Vous êtes sur la page 1sur 15

Lateral strength distribution

specification to limit
additional inelastic
deformation of torsionally
unbalanced structures
W. K. Tso and Hongshan Ying
Department of Civil Engineering and Engineering Mechanics, McMaster University,
Hamilton, Ontario, Canada
(Received March 1990; revised July 1990)

Using a single mass three-element model, a study is made on the


effect of strength distribution among elements on the inelastic
seismic responses of eccentric systems. Additional ductility demands
on elements and additional edge displacements are taken as response
parameters of interest in optimizing the strength distribution. It is
shown that for a given eccentricity, the stiffness eccentric system
and the mass eccentric system represent two extreme cases. The
inelastic responses of more general eccentric systems are bracketed
by the responses of these two extreme cases. To provide control of
additional ductility demands on the perimeter resisting elements, it is
necessary that the strength of the perimeter element at the opposite
side of the rigidity centre (as measured from the mass centre) is
suitably increased. In addition, the strength of the perimeter element
at the same side as the rigidity centre should not be decreased
drastically from its strength based on translational response alone. A
pair of design eccentricity expressions is suggested which can lead to
a strength distribution that limits the additional ductility demand
among elements to 20%. An approximate expression to estimate the
additional edge displacements due to structual asymmetry is
presented. This expression can be used in determining the separation
necessary between buildings to avoid pounding during an earthquake.

Keywords: seismic response, inelastic deformation, mathematical


modelling

Many buildings possessing pronounced asymmetry in


Mexico City suffered severe damage during the 1985
Michaocan earthquake I. Such damage might be the
result of excessive torsional responses of the buildings
caused by structural asymmetry. As a result, there is
considerable interest in understanding the relationship
.between inelastic seismic torsional responses and structural asymmetry of eccentric, or torsionally unbalanced
structual systems. A number of studies have been
reported on the inelastic torsional responses of single
mass torsionally unbalanced structural systems 2-7 as a
first step to understanding the seismic inelastic torsional
response problem of multistorey buildings. In all such

studies, the responses of the eccentric system are compared to those of a similar, but symmetric system.
Eccentricity, defined as the distance from the centre of
rigidity to the mass centre, is the commonly used
measure of asymmetry to distinguish the eccentric
system from its associated symmetric system.
Eccentricity, defined in this manner, is basically a
representation of the stiffness distribution relative to the
mass distribution and is a measure of structural asymmetry in the elastic range. For this reason, it will be
referred to as elastic eccentricity in this paper. Studies
have shown that this measure of asymmetry does not
correlate well with the inelastic torsional responses

014 I-O296/92/040263- ! 5
1992 Butterworth-Heinemann Ltd

Eng. Struct. 1992, Vol. 14, No 4

263

Inelastic deformation of torsionally unbalanced structures: VV. K. Tso and Hongshan Ying

when the structure is responding well beyond yield,


Recognizing the importance of strength distribution, the
concept of using resistance eccentricity was suggested
by Sadek and Tso 8. Resistance eccentricity is a
measure of uneven distribution of strength in relation to
mass distribution of the system. Their findings indicated
that the severity of inelastic torsional responses are well
correlated with the resistance eccentricity of the system,
For the particular eccentric system studied, the inelastic
torsional responses decrease with the system's resistance
eccentricity. Using a single mass three-element structural model, Tso and Ying 9 reconfirmed the desirability
of having small resistance eccentricity to minimize
torsional inelastic responses.
On the other hand, a recent study by Gomez et al. ~o
using a similar single mass three-element model, concluded that systems having resistance eccentricity much
smaller than the elastic eccentricity can lead to excessive
ductility demand on the resisting elements. Their view
is reflected in the current Mexican code ~] which
requires that ductile eccentric buildings be designed with
resistance eccentricity not drastically different from its
elastic eccentricity value. Such an apparently contradictory conclusion on the effect of resistance eccentricity on
inelastic torsional responses deserves clarification if
progress is to be made towards understandings the
inelastic seismic response of eccentric buildings.
This paper has three objectives. First, an overall view
on the effect of strength distribution on inelastic seismic
responses of the eccentric system is presented. It shows
the vulnerable areas for stiffness eccentric systems are
different from those for mass eccentric systems. As a
result, the apparently contradictory conclusions reached
by Tso 9 and Gomez et al. ]o on the desirable strength
distribution for eccentric systems based on the resistance
eccentricity concept is resolved as Tso used a stiffness
eccentric model while Gomez et al. used a mass eccentric model in their studies. Secondly, the necessary
ingredients to ensure a good control of additional ductility demand on the resisting elements for both the stiffness eccentric and mass eccentric systems are identified.
A procedure to reach a desirable strength distribution for
general eccentric systems is then presented in the form
of a pair of design eccentricity expressions. Inelastic
response calculations using eight earthquake records of
similar frequency content show that one can control the
additional ductility demand on the resisting elements to
no more than 20% of the corresponding ductility
demand in a symmetrical system. Thirdly, the additional
displacements of eccentric systems are less sensitive to
their strength distribution. Therefore, a simplified
expression can be developed to evaluate the additional
displacements for general eccentric systems. Such an
expression is presented for design use.

