stiffness
behavior is expected since the Global Express wing is
distribution of all the existing wings into as tight a cloud
longer than the Challenger 300, CRJ 200 and even
of points as possible. Several normalization techniques
the CRJ 700.
were investigated and applied to Bombardier wings.
Although a description of these techniques is given
below, only the results of the last technique will be
presented.
Challenger 300
CRJ 200
CRJ 700
(EI )
1 FEM Global Express a) The stiffness of the wing structure is divided by the
stiffness of a solid block material bounded by the
front and rear spar of the wing-box.

b) The stiffness of the wing structure is divided by the


stiffness of a solid block material bounded by the
h
0 0.2 0.4 0.6 0.8 1
front and rear spar or the auxiliary spar if there is
Figure 9 Spanwise variation of the vertical bending one at the station being considered.
stiffness

8
c) The stiffness of the wing structure is divided by
the stiffness of a solid block material bounded by
the leading and trailing edge of the wing (EI 2 ) FEM
(referred to as (EI )CATIA ). The stiffness values
= 0.0103h + 0.007
(EI 2 )CATIA

are obtained from the CATIA wing model.


(EI )
2 FEM

(EI )
None of these three techniques was clearly superior 2 CATIA

to the others. The third technique (using the inertia of


Global Express
the complete section as the reference) is the simplest CRJ 200
to implement and was ultimately selected for that Challenger 300

reason. The normalized stiffnesses are shown in CRJ 700

Figures 12-14.
0 0.2 0.4 h 0.6 0.8 1

For EI1 , the behavior of the normalized stiffness


appears to be different outboard and inboard of the Figure 13 In-plane bending stiffness of four of
break in the planform. Consequently, different Bombardier’s wings fitted with a straight line
relations were used to fit the data on the inboard and
outboard wing. The final empirical relation for the The results using the empirical relations are compared in
normalized stiffness is Figures 12, 13 and 14 with the real normalized stiffness.
Figure 13 shows the in-plane bending stiffness for
(EI1 )FEM RBreak - RRoot different wings fitted with one straight line and the
= ( h - h Root ) + RRoot torsional stiffness, which is fitted with one horizontal
(EI1 )CATIA h Break - h Root
line at 0.002, is shown in Figure 14.
for h Root £ h £ h Break
(EI1 )FEM
= RBreak for h Break £ h £ 1
(EI1 )CATIA

where R Root is the (EI1 )FEM / (EI1 )CATIA ratio at (GJ ) FEM

h Root and R Break is the (EI1 ) FEM / (EI1 )CATIA ratio (GJ ) CATIA

at h Break and if the wing has no kink then it is the


ratio at the tip, Rtip at h tip = 1 . It was determined Global Express
CRJ 200
that R Root = 0.03 and R Break = Rtip = 0.1 provides Challenger 300
CRJ 700
an acceptable fit for all airplanes.
0 0.2 0.4 h 0.6 0.8 1

Figure 14 Torsional stiffness of four of Bombardier’s


wings fitted with a straight horizontal line
(EI )
(EI )
1 FEM
5. Accuracy of the Empirical Relations
1 CATIA

Global Express The objective of coupling aerodynamic and structural


CRJ 200 models is to compute the flow over flexible wings. In the
Challenger 300
CRJ 700
work presented in this paper, the aerodynamic
Global Express_line characteristics of the flow are calculated using KTRAN
CRJ 200_line
Challenger 300_line
[12]. The fluid equations are solved using a transonic
CRJ 700_line perturbation formulation and a grid embedding
technique. The aerodynamic forces computed by
0 0.2 0.4 h 0.6 0.8 1
KTRAN are integrated to obtain the shear forces
required in NASTRAN. The shear forces are then
Figure 12 Vertical bending stiffness of four of applied to the stick model. The set of GRID and FORCE
Bombardier’s wings fitted with straight line segments cards is prepared automatically from KTRAN results.

