Dynamic Response of Tall Building To Wind Excitation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

DYNAMIC RESPONSE OF TALL BUILDING

TO W I N D EXCITATION

By Morteza A. M. Torkamani, 1 M . ASCE a n d Eddy Pramono 2


Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

ABSTRACT: This paper investigates the dynamic responses of tall buildings sub-
ject to wind loading. One of the objectives of this research is to study the im-
portance of the torsional dynamic response, coupled with translational re-
sponses. Finite element modeling is used to assemble the stiffness matrix of
the structure. Torsional degrees of freedom are considered in the stiffness for-
mulation of elements and systems. Aerodynamic forces on a tall building are
calculated assuming a deterministic, pseudo-turbulent approach. These aero-
dynamic forces are distributed over the height of the building. The equivalent
concentrated aerodynamic loads, acting at each floor level are calculated using
the principle of virtual displacements. The governing differential equations are
nonlinear. An iterative method of solution is used to calculate the responses.
In order to simplify the solution procedure, a method of linearization is applied
to the aerodynamic forces and the final result is a set of second order differ-
ential equations with constant coefficients. A 15-story building is modeled as
an application. One comparative study has been made between the finite ele-
ment model and an equivalent continuous cantilever beam model. A second
comparative study is between nonlinear and linear models. The results are pre-
sented as response spectra for different gust frequencies.

INTRODUCTION

Research on tall building response to w i n d load has been presented


frequently in literature (6,8,12,16). These investigations concentrated on
both deterministic and probabilistic approaches. Although torsional re-
sponse has been a problem in some tall buildings, not m a n y researchers
include the effects of torsional degrees of freedom in their research. Full
scale measurements of torsional response indicates quantitatively that
neglecting torsion in analysis could result in serious error in the re-
sponse at the outer corners of building floors (4).
The importance of the torsional response of tall buildings to wind loads,
coupled with translational motions, has already b e e n recognized (J.
Chrostowski, "Wind Effects on Tall Buildings," unpublished report, 4,
16) Patrickson a n d Friedmann have suggested a special class of buildings
in which a structure can be modeled as a h o m o g e n e o u s cantilever beam
with uniform cross section. The locations of the mass, elastic a n d aero-
dynamic centers at any section of the cantilever beam form three distinct
lines parallel to the vertical. Classical m o d e shapes for a cantilever beam
have been used in the analysis (5).
Yang, Lin and Samali (16) modeled a tall building composed of N
identical floors. The linear stiffness of each floor is provided by massless
shear columns or walls a n d is characterized by three stiffness constants,
kx, ky and k,, at the elastic center. The locations of the aerodynamic,
'Asst. Prof., Dept. of Civ. Engrg., Univ. of Pittsburgh, Pittsburgh, Pa. 15261.
2
Former Grad. Student, Dept. of Civ. Engrg., Univ. of Pittsburgh, Pittsburgh,
Pa. 15261.
Note.—Discussion open until September 1, 1985. To extend the closing date
one month, a written request must be filed with the ASCE Manager of Journals.
The manuscript for this paper was submitted for review and possible publication
on June 1, 1984. This paper is part of the Journal of Structural Engineering, Vol.
I l l , No. 4, April, 1985. ©ASCE, ISSN 0733-9445/85/0004-0805/$01.00. Paper No.
19672.

805

J. Struct. Eng. 1985.111:805-825.


elastic and mass centers at each floor are assumed to form lines not nec-
essarily coincidental with but parallel to the vertical. The mass at each
floor level is lumped at the mass center of that floor (16).
One of the objectives of this study is to investigate the importance of
the torsional dynamic response, coupled with translational responses,
of tall buildings. A more realistic model for tall buildings has been se-
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

lected in which the building is modeled as three-dimensional frames. A


finite element model for the building will be used in order to consider
the contributions of every member—beams, columns, panels and shear
walls. Furthermore, the masses, lumped at each floor, have three de-
grees of freedom, i.e., laterally in two mutually perpendicular directions
and torsionally about a vertical axis.
The aerodynamic center, in general, will be dependent on time and
on wind attack angle. Since an analytical procedure for defining the
aerodynamic center is not available, it is assumed that the locations of
the aerodynamic center of all floors is taken to form a line parallel to the
vertical. The mass centers are not necessarily alined, since their locations
may vary from floor to floor.
In this study, a deterministic sinusoidal model of the atmospheric tur-
bulence (gust) is used. A variety of loading conditions, such as a varied
offset of the aerodyanamic center from the mass center, different angles
of approach and different gust amplitudes are considered. These
pseudoturbulent flow fields can be useful for simulating real turbulent
fields since they are more convenient to analyze. However, care must
be exercised when using results obtained from this simplified represen-
tation. Furthermore, it should be noted that in the present study, buf-
feting due to vortex shedding is not included (3).
Two computer programs have been developed to analyze the response
of the building to wind loading. These programs are also capable of cal-
culating the element forces. These computer programs are modifications
of a well-known computer program, ETABS, developed by Wilson, Holl-
ing and Dovey from the University of California, Berkeley (15). The ETABS
computer program has been used to analyze the responses of a building
to earthquake loading. Modifications have been made so that the com-
puter programs can be used to analyze the response of a building to
wind excitation.