Notation
A
a*

b
di
e
e*

em

e*

264

peak ground acceleration


characteristic design acceleration
dimension of slab normal to ground motion
direction
distance of element i from centre of rigidity
eccentricity
e/b = normalized eccentricity
mass eccentricity
em/b = normalized mass eccentricity

Eng. Struct. 1992, Vol. 14, No 4

e~
resistance eccentricity
e*
er/b = normalized resistance eccentricity
stiffness eccentricity
es
e*
es/b = normalized stiffness eccentricity
(ed)a, (ed)b design eccentricity
nominal design strength of model
F
torsional shear in element i
(Fe),in-plane strength of element i
f.
lateral stiffness of model
K
torsional stiffness of model
Ke
in-plane stiffness of element i
ki
mass of rigid slab
m
strength reduction factor
R
structural period
T
codified design torsional methods
Ta, Tb
critical design torsional moments
T*
location of mass centre
Xm
location of rigidity centre
xs
displacement of asynunmetrical system
ma
displacement of associated asymmetrical
As
system
yield displacement of element 1 in asym(t~yl)a
metrical system
yield displacement of element 1 in
(6y~)s
associated symmetrical system
yield displacement of element 3 in asym(6y3)s
metrical system
yield displacement of element 3 in
(6y3)s
associated symmetrical system
ductility of asymmetrical system
/za
ductility of associated symmetrical system
/zs
coefficient to specify design eccentricity
X
mass radius of gyration about centre of
p
rigidity
uncoupled torsional to lateral frequency
fi
ratio

Statement of problem
The same structural model used by Tso and Ying 9 and
Gomez et al. 1o is adopted in this study. The structural
model consists of a slab of mass m, supported by
mree massless lateral load resisting elements, 1, 2,
and 3 spanning in the Y direction. Elements 1 and 3 are
located at equal distance b/2 from, but at opposite sides
of element 2. The system has the X axis as the axis of
symmetry, and it is subjected to ground motions in the
Y direction only. Each resisting element has bilinear
hysteretic force-deformation characteristics with a
post-yield stiffness equal to 3% of its initial stiffness.
The elements have in plane stiffness ki and yield
strength f/, (i = 1, 2, 3), while the out-of-plane stiffness and strengths of these elements are assumed to be
negligible.

Symmetrical (reference) model


In this model, the elements have the same stiffness, and
same strength. In addition, the mass centre is assumed
to coincide with element 2. Such an arrangement makes
the centre of rigidity, centre of resistance, and mass
centre, all coincide with element 2 and the model is
symmetrical in both the elastic and inelastic range, as
shown in Figure 1 (a). The responses of this symmetrical
model will form the basis of comparison for other

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying

responses obtained from the various eccentric models


studied in the paper.
The total strength of the system is related to the
nominal design strength F, determined from the following relation

F = ma*/R

(1)

in which a* denotes a characteristic acceleration and R


is a modification factor. In this study, a* takes the form

,'~

__

Mean

I
0.

0.2

I
0.q

0.6

I I I
0.8 1.0

2.

Period (s)

~1

b/2

}i~

Figure 2

Mean spectrum
N e w m a r k - Hall s p e c t r u m

b/ 2

CS CM
-------~ x

(9

(1)

for 8 earthquakes

records and

of the 5% damped Newmark-Hall spectrum ~2. It is


constant for a structural period T less than 0.5 s and
varies inversely proportional to the period when T
exceeds 0.5 s. When normalized with the peak ground
acceleration A, it has the appearance shown in Figure 2.
R can be considered as a reduction factor from elastic
strength demand. Structures designed with R _< 1 are
expected to remain essentially elastic, while a structure
designed with R > 1 will respond beyond its elastic limit
when excited by ground motions having spectra similar
to a*.

Eccentric systems

CS

Starting with the symmetrical model, there are a variety


of ways of creating eccentric systems. If xx and xm
denote the locations (to some fixed reference) of the
rigidity centre and mass centre, respectively, the eccentric e, by definition, is given by

tC M

~X

e = Xm-- xs

x
I

em

=='-I

Ground

system;

(b), SES; (c), M E S

One way to create an eccentric system is to modify the


stiffness distribution among the elements without changing the mass distribution. The resulting eccentric
systems are referred to as 'stiffness' eccentric systems
(SES), Another way to obtain the same eccentricity is to
keep the stiffness distribution the same, but to alter the
mass distribution and hence the location of the mass
centre. Systems created in such a way are referred to as
'mass' eccentric systems (MES). The SES and MES are
but two out of an infinite number of ways that equation
(2) can be satisfied for a given e and are illustrated in
Figures l(b) and 1 (c). In the elastic range, the. overall
response of an eccentric system is independent of the
way the eccentricity of the system is achieved. In
particular, the same elastic response will result irrespective of whether the eccentric system considered is to be
stiffness eccentric or mass eccentric. However, for
inelastic responses, the SES and MES have different
vulnerable areas and it is essential that a distinction be
made between them.
Stiffness eccentric system (SES)

motion

Figure 1 T h r e e e l e m e n t s i n g l e m a s s m o d e l s .

(2)

(a), s y m m e t r i c

Taking the reference Y axis to pass through element 2 in


the model, the locations of the mass centre and centre of

Eng. Struct. 1992, Vol. 14, No 4

265

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying

The loads on and hence strength of the elements of f


(i = I, 2, 3) can be determined as follows

rigidity in such a system are given by


x,,, = 0

(3a)

x~ -

(3b)

(9)
2K

where K = ~k i = lateral stiffness of the model.


Using equation (2), the eccentricity e is given by
e=e~ -

b(kl - k 3 )
2K

(4)

The subscript s in e~ denotes that the eccentricity is


caused by a shift of stiffness among the elements. In
addition to system eccentricity e~, the other system
parameters which affect the elastic dynamic responses
are the lateral period T, and the uncoupled torsional to
lateral frequency ratio f~ of the system. These
parameters are related to the elements' stiffness b y

/ m \ 1/2

[2= 1 ( ~ ) 1/2
-P

(6)

where Ko is the torsional stiffness of the model and p is


the radius of gyration of the deck about the rigidity centre. For simplicity, it is further assumed in this study
that elements 2 and 3 have the same stiffness
k2 = ks

(7)

The stiffness of the elements ki(i = 1, 2, 3) and the mass


radius of gyration O are adjusted in the system so that
they satisfy the specified values of es, T and fl according to equations (4)-(7).