9
The coupling procedure starts by computing the The empirical relations presented in the previous section
structural properties of the wing-box. In the first were used to generate approximate stick models for the
iteration, the aerodynamic loads are calculated for the Global Express, the CRJ 200, the Challenger 300 and the
undeformed wing (initial geometry, usually the jig CRJ 700. The actual and approximate stick models were
geometry). The initial values of wing vertical then coupled with KTRAN to predict the wing
deflection and twist are obtained from the deformation. The KTRAN aircraft configuration for all
NASTRAN solution. NASTRAN results are coupled the tests is Wing-Body-Engine-Winglet. Although
with KTRAN by modifying the twist specified in several test cases were conducted, only selected ones will
KTRAN. be presented. The flow conditions for the Global Express
test case are Mach=0.85, CL=0.376. The CRJ 200 and
A new set of loads is then obtained with KTRAN. Challenger 300 flow conditions are Mach=0.7, CL=0.3.
The shear forces are then given to NASTRAN to The CRJ 700 flow conditions are Mach=0.7 and CL=0.4.
obtain a more accurate set of wing deflection and
twist. This iterative procedure is repeated until the The wing deformations predicted by the actual stick
coupled solution converges. model (derived from the full FEM) and that generated
using the empirical relations are shown in Figures 15 to
Convergence is reached when the difference between 18. The prediction of the stick model generated using the
the twist of a previous iterative step and the current empirical relation is in good agreement with the
step is smaller than a user specified criterion, which deformation predicted by the actual stick model. The
is currently set at 0.1o. When a very accurate average deflection difference between the two models at
prediction of the wing twist is required, the the tip is ~5% while the twist difference is ~6%.
convergence criterion can be reduced to 0.01o .
Convergence represents the static aeroelastic
equilibrium state of the wing. The wing equilibrium
state is usually reached after 4 iterations.
Deflection (in.)

Actual Stick Model


Using Empirical Relations

Mach = 0.7
C L = 0.3
Alt = 30,000'
Deflection (in.)

Actual Stick Model Re = 12 x10 6


Using Empirical relations

Mach = 0.85
0 0.2 0.4
h 0.6 0.8 1

C L = 0.376
Alt = 45,000'
Re = 12 x10 6

0 0.2 0.4
h 0.6 0.8 1
Twist (Deg.)

Actual Stick Model


Using Empirical Relations

Mach = 0.7
C L = 0.3
Alt = 30,000'
Twist (Deg.)

Actual Stick Model Re = 12 x10 6


Using Empirical relations

Mach = 0.85
0 0.2 0.4
h 0.6 0.8 1

C L = 0.376
Alt = 45,000'
Figure 16 CRJ 200 wing deformation predicted by the
Re = 12 x10 6
actual stick model and the model generated using the
0 0.2 0.4
h 0.6 0.8 1 empirical relations

Figure 15 Global Express wing deformation


predicted by the actual stick model and the model
generated using the empirical relations

10
6. Conclusion
Two methodologies were developed to build wing stick
models with acceptable stiffness accuracy. Both methods
Deflection (in.)

Actual Stick Model are equivalent in terms of stiffness accuracy. The only
Using Empirical Relations
difference is that the first technique does not require
Mach = 0.7 building the full finite element model of the wing. The
C L = 0.3
second method requires the existence of a full finite
Alt = 35,000'
Re = 6.1x10 6 element model.

0 0.2 0.4
h 0.6 0.8 1 The second methodology was used to build wing stick
models of four of Bombardier’s wings namely, the
Global Express, the CRJ 200, the Challenger 300 and the
CRJ 700. With these stick models, CFD calculations can
be made on flexible wings to predict their deformation.
Twist (Deg.)

