You are on page 1of 19

Nonlinear Dynamic Analysis of Reinforced Concrete Building Structures

SHUNSUKE OTANI

Department of Civil Engineering, University of Toronto, Toronto, Ontario, Canada M5S 1A4
Received January 10,1979
Revised manuscript accepted March 3,1980

This state-of-the-art paper discusses nonlinear dynamic analysis of reinforced concrete


building structures. Nonlinear analysis of a reinforced concrete building is difficult (a)
because inelastic deformation is not confined at critical sections, but spreads throughout
the structure; and (b) because stiffness of the reinforced concrete is dependent on a strain
history.
The paper reviews the behaviour of reinforced concrete members and their
subassemblies observed during laboratory tests. Then different hysteresis and analytical
models of reinforced concrete members are reviewed, and their application to the
simulation of building model behaviour is discussed.
________________
Cet article traite de l’etat actuel des connaissances au sujet d’une analyse dynamique
non1ineare de structures de batiments en beton arme. L’analyse nonlineaire d’un
batiment en beton arme est difficile (a) puisque la deformation inelastique n’est pas
limitee a des sections critiques, mais s’etend a travers la structure et (b) parce que la
rigidite du beton arme depend d’une histoire de deformation unitaire.
Cet article fait la revue du comportement des membrures en beton arme et leurs
sous-assemblages obseves au cours d’essais en laboratoire. Ensuite.differents modeles
analytiques et hystesis de membrures en beton arme sont examines et leur application a
la simulation du comportement de modele de batiment est discutee.
Canadian Journal of Civil Engineering, Vol. 7、pp. 333-344 (1980)
[Traduit par la revue]

Introduction
Dynamic response of a structure can be caused by different loading conditions such as: (a) earthquake
ground motion; (b) wind pressure; (c) wave action; (d) blast; (e) machine vibration; and (f) traffic movement.
Among these, inelastic response is mainly caused by earthquake motions and accidental blasts. Consequently,
more research on nonlinear structural behaviour has been carried out in relation to earthquake problems. This
paper describes the research development in earthquake engineering.
Note that dynamic problems are different from static one in the following points: (a) inertial force; (b)
damping; (c) strain rate effect; and (d) oscillation (stress reversals). These need to be clarified in order to
analyze a structure under dynamic loading.
Dynamic characteristics up to failure cannot be identified solely through a dynamic test or a real structure
for the following reasons: (a) difficult to understand the behaviour due to complex interactions of various
parameters; (b) expensive to build a structure, as a specimen, for destructive testing; and (c) capacity of
loading devices insufficient to cause failure. Consequently, dynamic tests of real buildings are rather aimed
toward obtaining data (a) to confirm the validity of mathematical modelling techniques for a linearly elastic
structure; and (b) to obtain damping characteristics of different types of structures. A specifically designed
laboratory test becomes inevitable in order to complement the weakness of full scale tests and to study the
effect of individual parameters.

Damping
Any mechanical system possesses some energy-dissipating mechanisms, for example: (a) inelastic hysteretic
energy dissipation; (b) radiation of kinetic energy through foundation; (c) kinetic friction; (d) viscosity in
materials; and (e) aerodynamic effect. Such capacity or energy dissipation is vaguely termed “damping,” and is
most often assumed to be of viscous type because of its mathematical simplicity.
Damping capacity is often determined by the band width of the response curve during a sinusoidal
steady-state test. Figure 1 shows such acceleration response curves for a reinforced concrete building at
different excitation levels(Jennings and Kuroiwa 1968). Note the shift of resonant frequencies and the change
in amplitudes of damping with increase of excitation level despite low response amplitudes.

FIG. 1. Observed acceleration amplitudes from steady-state test (Jennings and Kuroiwa 1968).

Damping capacity is not a unique value of a structure, but it depends on the level of excitation. The
state-of-the-art does not provide a method to determine the damping capacity based on the material properties
and geometrical characteristics of a structure.

Strain Rate Effect


It is technically difficult to test structural members under dynamic conditions in a laboratory. Before
force-deformation relations already obtained from thousands of static tests can be studied for use in a dynamic
analysis, the effect of strain rate on the force-deformation relation needs to be examined.
The speed of loading is known to influence the stiffness and strength of various materials (Cowell 1965,
1966). Some member test results are available (Mahin and Bertero l972). Important findings from these
investigations are as follows: (a) high strain rates increased the initial yield resistance, but caused small
differences in either stiffness or resistance in subsequent cycles at the same displacement amplitudes; (b) strain
rate effect on resistance diminished with increased deformation in a strain-hardening range; and (c) no
substantial changes were observed in ductility and overall energy absorption capacity.
Note that strain rate (velocity) during an oscillation is highest at low stress levels, and that the rate gradually
decreases toward a peak strain. Cracking and yielding of a reinforced concrete member reduce the stiffness,
elongating the period of oscillation. Furthermore, such damage is normally caused by the lower modes of
vibration having long periods. Therefore, the strain rate is small in the case of earthquake response, and its
effect on the response is small.
Consequently, the static hysteretic behaviour observed can be utilized in a nonlinear dynamic analysis of
reinforced concrete structures.