(A) Stiffness eccentric proportional model


F

3~ =-~- (1 + 4e*)
F
f2 = ~- (1 - 2e*)

M a s s eccentric system ( M E S )
A mass eccentric system is derivable from the symmetric model by changing the location of the mass
centre, keeping the stiffness of the elements unchanged.
With reference to the coordinate system adopted,
xs = 0, and x,, is specified. From equation (2), the
resulting eccentricity is given by
e = e,, = x,,

The first term in equation (9) represents the translational


shear on element i, assuming that the roof deck has rigid
diaphragm action. The second term (Fe)~ is the torsional
shear on element i caused by the rotation of the deck.
To illustrate the differences in strength distribution for
SES and MES, two approaches to obtain the strength of
the elements will be considered. In the first approach,
we shall neglect the torsional shear component in equation (9). As a result, the strengths of the elements
become proportional to their stiffness and this approach
will be referred to as the 'proportional model' for
strength distribution. Such a strength distribution
implies that all elements have the same yield displacement. The second approach is based on the requirement
that the forces m the elements maintain static
equilibrium of the deck when the resultant of the applied
lateral loading passes through the mass centre. This
approach of strength distribution is referred to as the
'equilibrium model' in this paper.
By distributing strengths according to one of the two
ways mentioned above, one obtains four structural
eccentric models, namely, (A) a 'stiffness eccentric proportional' model, (B) a 'stiffness eccentric equilibrium'
model, (C) a 'mass eccentric proportional' model and
(D) a 'mass eccentric equilibrium' model. The strength
distribution of these models and its associated resistance
eccentricity e , can be expressed in terms of the
nominal design strength F and the normalized elastic
eccentricity of the system as follows

(8)

The subscript m in e,, denotes the eccentricity of the


system caused by the change of mass distribution. The
actual value of ki and the radius of gyration about the
rigidity centre of this system can be adjusted to satisfy
the specified values of period of T and frequency ratio
as defined by equations (5) and (6).

(10)

F
f3 = ~- (1 - 2e~*)
b(f~ - ~ )
e r

.=

--

es

(B) Stiffness eccentric equilibrium model


f l _ F ( 1 +4e*) [1
3
fz_F(1-2e*)

3e* ]
1 + 3e*J

6(e*) 2
]
1 + (1 - 2e*)(1 + 3e*)
(11)

Strength specifications
In design, the lateral strength of the elements are
designed based on the loads on the elements. Therefore,
the strength distribution is a function of the stiffness
distribution and the elastic eccentricity of the system.

266

Eng. Struct. 1992, Vol. 14, No 4

f3

er -

F(1 - 2e*) [
3e*(1 +2e*)
]
~
_ 1 + (1 ---2ee~*)O + 3 e * i J
b(f~ - f3)

2r.f

-0

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying


the SES. Gomez et al. 1o used a model similar to models
C and D for their study of inelastic torsional responses
in MES.
In terms of system parameters description, there are
some similarities in the models described. For example,
all models have a total strength independent of eccentricity, which is equal to the nominal design strength F.
Also, systems with strengths distributed proportional to
stiffness have resistance eccentricity equal to their
elastic eccentricity while systems with strengths
distributed based on equilibrium criteria have zero
resistance eccentricity. These similarities, however, are
deceptive if one considers the individual elements. The
strengths of individual elements are, in general, a function of the eccentricity. Its variation with eccentricity for
models A and B is shown in Figure 3. In model A the
strength of element 1 increases, while those of elements
2 and 3 will have very low strength when the eccentricity becomes large (say e, = 0.3b). The dependence
of strengths on eccentricity for model B is more complex. It has the effect of moderating the increase in
strength in element 1, and more importantly, reversing
the decreasing trend of strength in element 3.
In the mass eccentric model C the element strength
remains unchanged, resulting in a uniform strength
distribution among elements irrespective of the mass
eccentricity of the system, as shown in Figure 4(a). The
strength distribution variation of model D appears to be
the mirror image of model A. In model D, the strength
of element 3 increases, but that of element 1 decreases
linearly with the mass eccentricity. The rate of strength
decrement of element 1 in model D is even more rapid
than that for element 3 in model A. As a result, the
strength in element 1 in model D can reach a very low
value as shown in Figure 4(b).
From a comparison of elements' strength among the
four structural models, it is clear that using a single

(C) Mass eccentric proportional model


fl

F
3

F
f2 = 3
(12)
F
f3 = 3

b~ -A)
e r ~-- era

2~fi

-- e m

(D) Mass eccentric equilibrium model


F
f~ = ~- (1 - 3e*)

f2 = -

(13)

F
f3 = ~- (1 + 3e*)

e r ~

era

b~ -A) _ 0
2~f.

e* and e* denote the normalized stiffness and mass


eccentricity, respectively, using the distance b between
elements 1 and 3 as the normalizing constant.
Model A has been used by Bozorgnia and T s o 6 t o
evaluate the torsional inelastic responses of eccentric
systems. Tso and Ying 9 used both models A and B,
among others, to study the inelastic torsional response of

1.2

1.2

cz f;) / F

(z fi) / F

0.9

0.9

fl / F
0
+.,

0.6

0.6

01
-

fl

.......

f3/F

0.3

0.3

. . . . . . .

f2/F

r31F
0.0
0.0

0.1

0.2

0.3

Normalized

a
Figure 3

stiff,

eccen,

Model A (e r = es)

0.0

0.q

0.0

e~

0.1

0.2

0.3

Normalized stiff, eccen,

0.q

e~

Model B (e r = O)

Element strengths and total strength of SES. (a), stiffness proportional model; (b), equilibrium model

Eng. Struct. 1992, Vol. 14, No 4

267

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying


1.2

1.2

(~ ri) / F

(~. fi) / F
I

0.9

0.9
o

.9

.ioS

0.6

c~
t-

0.6

f3 IF

~I

+~

fl I F
f2/F

0.3

"~"
"'.,.,

0.3

f2 I F

f3/F
""
I

0.0

0.1

0.0

0.2

Normalized

0.3

mass eccen,

0.0

0.4

0.0

0.1

"1
0.3

mass e c c e n ,

I
0.4

em

Model D (er= O)