STRUCTURAL IDEALIZATION

In this study, a three-dimensional multi-story building has been se-


lected for structural modeling. The complete building is idealized as a
system that is composed of structural components which can be sepa-
rated into a series of rectangular frames in arbitrary planes. Isolated shear
walls are considered as frames consisting of a continuous column line
and a dummy column line to define the principle axis of the wall. Each
frame is treated as an independent substructure. The complete structure
stiffness matrix is obtained using the assumption that all frame and shear
wall components are connected by floor diaphragms which are rigid in
their own planes (15).
Several assumptions have been made to model this type of structure.
The assumptions are described as follows:
806

J. Struct. Eng. 1985.111:805-825.


1. The floor diaphragms are assumed to be rigid in their own planes.
All floor levels must be continuous and horizontal. Approximate bend-
ing stiffness of the floor may be included in modeling the individual
floor beams; axial deformation is not permitted.
2. The horizontal lateral loads are transferred to the columns and shear
wall elements through these rigid floor diaphragms. Thus, the horizon-
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

tal lateral loads are assumed to act at floor levels.


3. Compatibility is not enforced with regard to joint displacements at
joints which are common to more than one frame. Thus, axial defor-
mations in common columns will not be the same. However, for design
purposes, these column axial forces may be added directly to give rea-
sonable results. If the whole building is modeled as a single frame, com-
patibility will be satisfied (9).
4. The inertia of a typical unit is lumped at the floor level, and is char-
acterized by mass, m, and mass moment of inertia, I, about the mass
center.

Each frame substructure modeled consists of four structural elements:


columns, beams, panels and diagonal elements (9).

ASSEMBLY OF STRUCTURAL STIFFNESS

Each joint has six degrees of freedom (displacements and rotations


about each coordinate axis); within each frame three of these degrees of
freedom (the two translations and the rotation in the floor plane) can be
transformed, using the assumption of rigid in-plane floor diaphragms,
to the frame degrees of freedom at that floor level. Before each frame is
added to the total structural stiffness matrix, the remaining three joint
degrees of freedom are eliminated by static condensation. These pro-
cedures have been implemented in the ETABS program to compute the
stiffness matrix for the whole structure (15).
The complete structure stiffness matrix is obtained from the frame
stiffness matrices by the assembly process. The individual frame stiff-
ness matrices must be transformed to a common displacement coordi-
nate system (the global coordinate system) before assembling. The global
coordinate system consists of two translations and one rotation per story
level. The origin of these global displacement coordinates at each level
are taken at the center of mass of that story segment. This position may
vary from story to story.
After all of the assembly processes have been done, the complete
structure has three degrees of freedom at each floor level. Translations
in the x and y directions are denoted by the variables u and v, respec-
tively, and rotation about the center of mass is denoted by the variable
9. The motion of each point on a floor level may be calculated consid-
ering the motion of the center of mass of that floor. Fig. 1 shows the
structure model.
There are two distinct points of the floor cross section which are of
particular interest in this study:

1. The cross sectional center of mass (M.C.).


2. The cross sectional aerodynamic center (A.C.).
807

J. Struct. Eng. 1985.111:805-825.


Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

FIG. 1.—Model of Structure FIG. 2.—Schematic of Velocity and


Aerodynamic Forces

The cross sectional center of mass, the location of which may vary
from floor to floor, is a reasonably well defined concept. However, it is
difficult (or not possible) to determine the location of aerodynamic cen-
ter, which represents the point of action of air loads for bluff bodies.
Therefore, its hypothetical location is based on the experience and judg-
ment.

MODELING OF WIND LOAD ON STRUCTURE

Structural safety and serviceability are important factors in the design


of tall buildings subject to wind loading. In order to design a safe struc-
ture, designers are required to understand:
1. The variation of the wind environment.
2. The relation between the wind environment and the forces induced
on the structure.
3. The behavior of the structure under the action of aerodynamic forces.
The variation in mean wind velocity with height is an important factor
in estimating the design wind velocity on tall structures. An empirical
expression for the velocity distributions throughout turbulent boundary
layers, the power law, can be used to approximate experimental values
of mean wind velocity.