Stiffness Properties of Reinforced Concrete Members


It is not feasible to analyze an entire structure using microscopic material models. It is more important to
study the behaviour of isolated members and their subassemblies (beam-column, slab-column, and slab-wall
connections) so that their analytical models can be developed for use in the analysis of a complete structure.
A typical force-deflection curve of a cantilever column is shown in Fig. 2 (Otani et al. 1979). Note the
following observations: (a) tensile cracking of concrete and yielding of longitudinal reinforcement reduced the
stiffness; (b) when a deflection reversal was repeated at the same newly attained maximum amplitude (for
example, cycles 3 and 4) the loading stiffness in the second cycle was lower than that in the first cycle,
although the resistances at the peak displacement were almost identical; and (c) average stiffness
(peak-to-peak) of a complete cycle decreased with a maximum displacement amplitude. For example, the
peak-to-peak stiffness of cycle 5, after large amplitude displacement reversals, was significantly reduced from
that of cycle 2 at a comparable displacement amplitude. Therefore, the hysteretic behaviour of the reinforced
concrete is sensitive to loading history. Let us study some typical stiffness characteristics of reinforced
concrete.

FIG. 2. Hysteretic characteristics of reinforced concrete member (Otani et al. 1979).

Flexural Characteristics
The flexural deformation index (average curvature) is obtained from longitudinal strain measurements at
two levels assuming that a plane section remains plane. This flexural deformation index does not represent the
flexural deformation in a strict sense because a plane section does not remain plane in a region where an
extensive shear deformation occurs. However, the index is useful for understanding flexural deformation
characteristics qualitatively.
A typical moment-flexural deformation index curve obtained from a simply supported beam test (Celebi and
Penzien 1973) is shown.in Fig. 3. Note that the stiffness during loading gradually decreases with load, forming
a fat hysteresis loop, and absorbing a large amount of hysteretic energy. The hysteresis loops remain almost
identical even after several load reversals at the same displacement amplitude beyond yielding. Consequently,
vibration energy can be efficiently dissipated through flexural hysteresis loops without a reduction in
resistance. Many hysteretic models, as discussed later, are currently available to represent the nexural
behaviour.
FIG. 3. Flexural deformation characteristics (Celebi and Penzien 1973).

The increase in axial force decreases the flexural ductility of a reinforced concrete member, but increases
force levels corresponding to (a) tensile cracking of concrete; and (b) tensile yielding of longitudinal
reinforcement.

Shear Characteristics
Similar to the flexural deformation index, a shear deformation index is defined from strain measurements in
the two diagonal directions. Again, this index does not represent the true shear deformation because the
interference of shear and flexure exists.
A typical lateral load-shear deformation index curve (Celebi and Penzien 1973) is shown in Fig. 4. Unlike
what occurs in flexure, the stiffness during loading gradually increases with load, exhibiting a “pinching” in
the curve. The hysteretic energy dissipation is smaller. The hysteresis loop decays with the number of load
reversals, resulting in a smaller resistance at the same peak displacement in each repeated loading cycle.
Although the curve shows a “yielding” phenomenon, it is important to recognize that the shear force of the
member was limited by flexural yielding at the critical section rather than by yielding in shear. This yielding
clearly indicates the interaction of shear and bending.

FIG. 4. Shear deformation characteristics (Celebi and Penzien 1973).


The pinching in the force-deformation curve is obviously less desirable. The shear span to effective depth
ratio is the most significant parameter. Decreasing the shear span to depth ratio causes a more pronounced
pinching in the curve, and a faster degradation of the hysteretic energy-dissipating capacity. Considerable
improvements in delaying and reducing the degrading effects can be accomplished by using closely spaced ties.
Existence of axial force tends to retard the decrease in stiffness and resistance with cycles. However, it is hard
to eliminate this undesirable effect when high shear stress exists. Consequently, it becomes important to
include this degrading behaviour in a behavioural model for a short, deep reinforced concrete member. The
current state of knowledge is not sufficient to define the stiffness degrading parameters on the basis of the
member geometry and material properties.