E l e m e n t s t r e n g t h s and total s t r e n g t h of MES. (a), s t i f f n e s s p r o p o r t i o n a l

system parameter such as resistance eccentricity er to


denote the strength distribution of an eccentric system
can be misleading. For a SES, er =es implies wide
disparity in strengths among elements while e, = 0
implies the strengths are more uniformly distributed. On
the other hand, in a MES, er = e = implies uniform
distribution of strengths and e~ = 0 implies large
disparity in strength among elements.

fl I F
I

0.2

Normalized

em

Model C (er= em)

Figure 4

m o d e l ; (b), equilibrium

model

Inelastic responses
Each of these four models with elastic eccentricity varies
between 0 and 0.3b was subjected to the 1940 E1 Centro
N - S component record as base input. The structural
period is taken to be 0.5 s., torsional to lateral frequency
ratio fl is taken tO be unity and the nominal strength of
the overall system is designed using a reduction factor

3.3q at e s = 0.3

3.0

/
/

3.0

/
:~

2.0

=~

2.0

-t

/ p
" "s "s p

"C

1.0

1.0

E3

0.0
0.0

0.1

0.2

Normalized

268

eccen.

Model A (e r = es}

Figure 5
element

stiff,

J
0.3

0.0

0.4

0.0

e;

Eng. S t r u c t

1992,

Vol. 14, No 4

0.2

0.3

Normalized

Ductility ratios of SES. (a), s t i f f n e s s p r o p o r t i o n a l m o d e l ; (b), equilibrium m o d e l ; (

I
0.1

Model

stiff,

eccen,

0.4

es

B (e r = O )

), e l e m e n t 1 ; ( - - - ), e l e m e n t 2; ( - - -- --),

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying


8.76 at e m* = 0 . 3

3.0

3.0

Ut

~.

2.0

2.0

.2

.~
(3

1.0

o.o

0.0

0.1

0.2

0.3

Normalized mass eccen,

1.0

o.o

0.4

0.0

0.1

0.2

0.3

Normalized mass eccen,

em

Model C (e r = e m)

4.0

/I
I I
i I
I
I

e,l

3.0

SIIIIIIIIII
111
/'~
2.0

Z4-#"

I.O

0.0
0.0

I
0.1

I
0.2

Normalized e c c e n t r i c i t y

I
0.3

0.4

e*

Figure 7 Displacement ratios of element 3. (


), model A ;
(---), model B; ( - - - ) , model C; ( . . . . .
), model D

e_*

Model D (e r = O ]

Figure 6 Ductility ratios o f MES. (a), stiffness proportional model; (b), equilibrium model; (
element 3

R = 5. For each model, the ductility and also the


maximum absolute displacement of each resisting element are taken as the response parameters of interest.
These parameters are normalized with the corresponding parameters from the symmetrical model, resulting in
ductility ratios and displacement ratios for each element.
These ratios have values beyond unity which imply
additional responses caused by structural asymmetry.

0.4

), element 1; (---), element 2; ( - - - ) ,

The ductility ratios of the three elements as a function


of normalized stiffness eccentricity for models A and B
are shown in Figure 5. If the strengths are distributed
such that er = e,, as in model A, substantial additional
ductility is required for element 3. However, the ductility ratios for all elements are less than 1.4 if the
strengths are distributed such that er = 0. Therefore,
distributing element strength such that the resistance
eccentricity is zero is a preferred way of strength
specification for SES, as pointed out by Tso and Ying 9.
The ductility ratio versus normalized eccentricity
plots for the two MES are shown in Figure 6. Model C
leads to higher ductility demand on element 3; but of
more concern is the exceptionally large additional ductility on element 1 for large mass eccentricity value in
model D. For MES, therefore, it is preferable to
distribute the strengths so that er = era, a conclusion
observed by Gomez in his studies.
Irrespective of the system being stiffness or mass
eccentric, the displacement ratio of element 1 for all four
models is less than or equal to unity, indicating the
displacement at element 1 is at most the same as that for
the symmetrical system. The displacement ratio at element 3 represents the critical displacement to be considered in design, as shown in Figure 7. The order of
displacement increase is less affected by the strength
distribution although the additional displacements at element 3 in MES are less than the corresponding
displacements in SES.
To understand the very high ductility demand on element 3 in model A and on element 1 in model D, it is
instructive to relate the yield displacements of these
elements to the maximum dynamic displacement
responses Ai at these locations. The dynamic displace-

Eng. Struct.

1992,

Vol. 14, No 4

269

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying


3.0

3.0

t
//
//
/

t~

2.5

2.5

//
/

,.o
/
/ '

%
~,

2.0

2.0

,o

~o

1.5

1.5

11

E
0
1.0

"0

"$
>-

j.4f

t)

1.0
Q.

....