V(z) = F grad - ( — -) (1)


\Zgrad/
in which V(z) represents the mean wind velocity at height, z; V^a =
the mean wind speed at a reference height, z grad ; and A = a constant
depending on the roughness of terrain, typically ranging between 0.15-
0.50 (2).
808

J. Struct. Eng. 1985.111:805-825.


A structure immersed in a given flow field is subjected to aerodynamic
forces. For simple structures aerodynamic forces may be estimated using
available results of aerodynamic theory and experiments. However, if
unusual conditions of the environment or properties of the structure need
to be considered, special wind tunnel testing m a y be n e e d e d . The aero-
dynamic force on the bluff b o d y will have c o m p o n e n t forces along a n d
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

across flow directions, a n d are referred to as drag a n d lift forces, re-


spectively. If the distance between the mass center of the structure a n d
the aerodynamic center (i.e., the point of application of the aerodynamic
force) is large, the structure is also subjected to torsional m o m e n t s . Tor-
sional m o m e n t s may play a significant role in the structural design of
the frame (8).
The aerodynamic forces of wind on tall buildings, which are not well
understood, are primary d u e to the complicated nature of the unsteady
flow about a bluff object. Consideration also should be given to the buf-
feting forces k n o w n as buffeting d u e to vortex shedding, buffeting by
gusts, buffeting d u e to wake exposure a n d galloping. These complexities
cause uncertainty in modeling unsteady aerodynamic forces on struc-
tures. Additional research in the calculation of unsteady aerodynamic
loads is required for more accurate modeling of w i n d loads on struc-
tures.
However, for this study, wind loads on buildings can be simplified by
using a quasistatic representation (13). This implies that the aerody-
namic forces, due to the flow geometry, s h o w n in Fig. 2, w o u l d develop
instantaneously without accounting for the usual force deficiency and
phase lag encountered in unsteady fluid dynamics (8). With this as-
sumption, the unsteady wind loads denoted by lift, L(z,t), a n d drag,
D(z,t), acting respectively normal and along the relative velocity, can
be expressed by Eqs. 13 and 12.
When a bluff body is immersed in a flow of velocity, V, the flow will
develop forces, k n o w n as drag and lift. Drag and lift forces on a bluff
body in a turbulent flow can be represented as follows:

D(z,t) = - pbV(z,t)2CD(a,t) + pb2 V(z,t)C„,(a,t) (2)

L(z,t) = ^ pbV{z,t)2CL(a,t) (3)

in which CD(a, t) and CL(a,t) = lift a n d drag coefficients, and d e p e n d


on the angle of attack of the flow to the structure a n d time. In the ab-
sence of other data, these coefficients may be taken from the experi-
mental work d o n e by Vivekananda (14), wherein C D and CL only de-
pend on the angle of attack. In most applications, the term containing
the coefficient, CM (the virtual mass coefficient, the inertial force asso-
ciated with the mass of the fluid displaced by the body, and proportional
to that mass), is disregarded because in wind application the contribu-
tion of this term to overall response is small.

DETERMINISTIC MODEL OF ATMOSPHERIC WIND .

In this phase of study, a deterministic m o d e l of the w i n d (pseudo-

809

J. Struct. Eng. 1985.111:805-825.


turbulent) will be assumed as harmonic in time. Although, replacing a
random load by a simple deterministic function is not the most conve-
nient way to treat wind turbulence, this has been done by other authors
for a deterministic model (6). This assumption is justified for pilot-proj-
ects and to show how the valuable physical insight gained from this
relatively simple approach can be used for introducing simplifying ap-
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

proximations in the equation of motion. In particular it is shown that


this approach enables one to identify the importance of nonlinear terms
and terms associated with the time-dependent portion of the aerody-
namic damping.
A three-dimensional approach is used to model a building; however,
wind forces are two dimensional and parallel to the cross-section of the
building. The wind velocity is composed of a mean flow and two su-
perimposed sinusoidal fluctuating components, along and across the mean
wind. The fluctuating components of the wind, which are small com-
pared to the mean flow, have frequency, co, and amplitudes, ug0 and
vg0, respectively (5). This deterministic representation implies that:
1. The mean wind velocity is a function of z only, and has a profile
expressed by Eq. 1.
2. The gusts are functions of time only, that is, they are not depen-
dent on 2. Their amplitudes are relatively small compared to mean wind
velocity (about 20%) and their fluctuations are harmonic with a period
of 2ir/u).
Therefore, the wind velocity as a function of height and time can be
represented by
V(z,t) = V(z) + ug(t) + vg(t) (4)
The mean wind velocity will be denoted by the vector V(z), while the
gust components, along and across the mean velocity, will be repre-
sented by vectors ug(t) = ug0 sin cof; and vg(t) = vg0 sin (w£ + <$>), re-
spectively. Excitation of the building due to aerodynamic lift and drag
forces will produce velocity components of the building, varying with
time, denoted by the vectors u(z, £) and v(z, t), parallel to the x and y
axis, respectively. Fig. 2 shows all of the relevant vectors at a typical
floor cross section of the building.
The geometry of Fig. 2 shows the relationship.
yjei = D l - D2 + D3
in which D l = V2(z) + 2V(z) ug(t) + u\{t) + v2g(t)
D2 = ti(z, t) D„ + v(z,t) Dv
D3 = H (2/() + v%,t)

Du = 2{[V(z) + ug(t)] cos p - vg(t) sin 0} (9


Dv = 2{[V(z) + ug{t)\ sin p + vg(t) cos p} (10;
The fluctuating instantaneous effective angle of approach is given by
, v, + u sin P — v cos P
8 = tan' 1 s
- — (11)
V + ug - u cos p - v sin p
810

J. Struct. Eng. 1985.111:805-825.


and a(t) = (3 I 8(f) 6(f). The corresponding lift and drag forces are

D(z,0 = - pbVUzJ) CD(a,t) (12)

L(z,t) = -pbV2lel(z,t) CL(a,t)


Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

(13)

WIND ANALYSIS

As was mentioned before, it is assumed that a horizontal load will act


at each floor level. Drag and lift forces are distributed loads, in the di-
rections along and across relative wind velocities. They act at the aero-
dynamic center. The drag and lift forces can be projected in the x and
y directions and denoted by Px(z,t) and Py{z,t), respectively.