Bar Slip and Bond Deterioration


When a structural element is framed into another element, some deformation is initiated within the other
element. Consider a beam-column subassembly. Bertero and Popov (1977) reported a significant rotation at a
beam end caused by the slippage (pullout) of the beam's main longitudinal reinforcement within the
beam-column joint (Fig. 5). The general shape of the moment-bar slip rotation curve is similar to that shown in
Fig. 4, demonstrating a pronounced pinching of a hysteresis loop. The contribution of bar slip to total
deformation cannot be neglected, especially in a stiff member (short or deep).

FIG. 5. Rotation due to bar slip (Bertero and Popov 1977).

Biaxial Lateral Load Reversals


During an earthquake, columns of a framed structure must resist lateral forces simultaneously in longitudinal
and transverse directions. Recent tests at the University of Toronto (Otani et al. 1979) on reinforced concrete
columns showed that the columns under lateral load reversals in two perpendicular directions exhibited the
decay of resistance and stiffness at a faster rate than those under uniaxial lateral load reversals. This topic
needs further study.

Hysteretic Models for Reinforced Concrete


Nonlinear dynamic analysis of a reinforced concrete structure requires two types of mathematical modelling:
(a) modelling for the distribution of stiffness along a member; and (b) modelling for the force-deformation
relationship under stress reversals.
A hysteresis model must be able to provide the stiffness and resistance under any displacement history. At
the same time, the basic characteristics need to be defined by the member geometry and material properties.
The current state of knowledge is sufficient to define flexural hysteresis models. However, it is not sufficient to
determine the degree of stiffness degradation due to the deterioration of shear-resisting and rebar-concrete
bond mechanisms.

Bilinear Model
The elastic-perfectly plastic hysteretic model was used by many investigators because the model was simple.
The maximum displacement of an elasto-plastic simple system was found (Veletsos and Newmark 1960) to be
practically the same as that of an elastic system having the same initial period of vibration as long as the period
was longer than 0.5 s.
A finite positive slope was assigned to the postyield stiffness to account for the strain-hardening
characteristic, and the model was called a bilinear model. The bilinear model does not represent the
degradation of loading and unloading stiffnesses with increasing displacement amplitude reversals (Fig. 6),
and the model is not suited for a refined nonlinear analysis of a reinforced concrete structure.

FIG. 6. Bilinear hysteresis model.

Clough 's Degrading Stiffness Model


A qualitative model for the reinforced concrete was developed by Clough (1966), who incorporated the
stiffness degradation in the elasto-plastic model : the response point during loading moved toward the previous
maximum response point. The unloading slope remained parallel to the initial elastic slope. This small
modification improved the capability to simulate the flexural behaviour of the reinforced concrete. Compared
with the elasto-plastic model, less energy is absorbed per cycle beyond yielding by Clough's degrading model.
From the response analysis of a series of single-degree-of-freedom systems, Clough (1966) concluded that
(a) the degrading stiffness model did not cause any significant change in the ductility demand of long-period
structures (period longer than 0.6 s) compared with the elasto-plastic model; on the other hand, (b) the
degrading stiffness model required significantly larger ductility from short-period structures than the
corresponding elasto-plastic systems; and (c) the response waveform of a degrading stiffness model was
distinctly different from that of an ordinary elasto-plastic model.
The model is relatively simple, and has been used extensively in nonlinear analysis with the inclusion of
strain-hardening characteristics (Fig. 7).
FIG. 7. Clough’s degrading stiffness model.

Takeda's Degrading Stiffness Model


A more refined and sophisticated hysteresis model was developed by Takeda et al. (l970) on the basis of
experimental observation. This model included stiffness changes at flexural cracking and yielding, and also
strain-hardening characteristics. The unloading stiffness was reduced by an exponential function of the
previous maximum deformation. Takeda et al. also prepared a set of rules for load reversals within the
outermost hysteresis loop. These are major improvements over the Clough (1966) model.
Failure or extensive damage caused by shear or bond deterioration was not considered in the model. The
Takeda model, similar to the Clough model, simulates dominantly flexural behaviour (Fig. 8). Simplified
Takeda hysteresis models were proposed by Otani and Sozen (1972) and by Powell (1975), using a bilinear
backbone curve.

FIG. 8. Takeda’s degrading stiffness models.


To test the goodness of the Takeda model, cantilever columns tested on the University of Illinois earthquake
simulator were analyzed (Takeda et al.1970). Calculated waveforms were favourably compared with the
observed waveform as shown in Fig. 9.

FIG. 9. Takeda model applied to column analysis (Takeda et al. 1970)

Takeda-Takayanagi Models
The amplitude of the exterior column axial load varies greatly due to the earthquake overturning moment,
and changes its moment-carrying capacity. Takayanagi and Schnobrich (1976) incorporated the effect of axial
force variation in the Takeda model by preparing various backbone curves at different axial load levels (Fig.
10a).