"'<72

0.5

.v
>-

0.0
0.0

0.1

0.2

0.3

Normalized e c c e n t r i c i t y

0.5

o.o
0.4

e~, e ~

Figure 8

0.0

0.1

0.2

0.3

Normalized e c c e n t r i c i t y

Yield displacement ratios of models A , B , C and D. (a), for element 1; (b), for element 3. (
model C; . . . . ), model D

ment at element 1 of model D is no larger than the corresponding displacement in the symmetric system.
However, the yield displacement of element 1 in this
model becomes very low at large eccentricity, as shown
in Figure 8(a). Therefore, the very large ductility on element 1 in model D is not the result of large displacement
due to torsional responses, but due to low yield displacement of the element.
The large ductility demand on element 3 of model A
is due to other factors. From Figure 8(b), it can be seen
that the yield displacement of element 3 of model A is
not exceptionally low. In fact, it is the same as the symmetric case. But the dynamic displacement ratio at element 3 is substantially larger than unity as shown in
Figure 7. It is this large dynamic displacement at element 3 without a corresponding increase of yield
displacement of this element that leads to the very large
ductility demand of this element in model A. To prevent
excessive additional ductility demand in (the element 1
of) MES, one should not reduce the yield displacement
(and hence strength) of element 1 drastically. This can
be achieved by requiring er ~ era. To prevent large
additional ductility demand in (the element 3 of) SES,
one has to increase the yield displacement (and hence
strength) of element 3 of such a system. This would be
satisfied when er "~- O.
The essence to control the additional ductility demand
for MES is not to reduce the element 1 strength
drastically; and to control additional ductility demand
for SES is to increase the element 3 strength (and hence
its yield displacement) to compensate for the increase in
displacement at this location. The values of resistance
eccentricity, er, whether it should be similar to the
elastic eccentricity (for the MES) or close to zero (for
the SES), is a consequence of such dement strength
adjustments. It is apparent that resistance eccentricity
alone is not a sufficient index for proper specification for
strength distribution applicable to all classes of eccentric
systems.

270

Eng. Struct. 1992, Vol. 14, No 4

0.4

e s, e ;

), model A; (---), model B; ( - - - ),

General case of asymmetry


The SES or MES represent two of many eccentric
systems that can be derived from the reference symmetric system. In general, one can modify both the stiffness and mass distribution of the symmetrical system to
result in an eccentric system with a given ~ e n t r i c i t y e.
The resulting system is partially stiffness eccentric and
partially mass eccentric when referred to the symmetrical system. In this section, it is shown that SES and
MES can be considered to represent two extreme cases
of eccentric system for a given eccentricity e. The
responses of any general asymmetric system lie in
between these two extreme cases. This is illustrated
from the inelastic responses of a series of models having
different structural configurations. Each of the configurations has an elastic eccentricity value of 0.3b.
However, each configuration has a different mass and
stiffness distribution so that the location of their centre
of mass and centres of rigidity vary as shown in Figure
9. Configurations (a) and (d) correspond to SES and
MES which have been studied. Configurations (b) and
(c) represent the more general case where part of the
asymmetry is due to stiffness distribution change while
part is due to mass distribution change.
The stiffness of the elements of each configuration is
appropriately calculated such that they have a lateral
period of 0.5 s, torsional to lateral frequency ratio of
unity. A reduction factor R = 5 is used to evaluate the
design strength and the strength is distributed according
to both the proportional and the equilibrium criteria. The
1940 El Centro N - S component is taken as the ground
motion input. The ductility ratio of each of these configurations studied is shown in Figure 10. For systems
where the strength is distributed proportional to stiffness
of the elements, ductility demand in element 3 becomes
critical and the demand on this element is the highest for
the SES (configuration a) as shown in Figure lO(a). For
systems where the strength is distributed on the basis of

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying


Element

@
I..

I-"

b/2

._1_.

b/2

"-[-"

CS

._1
"l

CM

these configurations are shown in Figure 11. The stiffness accentric equilibrium model (model B) gives the
highest displacement demand.
The computation in this section snows that ~ES and
MES represent two systems which bracket the behaviour
of general eccentric systems. Conservative estimates of
the three design parameters of interest, namely, ductility
demands on elements 1 and 3, and dynamic displacement at element 3 location, can be determined using
either SES or MES. Specifically, a stiffness eccentric
proportional model (model A) leads to the largest ductility demand of element 3; a stiffness eccentric
equilibrium model (model B) leads to the largest
displacements at element 3 location; and a mass eccentric equilibrium model (model D) leads to the largest
ductility demand of element 1.

A compromise strength distribution


specification
Analysis in previous sections shows that no single
criteria of a strength distribution will lead to acceptable
additional ductility demand for both SES and MES. As
a result, the distribution of strength should be based on
the use of a pair of criteria. These would ensure sufficient strength in both elements 1 and 3 for any given
elastic eccentricity e (whether the system is stiffness
eccentric, mass eccentric, or somewhere in between).
These criteria can be expressed as a pair of design eccentricities (ea)a and (ed)b in the form

CM

CS

b
0.2b

O.lb

(ea) a = (1 + X)e + 0.1b


(ed)b = (1 -- X)e + 0. lb
CM

cs

(14a)
(14b)

w h e r e 0 _ h _ < 1.
The design torsional moments Ta and Tb are given by

Ta = F(ed),

(15a)

Tb = F(ed)b

(15b)

The torsional shear on element i is then given by


0.1b

0.2/)

(Fo) i = ~

t~o

CS

CM

(kidi)

(16)

where di is the distance from the ith element from the


centre of rigidity. T* is chosen to be equal to either Ta
or Tb, whichever produces a more severe strength
demand to the element concerned as defined in equation

(9).

d
Figure 9 Four structural configurations with elastic eccentricity
e = O.3b. (a), SES; (d) MES

equilibrium criteria, the ductility demand on element 1


represents a critical parameter, and the demand is
highest for a MES (configuration d) as shown in Figure
lO(b). The displacement ratio of element 3 in each of

The pair of eccentricity expressions given in equation


(14) take the familiar form that appears in many seismic
code provisions. In particular, h = 0 corresponds to the
New Zealand code formula j3 while h = 0 . 5 corresponds to the recommendation of the Canadian
code 14. In each expression the first term is related to
the structural asymmetry of the system and the second
term is the accidental eccentricity to allow for a number
of factors which cause torsion, but are difficult to be
quantified individually. The design expression (14a)
controls the design strength of element 3, while design
expression (14b) normally governs the design strength
of element 1. The design eccentricity expressions are

Eng. Struct. 1992, Vol. 14, No 4

271

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying


8.76 for c o n f i g u r a t i o n d
4.0

4.0

3.0

3.0

.9

2.0

2.0

I
u

~3
1.0

1.0

0.0

a
a

0.0

Configuration

c
Configuration

Figure 10 Ductility ratios of f o u r structural configurations; (a), stiffness proportional model; (b), equilibrium model; ( - 1; (

zx

Inelastic responses of systems designed


according to proposed design eccentricities
To evaluate the effectiveness of the proposed design
eccentricity expressions to limit excessive additional

4.0

3.0
<3

o
2.0

e~

1.0

0.0

I
b

I
c

Configuration

Figure 1 1 Displacement ratio of element 3 for 4 structural configurations. ( - - O - - ) ,


ness proportional

272

O --),

element

), element 3

presented in such a form that as X increases, one would


further increase the design strength of element 3, and
simultaneously provide a more severe restriction on the
decrease of the design strength of element 1.