Px(z,t) = - pbVUCD(<x,t) cos 7 - C L (a,f) sin 7] = - pbV^,C x (0 . . . . (14)

Pv(z,t) = -pbV^[CD(a,t) sin 7 + CL(a,t) cos 7] = - pbV2elCy(t).... (15)

in which 7 = (P + 8) is the effective angle of approach of VKi with respect


to the x-axis and
C.v(0 = CD(a, t) cos 7 - CL(a, t) sin 7 (16)
C,,(t) = C D (a,0 sin 7 + CL(a,t) cos 7 (17)
The equivalent concentrated load at each floor level must be deter-
mined to satisfy the previous assumption that a horizontal load acts at
each floor. This can be done by equating the work done by the distrib-
uted load to the work done by the concentrated load at each floor level.
The equation of dynamic equilibrium will be determined according to
Newton's second law, by direct equilibration of all forces acting on the
mass. The equilibrium equations for a structure may be written in the
following form:
MU + CU + KU = Q (18)
where: M = mass matrix; C = damping matrix; K = stiffness matrix; Q
= applied load vector; and U = vector of displacement of mass centers.
Mass matrix, M, defined in Eq. 18 is diagonal with m„ and /„, the
translational mass and the rotational mass moment of inertia at the n th
floor, respectively. The translational mass was obtained by lumping to-
gether all mass 1/2 the story height above and 1/2 the story height be-
low each floor. The rotational mass moment of inertia about the mass
center is obtained by the relationship
I„ = I0 + mr2 (19)
in which I0 = the rotational mass moment of inertia of a component
about its own centroid; and r = the distance from the mass center to
the centroid of the component. The total mass moment of inertia is the
sum of all the component mass moments of inertia.
811

J. Struct. Eng. 1985.111:805-825.


The damping matrix, C, has a characteristic such that pre- and post-
multiplication by the mode shape matrix will give
4> T C$ = [2£„a)„] : (20)
in which £„ = damping ratio of the nth mode; w„ = frequency of the
nth mode.
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

The stiffness matrix, K, is obtained from the assembly process where


each floor is considered to have three degrees of freedom. Thus, the
dimension of the matrix will be 3NST by 3NST, in which NST = the
number of stories in the building.
The mode shapes of vibration represent the solution of the undamped
free motion vibration. The mode shapes are normalized such that
<&r M«5 = 1 (21)
T
then * K $ = £1 (22)
in which <I> = mode shapes; and O = [oil] = square of frequencies.
The actual displacements, U, are now expressed as a linear combi-
nation of the mode shapes:
U = «6Y (23)
in which Y„(t) represents the response of the nth mode.

EQUIVALENT LOADING

In modeling the building, the reference displacements are taken at the


mass center of each floor. The aerodynamic forces are applied at the
aerodynamic center. For calculation of the equivalent aerodynamic forces,
the relationship between the displacements at the aerodynamic center
and the mass center is needed. A typical floor cross section is shown in
the Fig. 3.
At each floor, displacements at the aerodynamic center can be ex-
pressed in terms of displacements at the mass center, in matrix form as:
u ~1 0 -J/AC u
V = 0 1 *AC V (24)
0 AC 0 0 1 e MC

Therefore, the discrete shape functions at the aerodynamic center cor-


responding to the model matrix may be expressed as:
<&AC = A * M C (25)
in which <&AC = shape functions at the aerodynamic center; <E>MC = mode
shapes with respect to the mass center; and A = transformation matrix
defined by Eq. 24. Thus, the displacement at the aerodynamic center
can be written as:
UAc = *ACY (26)

When the building is considered to be a continuous system, the dis-


placements at the aerodynamic center can be expressed as the sum of
the shape functions of the aerodynamic center
812

J. Struct. Eng. 1985.111:805-825.


Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

^AC
x

FIG. 3.—Floor Cross Section

"(((2,0"
v{z,t) = *V*2... {Y} (27)
_9(2,oJ AC .WS-.JAC
or in more simple form
UAC = TACY (28)
The matrix, r A C , in Eq. 28 is a shape functions matrix. Each column of
this matrix is a shape function of the building at the aerodynamic center
for a particular mode and direction. The shape function matrix is a func-
tion of height, z.
The distributed loads Px(z,t) and Py(z,t) act at the aerodynamic cen-
ter. Therefore, equivalent concentrated loads at the aerodynamic center
can be obtained by equating the work done by distributed loads to the
work done by the equivalent concentrated loads.