(a) axial force variation (b) pinching and strength decay


FIG. 10. Takeda-Takayanagi models (Takayanagi and Schnobrich 1976):

A pinching action and strength decay are inevitable in a short and deep member due to bar slip and
deterioration in shear resistance. Takayanagi and Schnobrich (1976) introduced a pinching action and strength
decay in the Takeda model (Fig. 10b). Whenever a response point was located in the positive rotation-negative
moment range or the negative rotation-positive moment range, the pinching was introduced. After the moment
exceeded the yield level, a strength decay was incorporated. The values of guideline for strength decay and
pinching stiffness were not related to the member geometry and material properties.

Degrading Trilinear Hysteresis Model


A model that simulates dominantly flexural stiffness characteristics was developed in Japan (Fukada 1969).
The backbone curve is a trilinear shape with stiffness changes at cracking and yielding. Up to yielding, the
model behaves in the same way as the bilinear model. Once deformation exceeds the yield point, the model
behaves as a perfectly plastic system. Upon unloading, the unloading point is treated as a new “yield” point,
and unloading stiffnesses corresponding to pre- and postcracking are reduced proportionately so that the
behaviour becomes of the bilinear type in a range between the positive and the negative yield points (Fig. 11).

FIG. 11. Degrading trilinear model.

The degrading trilinear model can easily include strain-hardening characteristics. The hysteresis energy
dissipation per cycle beyond the initial yielding is proportional to the displacement, and the equivalent viscous
damping factor becomes constant. The fatness of a hysteresis loop is sensitive to the choice of a cracking point.

Comments
Many other hysteresis models have been proposed and used in the past. Figure 12 shows attained ductility
factors of single-degree-of-freedom systems with any of four flexural hysteresis models: bilinear; Clough;
Takeda; and degrading trilinear models. The four models have the same backbone curve except the cracking
point. The four models show similar variations of attained ductility factors with periods, but attained ductility
factors show a wide scatter from one model to another, especially in a short-period range.
Fig. 12. Effect of hysteresis models on earthquake response.

Further research is necessary to study the effect of hysteretic characteristics on earthquake response, and
develop a simple standard hysteresis model for general use.

Reinforced Concrete Member Model


Inelastic deformation of a reinforced concrete member does not concentrate in a critical location, but rather
spreads along the member (Fig. 13). Various member models have been proposed to represent the distribution
of stiffness within a reinforced concrete member. The effect of gravity load on the beam behaviour and the
contribution of slabs to the structural stiffness will not be discussed.

FIG. 13. Deformation of beam under gravity and earthquake loads.

One-component Model
An elasto-plastic frame structure was analyzed by placing a rigid plastic spring at the location where
yielding is expected. The part of a member between the two rigid plastic springs remains perfectly elastic. All
inelastic deformation is assumed to occur in these springs (Fig. 14). This one-component model was
generalized by Giberson (1967).

FIG. 14. One-component member model.

A major advantage of the model is that inelastic member-end deformation depends solely on the moment
acting at the end so that any moment-rotation hysteretic model can be assigned to the spring. This fact is also a
weakness of the model because the member-end rotation should be dependent on the curvature distribution
along the member, hence dependent on moments at both member ends. Consider two cases of moment
distribution along a member AB with corresponding curvature distributions as shown in Fig. 15. The inelastic
rotations at the A end are given by the shaded areas. For the same moments at the A end, case II causes larger
inelastic rotation at the A end. Consequently, this simple model does not simulate actual member behaviour.
Furthermore, it is not rational to lump all inelastic deformations at member ends.

FIG. 15. Inelastic rotation of beam: (a) moment; (b) curvature and inelastic rotation.

The stiffness of an inelastic spring is normally defined by assuming an asymmetric moment distribution
along a member with the infection point at midspan. The usage of the initial location of the inflection point in
evaluating spring properties was suggested by Suko and Adams (1971). However, once yielding is developed
at one member end, the moment at the other end must increase to resist a higher stress, moving the inflection
point toward the member centre. At the same time, a large concentrated rotation starts to occur near the critical
section. Despite rational criticisms against this simple model, the performance of the one-component model is
expected to be reasonably good for a relatively low-rise frame structure, in which the inflection point of a
column locates reasonably close to midheight.
A special-purpose computer program, SAKE (Otani 1974), for a regular rectangular reinforced concrete
frame structure and recent modifications (Powell 1975) to general-purpose computer program DRAIN 2D
(Kanaan and Powell 1973) used the one-component model.
Multi-component Model
In an effort to analyze frame structures well into the inelastic range under earthquake excitation, an
interesting model was proposed by Clough et al. (1965). A frame member was divided into two imaginary
parallel elements: an elasto-plastic element to represent a yielding phenomenon, and a fully elastic element to
represent strain-hardening behaviour. When the member-end moment reaches the yield level, a plastic hinge is
placed at the end of the elasto-plastic element. A member-end rotation depends on both member-end moments.
Aoyama and Sugano (1968) adapted the two-component model, creating the multicomponent model (Fig.16),
using four parallel beams to account for flexural cracking, different yield levels at two member ends, and
strain-hardening. The deformation compatibility of the imaginary components is satisfied only at their ends.