E
==

static equilibrium; ( - - - & - - - ) stiff-

Eng. Struct. 1992, Vol. 14, No 4

ductility demands on the resisting elements for both SES


and MES, two groups of model considered. The strength
distributions of these models are based on the design
eccentricity expressions; the first group of models uses
X = 0, and the second group uses X = 0.5. Four structural periods equal to 0.2s, 0.5s, 1.0s, and 2.0s are used
to cover structural systems with short, medium and long
periods. The reduction factor R is taken to be 5.
To minimize the dependence of the results obtained on
a single earthquake record, an ensemble of eight earthquake records are used as ground motion input. The
criterion in selecting these records is that the spectral
shapes of their 5% damped elastic response spectra
should be similar to the design spectrum a*. Such a
choice of ground motion inputs will minimize the effects
of mismatch of frequency content between input ground
motions and design spectrum as a cause for poor control
of ductility demand on the elements. The mean 5%
damped spectrum for this ensemble of records, sealed to
a peak acceleration value A, is shown in Figure 2.
Details of these eight records are given in Table 1.
For the jth earthquake record input, the ductility
demand (/~a)j and the maximum absolute displacement
(A~)j at the three elements were determined. These element ductilities and displacements were then normalized
with the corresponding quantities (/~,)j and (A,)j,
calculated using the symmetric model having the same
lateral period and reduction factor, to form the ductility
and displacement ratios for this earthquake record. The
means of these ductility and displacement ratios were
then computed and used as response quantities of
interest.
For models with design strength based on h = 0, the
mean ductility ratio for element 1 for SES and MF_~ are
shown in Figures 12(a) and 12(b). There is some
dependence of these ratios on the structural periods.

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying


Table 1 Description of earthquake records

Record
no.

used

Epic.
dist.
(km)

Comp.

Max.
acc.
A(g)

Max.
vel.
V (m/s)

A/V

SOOE

0.348

0.334

1.04

Stiff
soil

Soil
cond.

Earthquake

Date

Magnitude

Site

Imperial Valley,
California

May 18
1940

6.6

El Centro

Kern County,
California

July 21
1952

7.6

Taft Lincoln
School Tunnel

56

$69E

O. 179

O. 177

1.01

Rock

Kern Cnunty,
California

July 21
1952

7.6

.Taft Lincoln
School Tunnel

56

N21E

O. 156

O. 157

0.99

Rock

San Fernando,
California

Feb. 9

6.4

Hollywood
Storage, Los
Angeles

35

N90E

0.211

0.211

1.00

Stiff
soil

San Fernando
California

Feb. 9
1971

6.4

234 Figueroas
St.. Los Angeles

41

N37E

O. 199

O. 167

1.19

Stiff
soil

Near E Coast of
Honshu, Japan

Nov. 16
1974

6.1

Kashima Harbour
Works

38

NOOE

0.070

0.072

0.97

Stiff
soil

Monte Negro,
Yugoslavia

Apr. 15
1979

7.0

Albatros Hotel,
Ulcinj

17

NOOE

O. 171

O. 194

0.88

Rock

Mexico earthquake

Sept. 19
1985

8.1

La Villita,
Guerrero Array

44

N9OE

O. 123

O. 105

1.17

Rock

governs the strength requirement for element 1. In view


of the large mean ductility ratio in MES, particularly
when the eccentricity e is large, it shows that the value
of (ed)b used allows too large a reduction of strength to
element 1. To restrict such reduction of strength, one
should decrease the coefficient of e in this expression
from unity. This will be achieved when using X = 0.5 in
the proposed design expression (14).
The major difference in the inelastic responses for
models with design strength based on the design eccentricities using X = 0.5 appears in the mean ductility
ratios for element 1. For both SES and MES, the mean

Shorter period structures tend to have a lower mean ductility ratio. For SES, the mean ductility ratios are in the
neighbourhood of unity, with a maximum value of 1.4.
Of more concern is the trend and 'magnitude of these
ratio for MES. They appear to be an increasing function
of the elastic eccentricity and in the worst scenario reach
a value of two. For element 3, the mean ductility ratios
never exceed 1.2 for both SES and MES, as shown in
Figures 13(a) and 13(b). Since the additional ductility
demand of element 3 is fairly well controlled in both
SES and MES, the design eccentricity expression (ea)a
is satisfactory. The expression (ea)b = e -- O. lb

q.O

4.0

:~ 3.0

:~

~3
-s

:~.

3.0

2.o

2.o
t

:~

1.0

i.t

1.0

%%,,.
0.0
0.0

0.1

0.2

0.3

Normalized stiff, eccen,

e;

0.0
0.q

0.0

Figure 12 Mean ductility ratios of element 1 (;k = O). (a), SES; (b), MES; (
(.....
), Tv= 2 . 0 s

0.1

0.2

0.3

Normalized
) T v = 0 . 2 s; ( -

- -)

mass e c c e n ,
T v = 0 . 5 s; ( -

0.4

em
-

-),

Tv =

Eng. Struct. 1 9 9 2 , Vol. 14, No 4

1 . O s,

273

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying


4.0

4.0

3.0

3.0

-1

-1

.9
2.0

u
"0
i-

1.0

1.0

0.0

I
0.0

0.1
Normalized

Figure 13

0.2

0.3

stiff,

eccen.