p.;
QACUAC = dz (29)
Jo l _ 0 j
In terms of generalized displacement:

QAC*ACY= PACrACY<fz (30)

and Qj
>AC*AC = 1 (31)
Jo
Substituting for <J>AC from Eq. 25 into the transpose of Eq. 31:
813

J. Struct. Eng. 1985.111:805-825.


*MCATQAC = I rTACFACdz (32)
Jo
in which Q AC is denoted as an equivalent concentrated load vector at
the aerodynamic center.
Considering that the displacements at each floor are referred to the
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

mass center, the concentrated loads at the aerodynamic center need to


be transformed to equivalent loads at the mass center. This relationship
can be expressed for the whole system in compact notation as:
QMC = A T Q AC (33)
Premultiplying by <&MC leads to:
*MCQMC = 3>McArQAC (34)
Eq. 32 can be substituted into the right hand side of Eq. 34 and the result
is

<*>MCQMC = r AC P AC dz (35)
Jo

in which Q MC is denoted as the equivalent concentrated loads vector at


the mass center.

DYNAMIC RESPONSE ANALYSIS

Combining Eq. 23 and Eq. 18, the equation of motion may be ex-
pressed as:
M<*>MCY + C<t>MCY + K* M C Y = Q MC (36)
Premultiplication of Eq. 36 by <&MC yields to the second order differential
equations, which are coupled by the forcing functions.
M*Y + C*Y + K*Y = P* (37)
The term P* = <&MCQMC has already been defined in Eq. 35 as:

P* = r AC P AC dz (38)
Jo
Thus, the final equation of motion may be rewritten as:

[J]Y + [2£„u>„]Y + [co?,] Y = r AC P AC dz (39)

SOLUTION OF WIND RESPONSE

The right hand side of Eq. 39 represents the aerodynamic loads where
the matrix PAC has already been defined as:

1
PAC = -p&[Dl - D 2 + D3] (40)

814

J. Struct. Eng. 1985.111:805-825.


It should be noted that, the terms in the aerodynamic forces containing
Dl, D2 and D3 have important physical interpretations (5): D l repre-
sents the main aerodynamic excitation; D2 represents the aerodynamic
damping effect; D3 represents the nonlinear damping term, which is
quite small in actuality. Thus, Eq. 39 can be considered as a coupled
nonlinear second order differential equation.
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

NONLINEAR EQUATION OF MOTION

In the section entitled "Deterministic Model of Atmospheric Wind,"


the deterministic modeling of wind gusts is approximated closely by a
special case of harmonic function in the time variable. This indicates that
the aerodynamic forcing functions are periodic. A periodic function can
be separated into its harmonic components by means of a Fourier series
expansion. Although theoretically a Fourier series representation of a
periodic function may require an infinite number of terms, in actual
practice periodic functions may generally be approximated with suffi-
cient accuracy by a relatively small number of terms (1).
Having a periodic forcing function, the steady state solution is also a
periodic function of time. The response to each term of the series is,
then, merely the response to a harmonic loading. By the principle of
superposition, the total response is the sum of the responses to the sep-
arate load terms. A particular solution for the second order differential
equation with a periodic forcing function is available in the literature (7).
Eq. 39 is a coupled, nonlinear equation of motion. The nonlinearity of
the right hand side of Eq. 39 is due to the nonlinear damping term D3.
An iteration procedure is applied to solve for the response. This solution
is developed by initially disregarding the effect of the small nonlinear
damping. Then, improved iterative solutions are obtained by a series of
iterations until convergence criterion is achieved, that is, when the input
solution and the improved one are almost equal.

LINEAR EQUATION OF MOTION

Solving the nonlinear equation of motion using an iterative procedure


requires a great amount of computer time. In practice, it is sufficiently
accurate to neglect the nonlinear term D3 (5,11). A suitable linearization
is applied by the following considerations:
1. The nonlinear aerodynamic damping, D3 = u2 + v2, in the expres-
sion of the aerodynamic forces is neglected.
2. The fluctuating part of the effective angle of approach of the wind
is approximated by

8 = tan" 1 — ^ — (41)
V+ug
3. The periodic aerodynamic damping coefficients due to the term D2
= iiJDu + vDv are replaced by time averaged values
C C rf ; T= : (42)
Sf " ' ( T)
815

J. Struct. Eng. 1985.111:805-825.


After linearization, the equation of motion yields a system of second
order linear differential equations with constant coefficients.
[I]Y + ([2L<o„] + [Cav]) Y + [a)?,] Y = F(f) (43)
av
The details of calculating matrix [C ] are given in Ref. 10. Now this
linear coupled equation of motion subjected to a periodic external forc-
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

ing function can be solved without using an iterative procedure.