FIG. 16. Multi-component member model.

The multi-component model appears to have merit; rotation at one end of a member depends on both
member-end moments. In other words, the moment distribution along a member can be approximately
reflected in the analysis. However, the stiffness of the multi-parallel components must be evaluated under a
certain assumed moment distribution. Therefore, the stiffness parameters are valid only under such a moment
distribution, and are bound to be approximate when the moment distribution becomes drastically different.
Giberson (1967) discussed the advantages and disadvantages of the one-component and the two-component
models, and concluded that the one-component model was more versatile than the two-component model
because the two-component model was restricted to the bilinear-type hysteresis characteristics. This
two-component model was used in a general-purpose computer program DRAIN 2D (Kanaan and Powell
1973) for plane structure. The interaction of the bending moment and the axial force was easily incorporated
by simply changing the yield value of the elasto-plastic component depending on the existing axial force.

Connected Two-cantilever Model


When a frame is analyzed under lateral loads only, the member moment distributes linearly. From the
similarity of moment distribution, the member can be considered to consist of two imaginary cantilevers, free
at the point of contraflexure and fixed at the member end, and connected at the inflection point, satisfying the
continuity of displacement and rotation (Otani and Sozen 1972).
The flexibility relation of a member was formulated by assuming: (a) the inflection point did not shift much
during a short time increment; (b) free-end rotation and displacement were proportional to the beam length and
the square of the beam length, respectively; and (c) instantaneous stiffness for shear-rotation and
shear-displacement curves of a unit length reference cantilever could be defined by hysteretic models.
The weakness of this method is that the member flexibility matrix is a function of the location of the
inflection point, which tends to shift rapidly when the sign of a member-end moment changes. This causes a
numerical problem. Consequently, this method cannot be recommended for a general dynamic analysis.
However, the method is useful for incremental static load analysis of a structure.

Discrete Element Model


In order to overcome difficult problems of variable stiffness distribution along a member, the member can be
subdivided into short line segments along the length, with each short segment assigned a nonlinear hysteretic
characteristic. The nonlinear stiffness can be assigned within a segment, or at the connection of two adjacent
segments.
Wen and Janssen (1965) presented a method for dynamic analysis of a plane frame consisting of
elasto-plastic segments. Consequently, the mass and flexibility of a member were lumped at the connecting
points on a tributary basis. Powell (1975) suggested using a degrading stiffness hysteresis model for rigid
inelastic connecting springs (Fig. 17a). Shorter segments were recommended in a region of high moment, and
longer segments in a low-moment region.

FIG. 17. Discrete element model: (a) lumped inelastic stiffness; (b) distributed inelastic stiffness.

An alternative method is to divide a member into short segments, each segment with a uniform flexural
rigidity that varies with a stress history of the segment (Fig. 17b). Local concentration of inelastic action can
be easily handled by arranging shorter segments at the location of high concentration of inelastic deformation
(Takayanagi and Schnobrich 1976).
These methods are useful when more accurate results are required, or in the analysis of walls. More
computational effort is required compared with the other simple models.

Distributed Flexibility Model


Once cracks develop in a member, the stiffness becomes nonuniform along the member length. Instead of
dividing a member into short segments, Takizawa (1973) developed a model that assumed a prescribed
distribution pattern of cross-sectional flexural flexibility along member length. A parabolic distribution with an
elastic flexibility at the infection point was used (Fig. 18). The flexura1 flexibility at the member ends was
given by a hysteretic model dependent on a stress history.

FIG. 18. Distributed flexibility model.


This is an interesting concept in analyzing an inelastic member. However, the parabolic flexibility
distribution may not describe the actual concentration of deformation at critical sections (normally at member
ends) due to flexural yielding and deformation attributable to slippage of longitudinal reinforcement within a
beam-column connection. The usage of inelastic springs at locations of concentrated deformation in
conjunction with this model may be a useful solution.

Summary
Various member models are reviewed, and their advantages and disadvantages are discussed. These models
have been developed specifically for earthquake response. Development of a simple model for simultaneous
gravity and earthquake situations is desired.