0.0
0.4

0.0

e;

0.1

0.2

0.3

Normalized mass eccen,

M e a n ductility ratios of e l e m e n t 3 (k = O). (a), SES; (b), M E S {Key as in

0.4

era

Figure 1 2 )

Additional inelastic edge displacement due to torsion

ductility ratios for this element do not exceed 1.2, as


shown in Figure 14. This is a significant improvement
over the performance of element 1, when the models are
designed based on X = 0. The improvement in further
reducing ductility ratio for element 3 is minimal. Since
the ductility ratios for element 3 are already satisfactory
when designed according to X = 0, such improvement is
not really necessary.
Similar computations have been carried out with system strengths design based on the reduction R = 3. Both
the mean ductility ratios and mean displacement ratios
for the different elements showed a similar trend as
systems designed using R = 5 is.

~.~

8 , i t -dl"m
- - - ~ :-"~'~
~ =I-IL
- ~-~"--~2

Observation after the 1985 Michaocan earthquake


showed a number of buildings that suffered severe
damge in Mexico City because of pounding between
adjacent buildings during the earthquake. To avoid
future occurrence of this type of damage, the edge
displacement of buildings needs to be evaluated on a
more realistic basis so that there will be adequate seismic
separation between buildings. Specifically, for torsionally unbalanced buildings, the additional edge
displacement caused by torsional responses should be
allowed for.

4.0

4.0

3.0

3.0
-1
o

2.0

2.0
u
~3
"O

u
"I0

:~

1.0

1.0

- - - - - - - - - --7

0.0
0.0

0.2

0.3

Normalized

Figure 14

274

I
0.1

stiff,

eccen,

es

0.0
0.4

0.0

M e a n ductility ratios of e l e m e n t 1 (X = O.5).(a), SES; (b), M E S (Key as in

Eng. Struct. 1992, Vol. 14, No 4

0.I

0.2

0.3

Normalized mass eccen,

Figure 12)

era

0.4

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying

-~

4.0

4.0

3.0

3.0

o
O

"

i.

2.0

Ao / As =

I + 3e ~

2.0

fll

:5
1.0

1.0
A a / A s = 1 + 3e*

o.o
0.0
a

0.1

0.2

0.3

Normalized s t i f f ,

eccen,

0.1

0.2

0.3

0.0
0.4

0.0

es

Normalized mass eccen,

0.4

v,.

em

Figure 15 M e a n d i s p l a c e m e n t ratios of e l e m e n t 3 (k = O). (a), SES; (b), M E S (Key as in Figure 12)

For the two groups of structural model designed using


k = 0 and k = 0.5, respectively, the mean displacement
ratio at element 1 location is less than unity. The additional displacements due to torsional responses occur at
the element 3 location. Figure 15 shows the mean
displacement ratios at element 3 location of SES and
MES designed using h = O, while Figure 16 shows
similar quantities for systems designed using h = 0.5. In
all cases, the displacement ratios exceed unity and are
monotonically increasing functions of the elastic eccentricity value. The mean displacement ratios associated
with SES are larger than MES. This can be expected

4.0

-~.

because for the same elastic eccentricity, the lever arm


from element 3 to the centre of rigidity in SES is longer
than that in MES. As a result, the same rotation about the
centre of rigidity will lead to a larger displacement at
element 3 location for SES. The mean displacement
ratios at the element 3 location are relatively insensitive
to the period of the structure, and also to the strength
distribution based on h = 0 or h = 0.5. Therefore, a
simple expression for the additional inelastic displacement applicable to a wide class of torsionally unbalanced
structures can be developed. It is suggested that the mean
displacement ratio at the element 3 location can be

4.0

3.0

"-.

0
.~

3.0

.o
A e / A s = 1 + 3e*

20

20

"0

1.0

AalA s =

0.o
0.0
a

Figure 16

0.1

0.2

0.3

Normalized s t i f f ,

eccen,

es

c
~

I +3e*

1.0

o.o
0.q

0.0

0.1

0.2

0.3

Normalized mass eccen.

0.q

em

M e a n d i s p l a c e m e n t ratios o f e l e m e n t 3 (k = O.5). (a), SES; (b), M E S (Key as in Figure 12)

Eng. Struct. 1992, Vol. 14, No 4

275

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying

approximated by the expression


A~/As = 1 + 3e*

(17)

where e* = e/b. Equation (17) is a compromise. It


slightly underestimates the displacement ratio for SES
and overestimates the same ratio for MES. The degree
of approximation can be judged from Figures 15 and 16.
Since SES and MES represent extreme circumstances, it
is believed the simple formula presented is of sufficient
accuracy to be used for the design of most torsionally
unbalanced structures.
To evaluate the critical seismic edge displacement of
a torsionally unbalanced structure, one needs to estimate
the translational inelastic displacement As first. Guidelines are available j4'16 for such estimation. Once As is
found, equation (17) can be used to obtain Aa. /X
represents the critical seismic edge displacement and can
be used to establish the separation necessary to avoid
damage caused by pounding between buildings during a
major earthquake.