APPLICATIONS AND COMPUTER PROGRAMS

Using the theory presented in previous sections of this paper two


computer programs have been developed. Both programs have the ca-
pability of calculating displacements of the mass center of each floor or
stresses, or both, in each member. The first program considers three
dimensional nonlinear analysis of the buildings. The linearization pro-
cedure presented in the section entitled "Linear Equation of Motion" is
incorporated in the second computer program.
A 15-story building known as the University of Pittsburgh's Natural
Science Facility is chosen for analysis. The data for this building has
been presented in Ref. 9. The main tower is 144 ft (43.89 m) by 81 ft
(24.69 m) in plan and rises 237 ft (72.24 m) above ground level (9).
A comparison has been made between the theory presented in this
report and the work of Patrickson and Friedman (5). A third computer
program, which was developed by Patrickson, has been used for this
comparison. This program considers the building as a cantilever beam
with three degrees of freedom at elastic center. These degrees of free-
dom are two lateral displacements and a rotational degree of freedom.
Modula of rigidity E/,/v, Elxx and GJ or the fundamental periods of the
building, Tx, T„ and Te are part of the input data to this program. In
order to have a cantilever building equivalent to the finite element model,
fundamental periods of the building using program TLWAB (Three Di-
mensional Linear Wind Analysis of Buildings) (10) is calculated. Then,
these data are used as input for the continuous model.
Figs. 4(«), (b), (c) are the first three mode shapes of the building with
the corresponding periods. Mode 1 is dominated by lateral displace-
ments in the x-direction. Modes 2 and 3 are combined lateral-torsional
modes. For the sake of this comparison, considering Figs. 4(b), (c), it is
decided that modes 2 and 3 are dominated by lateral displacements in
the y-direction and torsional motions.
Figs. 5(a), (b), (c) show the dimensionless response spectra u/h, v/h
and 6 • otc/h, at the roof of the building, in which otc (outer corner) =
the farthest distance of the columns from the mass center or elastic cen-
ter at each floor. These figures show that the response spectra of the
top floor in the x-direction are relatively close for the two models. Dif-
ferences between the response spectra of the two models in the y-direction
and the 6-direction are larger. The results indicate that the finite element
model is more conservative. It should be noted that iterative procedures
have been used in both programs for this comparison.
Fig. 5(c) shows that the fundamental periods of the torsional mode of
the two models are different, even though the input for cantilever model
is the output from the finite element model. This should not be a sur-
816

J. Struct. Eng. 1985.111:805-825.


Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

MODE SHAPE NO. 1

MODE SHAPE NO. 3

FIG. 4.—First Three Mode Shapes of Building: (a) Model No. 3, r = 1.7856 sec;
(b) Model No. 3, T = 0.8733 sec; (c) Model No. 3, T = 0.6207 sec

817

J. Struct. Eng. 1985.111:805-825.


LEGEND
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

Finite Element Model


Cantilever Beam Model

l.Q 1.5 2.0 2.5 3.0


Gust Frequencies C.P.S.
(a)

LEGEND
Finite Element Model

Gust Frequencies C.P.S.


(b)

818

J. Struct. Eng. 1985.111:805-825.


LEGEND
n Finite Element Model
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0


Gust Frequencies C.P.S.
(c)

FIG. 5.—Comparison Between Finite Element Mode and Cantilever Beam Mode,
Response Spectra for 15-Story Building

prising observation, since the calculated periods using finite element


method are coupled lateral-torsional periods. For the cantilever model
Galerkin's method is used to convert the equation of motion to a set of
ordinary differential equations. The selected mode shapes for this re-
duction procedure are pure lateral or torsional modes of the cantilever
beam.
The iterative solution of the nonlinear formulation of the equation of
motion is very time-consuming and expensive. Therefore, the idea of
linearization appears very attractive. On the other hand, the accuracy of
the results is of immediate concern. Considering the previously men-
tioned factors, a study comparing the results of the iterative and linear-
ized methods is undertaken.
The effect of the nonlinear and linearized aerodynamic forces on the
responses are evaluated by considering the following aerodynamic loads:

1. An exact form, where all terms in the aerodynamic forces are in-
cluded. An iterative method of solution is used in the computer program
for this form.
2. An almost exact form, where the term D3 is excluded. This also
requires an iteration procedure for the solution.
3. A linearized form, where linearization is performed as examined in
"Linear Equation of Motion."
4. A linearized form, similar to item 3 where, in addition, the term
ul + vl is excluded.
819

J. Struct. Eng. 1985.111:805-825.


LEGEND
D Exact Solution
& RLmost Exact Solution
x Linearized Solution
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

* Linearized Solution
2 2
+
NegLectung u v

Tx - 1.7856
TT - 0.8733
T 9 - 0.6207
- 0.10
- 0.00
- 0.00
0.2 V

Gust frequencies C.P.S.


(a)

LEGEND
Exact Solution
Rlmost Exact Solution
Linearized Solution
* Linearized Solution
2 2
Neglecting u + v

T» - 1. 7856
Ty - 0.8733
Tn - 0 6207
t - 0 10
^ - 0 00
» - 0 00
u -V - 0.2

1.0 1.5 2.0 2.5 3.0


Gust Frequencies C.P.S.