Reliability of Analytical Models


Earthquake simulator tests provide interesting opportunities to examine the goodness of different analytical
models in simulating the observed response of small- to medium-scale highly inelastic model structures. This
section reviews the reliability of different analytical models in relation to the capability to simulate the
observed behaviour.
These test structures were designed to behave dominantly in flexure, being prevented as much as possible
from failing in shear or anchorage because the two types of failure are not desirable in real construction and
are avoided in a design process.

Three-storey One-bay Frames (I)


Small-scale three-storey one-bay reinforced concrete frames were tested on the University of Illinois
earthquake simulator, and were analyzed using the connected two-cantilever member model (Otani and Sozen
1972). The stiffness properties of individual members were calculated on the basis of the geometry and
material properties. The Takeda model was used to represent the force-deflection of each cantilever model.
Member-end rotation due to bar slip was approximated by the simplified bilinear Takeda model, which did not
simulate the pinching behaviour. The damping matrix was assumed to be proportional to the instantaneous
stiffness matrix.

FIG. 19. Connected two-cantilever model applied to three-storey frame analysis (Otani and Sozen 1972):
(a) measured; (b) calculated (h=0.0); (c) calculated (h=0.02)
The model structure was subjected to a base motion simulating the El Centro (NS) 1940 accelerogram. The
first-floor displacement was measured to be as much as four times the yield displacement calculated under
static lateral loads. The analytical models with and without viscous damping favourably simulated the
large-amplitude oscillations at 1.0, 2.0, and 5 s from the beginning of the motion (Fig. 19). The analytical
models, however, failed to simulate the medium- and low-amplitude oscillations. Note that the frequencies at
the medium- to low-amplitude oscillations are higher for the analytical model, which indicates that the test
structure was more flexible at low stress levels than the analytical model. In order to reproduce lower
amplitude oscillations of the observed response waveforms, the pinching behaviour needs to be incorporated in
a hysteretic model.

Three-storey One-bay Frames (II)


Another set of three-storey one-bay small-scale reinforced concrete frame structures was tested on the
University of Illinois earthquake simulator (Otani 1976). The base motion is significantly more intense than a
design earthquake motion.
A member was represented by the one-component model with two inelastic rotational springs at each
member end: one for the flexural deformation and the other for the member-end rotation due to bar slip (Otani
1974). Takeda models with trilinear and bilinear backbone curves were assigned to the two inelastic springs.
Two types of damping were used in the analysis: (a) a damping matrix proportional to the constant mass
matrix; and (b) a damping matrix proportional to an instantaneous stiffness matrix. The first mode damping
factor was 5% of critical at the initial elastic stage. Observed and calculated third-level displacement
waveforms are compared in Fig. 20 (Otani 1976). The comparison is fair for large-amplitude oscillations, and
poor at low-amplitude oscillations. Again in this analysis, the pinching characteristic was not incorporated. A
fair agreement between the computed and the observed may be attributable to the fact that the yielding was
developed at most member ends at the large-amplitude oscillations, and that the inflection point tended to be
near the midpoint of each member in such a low-rise frame structure.

FIG. 20. One-component model applied to three-story frame analysis (Otani, 1976):
(a) measured; (b) calculated (mass proportional damping); (c) calculated (stiffness proportional damping).
Two-storey One-bay Frame
A two-storey one-bay medium-scale frame structure with slabs was tested on the University of California
earthquake simulator (Hidalgo and Clough 1974). The structure was analyzed using the two-component model.
In an effort to improve the correlation, the elastic stiffness of the two parallel components was degraded as a
function of the first-mode-response amplitude history. The observed and the calculated second-floor
displacement waveforms are satisfactorily compared in Fig. 21. However, the parameters controlling stiffness
degradation could not be determined from the theory.

FIG. 21. Two-component model applied to two-story frame analysis (Hidalgo and Clough 1974).

Ten-storey Coupled Shear Walls


Ten-storey coupled shear walls were tested on the University of Illinois earthquake simulator
(Aristizaba1-Ochoa and Sozen 1976). Takayanagi and Schnobrich (l976) divided a wall into short segments of
uniform stiffness, and represented connecting beams by the one-component model. The Takeda-Takayanagi
model with changing axial force was assigned to a wall element, and the Takeda-Takayanagi model with
pinching action and strength decay was used in a beam. It was judged that the usage of two-dimensional plane
stress elements for the walls was less desirable because such an approach might cost more computational effort
without any compensating increase in accuracy.

(a) displacement at level 10, in inches (1 in. = 25.4 mm) (b)Acceleration at level 10, g.
FIG. 22. Analysis of ten-story coupled shear wall (Takayanagi and Schnobrich 1976):

The comparison of the measured and calculated displacement and acceleration is excellent, as shown in Fig.
22. It is necessary to include the effects of inelastic axial rigidity of the wall section and pinching action and
strength decay of the connecting beams to reproduce the maximum displacement response and the elongation
of the period. Some stiffness parameters for the walls and connecting beams were defined on the basis of static
tests of connecting beam-wall assemblies.