Conclusions
The following conclusions can be drawn based on this
parametric study of a single mass, three element
monosymmetric structural system subjected to seismic
ground motions.
(1)

For any given eccentric system, asymmetry can


arise due to the uneven distribution of stiffness,
and/or uneven distribution of mass. For understanding, and to develop guidelines for strength
distribution in eccentric systems excited beyond
yielding, it is convenient to study the seismic
inelastic behaviour of two special cases, namely,
SES and MES. The vulnerable areas are different
in SES and MES. The perimeter elements on the
opposite side of the centre of rigidity (measured
from the mass centre) as represented by element 3
in the three-element model can experience large
additional ductility demand for SES. The perimeter
elements on the same side as the centre of rigidity
(measured from the mass centre) as represented by
element 1 in the three-element model, can experience very large additional ductility demand in
MES. In both SES and MES, there will be significant additional displacement at the element 3 location.
(2) Guidelines based on the resistance eccentricity concept to distribute lateral strength of elements that applies to SES lead to poor response performance when
they are applied to MES, and vice versa. Therefore, using a single criterion based on resistance
eccentricity for strength specification for a wide
class of eccentric structures is unlikely to be effective.
(3) To provide control of additional ductility demands
on both elements 1 and 3 in a general eccentric
system, two steps need to be taken. Firstly, one
should ensure that the strength of element 3 is
suitably increased to allow for the inevitable larger
displacement at that element due to torsional effect.
Secondly, one should not allow drastic decrease of

276

Eng. Struct. 1992, Vol. 14, No 4

strength in element 1 from its strength based on


translational response alone.
(4) A pair of design eccentricities is proposed that
would achieve the desired strength distribution to
control additional ductility demand at the elements.
As shown by response calculations, to limit the
ductility demand on element 1, the parameter X in
the proposed design eccentricity expressions
should be larger than zero. Eccentric systems
designed based on the pair of design expressions
using X = 0.5, namely, (ed)" = 1.5e + 0.1b, and
(ed) b = 0 . 5 e 0.1b lead to good control of additional ductility demand on both elements 1 and 3.
Under the worst scenario, the additional ductility
demand is 20% of that of an associated symmetrical
system. It is believed that the proposed form of
design eccentricity expressions with X > 0 is a
viable means of strength distribution specification
to control additional ductility demand. The optimal
value of ~, to be adopted depends on the degree of
conservatism desired.
(5) Torsional responses always caused additional
inelastic displacements at the building edge which
is opposite to the centre of rigidity (as measured
from the mass centre). This additional inelastic
displacement can be substantial, particular for SES.
It is relatively insensitive to the structural period,
or the ~, value chosen for strength distribution. A
simple formula is presented to estimate this additional inelastic edge displacement due to torsion.
Together with the guidelines provided by building
codes to estimate inelastic translational displacement of buildings, this formula provides a convenient means for designers to estimate the seismic
edge displacement of torsionally unbalanced
buildings so that adequate separation between
buildings will be allowed for to avoid pounding
damage during a major earthquake.

Acknowledgement
The writers wish to acknowledge the support of the
Natural Sciences and Engineering Research Council for
Canada (NSERC) for the work presented herein.

References
1 Esteva, L. 'Earthquake engineering research and practice in Mexico
after the 1985 earthquake, Bull. of New Zealand Nat. Soc. for Earthquake Eng., 1987, 20, (3), 159-200
2 Irvine, H. M. and Kountrouris, G. E. 'Peak ductility demands in
simple torsionally unbalanced building models subjected to earthquake excitations', Proc. 7th Worm Conf. of Earthquake Eng., Istanbul, 1980, part 4, pp. 117-120
3 Yamazaki, Y. 'Inelastic torsional response of structures subjected to
earthquake ground motions', Report No. UCB/EERC-80/07, University of California, Berkeley, 1980
4 Kan, C. L. and Chopra, A. K. 'Torsional coupling and earthquake
response of simple elastic and inelastic systems', J. Struct. Div.,
ASCE, 1981, 107, (8) 1569-1588
5 Tso, W. K. and Sadek, A. W. 'Inelastic seismic response of simple
eccentric structures', J. Earthquake Engng, Struct. Dyn. 1985, 13,
(2) 255-269
6 Bozorgnia, Y. and Tso, W. K. 'Inelastic earthquake response of
asymmetric structures', J. Struct. Engng., ASCE, 1986, 112, (2),
383 - 4 0 0

Inelastic deformation of torsionally unbalanced structures: W. K. Tso and Hongshan Ying


7 Tso, W. K. and Bozorgnia, Y. 'Effective eccentricity for inelastic
seismic response of buildings', J. Earthquake Engng. Struct. Dyn.,
1986, 14, pp. 413-427
8 Sadek, A. W. and Tso, W. K. 'Strength eccentricity concept for
inelastic analysis of asymmetrical structures', J. Engng. Struct.,
1989, 11, (3), 189-194
9 Tso, W. K. and Ying, Hongshan, 'Additional seismic inelastic deformation caused by structural asymmetry', J. Earthquake Engng.
Struct. Dyn., 1990, 19, 243-258
10 Gomez, R., Ayala, G. and ]aramillo, D. 'Respuesta Sismica de
Edificios Asimetricos', Instituto de Ingenieria, UNAM, Mexico City,
May 1987
11 Gomez, R., Garcia-Ranz, F. 'The Mexico earthquake of September
19, 1985 - complementary technical norms for earthquake resistant
design, 1987 edition, Earthquake Spectra, 1988, 4, (3) 441-459

12 Newmark, N. M. and Hall, W. J. Earthquake spectra and design,


Earthquake Engineering Research Institute, Berkeley, California,
1982
13 New Zealand Standard NZS 4203:1984, 'Code of practice for general
structural design and design loadings for buildings', Standards
Association of New Zealand, Wellington, New Zealand, 1984
14 National Building Code of Canada, Subsection 4.1.9., National
Research Council of Canada, Ottawa, 1990
15 Ying, Hongshan, 'Torsional seismic inelastic response for stiffness
and mass eccentric systems', M.Eng. Thesis, McMaster University,
Hamilton, Ontario, Canada, 1989
16 Uniform Building Code, 'Section 2313: earthquake regulations',
International Conference of Building Officials, Uniform Building
Code, Whittier, Ca., 1988

Eng. Struct. 1992, Vol. 14, No 4

277

Vous aimerez peut-être aussi