820

J. Struct. Eng. 1985.111:805-825.


LEGEND
o Exoct S o l u t i o n
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

FIG. 6.—Comparison of Various Approximations in Responses, Response Spec-


tra for 15-Story Building

The responses of the top floor of the 15-story building have been cal-
culated using these aerodynamic forces for a mean wind velocity of 115
ft/sec (35.05 m/s) and sinusoidal gusts ug(t) = vg(t) = 0.20V sin (2irt/
T) ft/sec or m/s. It should be noted that different gust amplitudes with
phase angle could be used for alongwind and acrosswind turbulence
flow. Using this option of the computer program one has control on the
alongwind and acrosswind responses of the building.
Fig. 6 is the dimensionless response spectra of the top floor in the x-,
y- and 6-directions for these four cases. It indicates that, the response
of the top floor of the building due to the linearized form of the aero-
dynamic forces is very close to the nonlinear form. These results are very
encouraging.
A parametric study has been conducted to study the effect of the var-
ious parameters on the response of the building considering the linear-
ized form of the aerodynamic forces. Several cases have been considered
for this study and the results are presented in the conclusion section.
The response spectrum of each coordinate indicates that the maximum
response occurs when gust frequencies are close to the fundamental fre-
quencies of the building. The maximum response in the x-direction for
this particular building occurs when the gust frequencies are close to
0.56 cps. The maximum responses in the y-direction and 0-direction oc-
cur when the gust frequencies are close to 1.16 cps.
Fig. 7 shows a comparison between the nondimensional response
821

J. Struct. Eng. 1985.111:805-825.


Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0


Gust Frequencies C.P.S.

FIG. 7.—Comparison of Responses in x, y, and 9 Directions, Response Spectra


for 15-Story Building, Linearized Solutions

spectra of x, y, and 6 directions (u/h, v/h and 0-otc//z). Larger magni-


tude of the torsional displacements than displacements in the y-direction
indicates the importance of the torsional response. The torsional re-
sponse is found to be even more significant when the lateral and tor-
sional first fundamental frequencies are close, and the torsional response
has a higher order of magnitude than the lateral response. Since for this
type of building, the lateral and torsional fundamental frequencies are
close, inclusion of the torsional degree of freedom makes a significant
contribution to the dynamic response of the building.

CONCLUSIONS

An analytical approach is presented for calculation of the displace-


ments or stresses, or both, for high rise buildings subjected to wind ex-
citation. The following conclusions are drawn from this study:

1. A comparative study between the finite element model and a can-


tilever model of the building shows the following results: (a) A cantilever
model equivalent to a finite element model is not possible; (b) torsional
responses of the building are more significant in the finite element model
Q£.£,

J. Struct. Eng. 1985.111:805-825.


than they are in the continuous cantilever model. Since analysis using
the iterative method of solution requires a great a m o u n t of computer
time, two methods of linearization are considered. The m e t h o d of lin-
earization which keeps D l terms in the analysis is quite accurate and
recommended for future analysis.
2. Other parametric studies show that: the angle of approach of wind
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

velocity has a major effect on the responses of the building. Different


angles of approach should be considered in the analysis of tall buildings.
Modal damping ratios are very important in the response calculations
and further study is needed in this area. Gust amplitude has direct effect
on the responses. The selection of the correct gust velocity is important
in the response analysis. The gust phase angle will not significantly af-
fect the response. The position of the aerodynamic centers relative to
the mass centers is very important in the calculation of the torsional
displacements.

ACKNOWLEDGMENT

The writers are grateful for partial support from National Science
Foundation Grant CEE-8206909 a n d to Dr. John B. Scalzi, program di-
rector. Special thanks to Joel I. Abrams for reviewing this paper a n d for
his suggestions.

APPENDIX I.—REFERENCES

1. Craig, R. R. Jr., Structural Dynamics, An Introduction to Computer Methods, John


Wiley and Sons, New York, N.Y., 1981.
2. Davenport, A. G., "The Spectrum of Horizontal Gustiness Near the Ground
in High Winds," Quarterly Journal Royal Meteorological Society, Vol. 87, Apr.,
1961, pp. 194-211.
3. Davenport, A. G., "The Buffeting of Structures by Gusts," Proceedings of the
First Conference on Wind Effects on Buildings and Structures, London, England,
1963, pp. 357-392.
4. Hart, G. C., Dijulio, R. M., and Lew, M., "Torsional Response of High-Rise
Buildings," Journal of the Structural Division, ASCE, Vol. 101, No. ST2, Feb.,
1975, pp. 397-416.
5. Patrickson, C. P., and Friedmann, P., "A Study of the Coupled Lateral and
Torsional Response of Tall Buildings of Wind Loads," Report UCLA-ENG-
76126, University of California, Los Angeles, Calif., Dec, 1976.
6. Patrickson, C. P., and Friedmann, P. P., "Deterministic Torsional Building
Response to Winds," Journal of the Structural Division, ASCE, Vol. 105, No.
ST12, Dec, 1979, pp. 2621-2637.
7. Pipes, L. A., Applied Mathematics for Engineers and Physicists, McGraw Hill,
New York, N.Y., 1974.
8. Simiu, E., and Scanlan, R. H., Wind Effects on Structures, An Introduction to
Wind Engineering, John Wiley and Sons, New York, N.Y., 1978.
9. Torkamani, M. A. M., and Meyers, A. J., "A Comparison of Modeling As-
sumptions for the Dynamic Analysis of Reinforced Concrete Buildings," Re-
search Report SETEC-CE-81-057, University of Pittsburgh, Pittsburgh, Nov.,
1981.
10. Torkamani, M. A. M., and Pramono, E., "Dynamic Response of High-Rise
Building Subject to Wind Excitation," Research Report SETEC-CE-84-002, Uni-
versity of Pittsburgh, Pittsburgh, Sept., 1983.
11. Vaicaitis, R., Shinozuka, M., and Takeno, M., "Response Analysis of Tall
Buildings to Wind Loading," Journal of the Structural Division, ASCE, Vol. 101,
No. ST3, Mar., 1975, pp. 585-600.