Summary
The favourable comparison of the measured and the calculated response waveforms encourages the use of
correct analytical and hysteretic models. It is desirable in developing a mathematical model that all parameters
of the proposed model should be evaluated on the basis of the geometry of a structure and the properties of
materials.

Three-dimensional Building Analysis


The development of analytical methods has made it feasible to discuss nonlinear behaviour of reinforced
concrete plane structures with a certain confidence. However, we have not yet reached a point to discuss, with
any confidence, the nonlinear behaviour of three-dimensional building structures. Columns of a framed
structure must resist lateral forces in two horizontal directions. The stiffness is reduced significantly under
biaxial lateral load reversals (Otani et al. 1979).
The first nonlinear analysis of a frame structure under horizontal biaxial ground motion was made by Nigam
(1967) with elasto-plastic columns. Multi-storey frames with rigid beams and floor slabs under horizontal
biaxial ground motion were studied by Pecknold (1974). Prager's kinematic hardening theory as modified by
Ziegler (1959) was used. The basic effect of biaxial inelastic interaction is to produce a softer structure.
Pecknold (1974) confirmed Nigam's finding that: (a) horizontal biaxial ground motion increased ductility
demand for stiff structures (initial natural period less than 0.3 s); and (b) horizontal biaxial ground motion had
little effect on the ductility demand of flexible structures. Neither study included the stiffness degradation
property.
Aktan et al. (1973) used the finite element technique to include the stiffness degradation through the
degradation of material properties. The biaxial ground motion was found to cause 20-200% larger response
than the uniaxial ground motions from columns for which the calculated deflection under uniaxial ground
motion exceeded approximately twice the yield deflection.
Takizawa and Aoyama (1976) extended the one-dimensional degrading trilinear hysteretic model into a
two-dimensional model on the basis of plasticity theories (Ziegler 1959). The proposed model was judged to
predict the significant trends of the biaxial behaviour of reinforced concrete test columns. The effect of biaxial
response interaction was reported to be significant for degrading stiffness models, and not so important for
nondegrading type hysteretic models.
A brief review of nonlinear analysis of three-dimensional structures indicates the necessity of further study
in this area.

Summary
The behaviour of reinforced concrete buildings, especially under earthquake motion, was briefly reviewed.
When a structure can be idealized as plane structures, the current state-of-the-art provides useful and reliable
analytical methods.
However, more research is required to understand the effect of slabs, gravity loads, and biaxial ground
motion on nonlinear behaviour of a three-dimensional reinforced concrete structure.