Actual Stick Model


Using Empirical Relations
By comparing the stiffness distribution of the four stick
models, empirical relations were obtained for the
Mach = 0.7
C L = 0. 3
stiffness distribution. Stick models created using these
Alt = 35,000' empirical relations gave acceptable deformations, both in
Re = 6.1x10 6 twist and in vertical displacements.
0 0.2 0.4 0.6 0.8 1
As long as the engines are aft-mounted, the empirical
relations provide a quick and easy way to obtain a stick
Figure 17 Challenger 300 wing deformation model for a new wing. They can also be used as a reality
predicted by the actual stick model and the model check for a stick model obtained from different sources.
generated using the empirical relations
7. References
[1] Abdo, M., Kafyeke, F., Piperni, P. and Pépin, F.
“Multi-Disciplinary Optimization Of Wings”, AIAC
Conference Proceedings, May 2002
Deflection (in.)

Actual Stick Model


Using Empirical Relations
[2] Piperni, P., Abdo, M. and Kafyeke, F. “The Building
Mach = 0.8 Blocks of A Multi-Disciplinary Wing Design Method”,
C L = 0.5
CASI 50th Annual General Meeting 2003
Alt = 35,000'
Re = 19.5 x10 6
[3] Abdo, M., Kafyeke, F. and Pépin, F. “Transonic
0 0.2 0.4 0.6 0.8 1
Aerodynamics of Flexible Wings”, CASI 48th Annual
Conference Proceedings 2001- (47-53).

[4] Abdo, M., Kafyeke, F. and Pépin, F. “Multi-


Disciplinary Design & Analysis of Flexible Wings”,
CSME-MDE conference Proceedings, November 2001
Twist (Deg.)

Actual Stick Model


Using Empirical Relations

Mach = 0.8
[5] Abdo, M. and Pépin, F. “Transonic Aerodynamics of
C L = 0.5 Flexible Wings”, Bombardier Report. MAA-000-387
Alt = 35,000'
Re = 19.5 x10 6
[6] Reymond, M. and Miller, M. MSC/NASTRAN,
0 0.2 0.4 h 0.6 0.8 1
“Quick Reference Guide”, Version 68, 1994

[7] Bruhn, E.F. “Analysis and Design of Flight Vehicle


Figure 18 CRJ 700 wing deformation predicted by Structures”, S.R. Jacobs & Associates Inc., June 1973.
the actual stick model and the model generated using
the empirical relations

11
[8] Megson, T. “Linear Analysis of Thin-Walled [14] Dassault Systems, “CATIA Basic 3D: Base,
Elastic Structures”, John Wiley & Sons, New York, Surfaces, Solids and Drafting”, Version 4, 1996
1974.
[15] Beckert, A. “Coupling (CFD) and Structural (FE)
[9] Peery, D.J. and Azar, J.J. “Aircraft Structures”, models using finite-interpolation elements”, Aerosp. Sci.
McGraw-Hill Book Company, 1982 Technol. 4 (2000) 13-22.

[10] Kuhn, P. “Stresses in Aircraft and Shell [16] Young, W.C. “Roark’s Formulas for Stress &
Structures”, McGraw-Hill Book Company, 1956 Strain”, McGraw-Hill Book Company, 1989

[11] Abdo, M., Piperni, P. and Kafyeke, F. [17] Borowiec Z., “The Margin of Safety in the Plastic
“Conceptual Design of Stringer Stiffened Range for Combined Bending and Axial Loads”, 2nd
Compression Panels”, CASI 16th Aerospace CASI Symposium on Aerospace Structures and
Structures & Materials Symposium 2003 Materials, Ottawa, 1984.

[12] Kafyeke, F. and Piperni, P. “ Applications of [18] Timoshenko, S. and Gere, J. “Theory of Elastic
KTRAN Transonic Small Disturbance code to the Stability”, McGraw-Hill Book Company, 1961
Challenger Business Jet Configuration with
Winglets”, SAE Paper 881483, October 1988. [19] Niu, M. “Airframe Structural Design”, Conmilit
Press Ltd., 1993
[13] Yoo, C.H. and Acra, S.V. “Cross-Sectional
Properties of Thin-Walled Multi-Cellular Sections”, [20] Freudenthal, A. “The Inelastic Behavior of
Computers & Structures, Vol.22, No.1, pp. 53-61, Engineering Materials and Structures”, John Wiley &
1986. Sons, Inc., 1950

12

You might also like