823

J. Struct. Eng. 1985.111:805-825.


12. Vellozzi, J., and Cohen, E., "Dynamic Response of Tall Flexible Structures
to Wind Loading," Building Science Series 30, Wind Loads on Buildings and
Structures, National Bureau of Standards, Washington, D.C., 1970, pp. 115-
128.
13. Vickery, B. J., and Kao, K. H., "Drag or Along-Wind Response of Slender
Structures," Journal of the Structural Division, ASCE, Vol. 98, No. ST1, Jan.,
1972, pp. 21-36.
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

14. Vivekananda, M., "Wind Excited Vibration of Square Section Beam and Sus-
pended Cable," thesis presented to the Department of Aeronautics and
Astronautics, Massachusetts Institute of Technology, at Cambridge, Mass.,
in 1972, in partial fulfillment of the requirements for the degree of Doctor of
Engineering.
15. Wilson, E. L., Holling, J. P., and Dovey, H. H., "Three Dimensional Anal-
ysis of Building Systems," Report No. EERC 75-13, University of California,
Berkeley, Calif., Apr., 1975.
16. Yang, J. N., Lin, Y. K., and Samali, B., "Coupled Motion of Wind-Loaded
Multi-Story Building," Journal of Engineering Mechanics Division, ASCE, Vol.
107, No. EM6, Dec, 1981, pp. 1209-1225.

APPENDIX II.—NOTATION

The following symbols are used in this paper:

A = exponential coefficient d e p e n d i n g o n terrain;


A = transformation matrix;
b = width of building;
C = d a m p i n g matrix;
C„ = aerodynamic damping;
Cav = time averaged aerodynamic damping;
Cx{t),Cy(t) = aerodynamic coefficients;
CD(a,t),CL(a,t) = drag a n d lift experimental coefficients;
C,„ = virtual mass coefficient (inertial force associated
with mass of fluid displaced by body, propor-
tional to that mass, acceleration and coefficient C,„);
D(z, t) = aerodynamic drag force per unit length of build-
ing;
D1,D2,D3 = squared velocity terms defined b y Eqs. 6, 7, a n d
8;
F(t) = aerodynamic forces after linearization;
h = height of building;
K = stiffness matrix of building;
L = identity matrix;
L(z,t) = aerodynamic lift force per unit length of building;
M = mass matrix of building;
PAC = force matrix containing Px and Py, acting at aero-
dynamic center;
Px(z,t),Py(z,t) = distributed forced defined by Eqs. 14 a n d 15;
Q = applied load vector;
QAC = force matrix that contains concentrated load at
aerodynamic center at each floor level;
=
QMC force matrix that contains concentrated load at mass
center at each floor level;
T = period;
824

J. Struct. Eng. 1985.111:805-825.


H,U,8 = displacement coordinates in x, y and 0 directions;
U AC = displacement matrix at aerodynamic center as dis-
crete system;
UAC = displacement matrix at aerodynamic center as
continuous system;
ug(t),vg(t) = gust components along and across mean velocity;
Downloaded from ascelibrary.org by University Of Ulster AT on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

ug0,vg0 = constant gust velocity amplitudes;


^grad = gradient mean velocity;
V(z), also V = mean wind velocity;
Vrel = relative wind velocity, referred to xyz coordinate
system;
=
*AC /]/AC X and y distance from mass center to aerodynamic
center of each floor;
Y = generalized displacements vector;
Zgrad = gradient height;
a = angle of attack of relative velocity Vrel;
p = angle of attack of mean wind velocity;
TAC = matrix that contains shape function of building as
continuous system at aerodynamic center;
=
TMC matrix that contains shape function of building as
continuous system at mass center;
7 = effective angle of approach (P + 8);
8 = fluctuation of angle of attack a;
6 = torsional displacement;
p = air density;
<j> = mode shapes matrix;
(j> = phase angle;
^ = shape function of building as continuous system;
and
o) = frequency in radians per second.

825

J. Struct. Eng. 1985.111:805-825.

You might also like