References
AKTAN, A. E., PECKNOLD, D. A. W., and SOZEN, M. A. 1973. Effect of two-dimensional
earthquake motion on a reinforced concrete column, University of Illinois, Urbana, IL, SRS
No. 399.
AOYAMA, H., and SUGANO, T. l968. A generalized inelastic analysis of reinforced concrete
structures based on the tests of members. Recent researches of structural mechanics.
Contribution in Honour of the 60th Birthday of Professor Y. Tsuboi, Uno-Shoten, Tokyo, pp.
15-30.
ARISTIZABAL-OCHOA,J. D., and SOZEN, M. A. 1976. Behaviour of ten-storey reinforced
concrete walls subjected to earthquake motion, University of Illinois, Urbana, IL, SRS No.
431.
BERTERO, V. V., and POPOV, E. P. 1977. Seismic behaviour of moment-resisting reinforced
concrete frames. In Reinforced concrete structures in seismic zones, American Concrete
Institute, Special Publication No. 53, pp. 247-292.
CELEBI, M., and PENZIEN, J. l973. Experimental investigation into the seismic behaviour of
critical region of reinforced concret components as influenced by moment and shear,
Earthquake Engineering Research Center, University of California, Berkeley, CA, EERC 73-4.
CLOUGH, R. W. l966. Effect of stiffness degradation on earthquake ductility requirements,
Structural and Materials Research, Structural Engineering Laboratory, University of
California, Berkeley, CA, Report 66-16.
CLOUGH, R. W., BENUSKA, K. L., and WILSON, E. L. l965. Inelastic earthquake response of
tall buildings, Proceedings, 3rd World Conference on Earthquake Engineering, New Zealand,
Vo1. II, Session II, pp. 68-89.
COWELL, W. L. 1965. Dynamic tests of concrete reinforcing steels, U.S. Naval Civil
Engineering Laboratory, Port Hueneme, CA, Technical Report 394.
COWELL, W. L. 1966. Dynamic properties of plain Portland cement concrete, U.S. Naval Civil
Engineering Laboratory, Port Hueneme, CA, Technical Report 447.
FUKADA, Y. 1969. Study on the restoring force characteristics of reinforced concrete buildings
(in Japanese), Proceedings, Kanto District Symposium, Architectural Institute of Japan,
Tokyo, Japan, No. 40.
GIBERSON, M. F. 1967. The response of nonlinear multi-story structures subjected to
earthquake excitation, Earthquake Engineering Research Laboratory, California Institute of
Technology, Pasadena, CA, EERL Report.
HIDALGO, P., and CLOUGH, R. W. l974. Earthquake simulator study of a reinforced concrete
frame, Earthquake Engineering Research Center, University of California, Berkeley, CA,
EERC 74-13.
JENNINGS, P. C., and KUROIWA, J. H. 1968. Vibration and soil-structure interaction tests of a
nine-story reinforced concrete building, Bulletin of the Seismological Society of America, 58,
pp. 891-916.
KANAAN, A. E., and POWELL, G. H. l973. DRAIN-2D, A general purpose computer program
for dynamic analysis of inelastic plane structures, Earthquake Engineering Research Center,
University of California, Berkeley, CA, EERC 73-6.
MAHIN, S. A., and BERTERO, V. V. 1972. Rate of loading effect on uncracked and repaired
reinforced concrete members, Earthquake Engineering Research Center, University of
California, Berkeley, CA, EERC 72-9.
NIGAM, N. C. 1967. Inelastic interactions in the dynamic response of structures, Ph.D. thesis,
California Institute of Technology, Pasadena, CA.
OTANl, S. 1974. SAKE-A computer program for inelastic response of R/C frames to earthquakes,
University of Illinois, Urbana, IL, Structural Research Series, No. 413.
OTANI, S., 1976. Earthquake tests of shear wall-frame structures to failure, Proceedings,
ASCE/EMD (Engineering Mechanics Division) Specialty Conference, Dynamic Response of
Structures, University of California, Los Angeles, CA, Mar. 1976, pp. 298-307.
OTANI, S., CHEUNG, V. W. T., and LAI, S. S. 1979. Behaviour and analytical models of
reinforced concrete columns under biaxial earthquake loads, Proceedings, 3rd Canadian
Conference on Earthquake Engineering, Montreal, P.Q., ,pp. 1141-1168.
OTANI, S., and SOZEN, M. A. 1972. Behaviour of multi-story reinforced concrete frames during
earthquakes. University of Illinois, Urbana, IL, Structural Research Series, No. 392.
PECKNOLD, D. A. W. 1974. Inelastic structural response to 2D ground motion, ASCE Journal
of the Engineering Mechanics Division, 100(EM5), pp. 949-963.
POWELL, G. H. l975. Supplement to computer program DRAIN-2D, Supplement to report,
DRAIN-2D user's guide, University of California, Berkeley, CA.
SUKO, M., and ADAMS, P. F. 197l. Dynamic analysis of mu1tibay multi-story frames, ASCE
Journal of the Structural Division, 97(ST10), pp. 2519-2533.
TAKAYANAGI, T., and SCHNOBRICH, W. C. 1976. Computed behaviour of reinforced concrete
coupled shear walls, University of Illinois, Urbana, IL, Structural Research Series, No. 434.
TAKEDA, T., SOZEN, M. A., and NEILSEN, N. N. 1970. Reinforced concrete response to
simulated earthquakes, ASCE Journal of the Structural Division, 96(ST12), pp. 2557-2573.
TAKIZAWA, H. 1973. Strong motion response analysis of reinforced concrete buildings (in
Japanese), Concrete Journal, Japan National Council on Concrete, II (2), pp. l0-21.
TAKIZAWA, H., and AOYAMA, H. 1976. Biaxial effects in modelling earthquake response of R/C
structures, Earthquake Engineering & Structural Dynamics, 4, pp. 523-552.
VELETSOS, A. S., and NEWMARK, N. M. l960. Effect of inelastic behaviour on the response of
simple systems to earthquake motions, Proceedings, 2nd World Conference on Earthquake
Engineering, Tokyo and Kyoto, Vol. II, pp. 895-912.
WEN, R. K., and JANSSEN, J. G. 1965. Dynamic analysis of elasto-inelastic frames,
Proceedings, 3rd World Conference on Earthquake Engineering, Wellington, New Zealand,
Jan. 1965, Vol. II. pp. 713-729.
ZIEGLER, H. 1959. A modification of Prager's hardening rule. Quarterly of Applied Mechanics,
17(1), pp. 55-65.

You might also like