Earthquakes and Structures, Voi. 4, No.

5 (2013) 451-470
DOI: http://dx.doi.Org/10.12989/eas.2013.4.5.451 451

Nonlinear static and dynamic analyses of reinforced concrete

buildings - comparison of different modelling approaches
GONÇALO CARVALHO, RITA BENTO* and CARLOS BHATT
Department of Civil Engineering, Instituto Superior Técnico, Technical University of Lisbon,
Av. Rovisco Pais 1049-001 Lisboa, Portugal

(Received June 8, 2012, Revised September 25, 2012, Accepted October 29, 2012)

Abstract. It generally accepted that most building structures shall exhibit a nonlinear response when
subjected to medium-high intensity earthquakes. It is currently known, however, that this phenomenon is not
properly modelled in the majority of cases, especially at the design stage, where only simple linear methods
have effectively been used. Recently, as a result of the exponential progress of computational tools,
nonlinear modelling and analysis have gradually been brought to a more promising level. A wide range of
modelling alternatives developed over the years is hence at the designer's disposal for the seismic design and
assessment of engineering structures. The objective of the study presented herein is to test some of these
models in an existing structure, and observe their performance in nonlinear static and dynamic analyses.
This evaluation is done by the use of two of a known range of advanced computer programs: SAP2000 and
SeismoStruct. The different models will focus on the element flexural mechanism with both lumped and
distributed plasticity element models. In order to appraise the reliability and feasibility of each alternative,
the programs capabilities and the amount of labour and time required for modelling and performing the
analyses are also discussed. The results obtained show the difficulties that may be met, not only in
performing nonlinear analyses, but also on their dependency on both the chosen nonlinear structural models
and the adopted computer programs. It is then suggested that these procedures should only be used by
experienced designers, provided that they are aware of these difficulties and with a critical stance towards
the result ofthe analyses.

Keywords: material nonlinearity; reinforced concrete sections; lumped plasticity models; distributed
plasticity models; nonlinear seismic analysis

1. Introduction

The response of a building to an accelerated ground motion is carried by a number of physical

mechanisms developed by a complex association of structural elements. These elements are
assembled together to grant civil engineering structures with three fundamental components:
stiffness, resistance and ductility. When provided with a balanced proportion between these
characteristics, each structure shall be able to: (1) control displacements, (2) avoid damage under
low seismic intensities and withstand the remaining design actions and (3) control damage spread,
accommodate large displacements without early collapse and dissipate energy.

•Corresponding author. Professor, E-mail: rbento@civil.ist.utl.pt

Copyright © 2013 Techno-Press, Ltd.

http://www.techno-press.org/?journal=eas&subpage=7 ISSN: 2092-7614 (Print), 2092-7622 (Online)
452 GONÇALO CAR VALHO, RITA BENTO and CARLOS BHA TT

Structural designers, however, have mostly relied the seismic design and assessment of
buildings on linear methods, which have been proved not to describe in a correct fashion the actual
behaviour of asymmetric structures (Priestly 2003). In fact, the structural response is characterized
by a complex nonlinear behaviour where the internal forces are continuously redistributed as the
structure goes through the inelastic stage. This complex behaviour has been studied for the last few
decades in order to grant engineers with more trusted means to predict the seismic response of
building structures. A wide number of nonlinear modelling altematives, analyses and computer
programs have been developed and reported in several research studies. The better understanding
of material behaviour, the growing performance of the existing element models and the increasing
development of computational methods may turn nonlinear analysis into a generalized tool. The
current challenge lies on the development of user-friendly modelling software and
simplified/adequate nonlinear methods compatible with the time constraints found in design
offices, and adapted to the possible lack of knowledge of engineers on nonlinearity issues.
The main goal of this work is to initially summarise some of the existing nonlinear models and
to use some of those available in two different computer programs for a specific case study. In this
paper, a brief description of nonlinear models and analyses is made, with main focus on those that
are featured in SAP2000 (2008), essentially a linear analysis program that has lately started to
develop nonlinear analysis; and SeismoStruct (2010), a more recent program specialized in the
nonlinear field. Six different three-dimensional models were built with these software applications,
and a careful accuracy evaluation was performed using nonlinear static and dynamic analyses.
As an important fact, this work will focus on the flexural behaviour of structural elements,
meaning that no shear failures are expected to affect the nonlinear response of the case study,
which is known to be a rather optimistic hypothesis, especially on older RC buildings. In fact,
shear models are currently under development and their applicability to complex systems is still
far from being straight forward. Moreover, recent capacity design standards prescribe that shear
resistance be assured so as not to harm flexural ductility, i.e., to remain linear.
The amount of work required to build each model in the aforementioned software and the time
consumed by each analysis and procedure were taken into account to evaluate the efficiency of
eaeh alternative. Moreover it intends to wam potential users for the difficulties in performing
nonlinear analyses and for the strong dependence of the results on the nonlinear structural model
adopted.
In the end of the endeavour, final conclusions were outlined in order to support future users on
the choice among the selected models for the seismic assessment of reinforced concrete buildings.

2. Nonlinear models

Structural elements designed to absorb deformation without brittle failures are expected to form
plastic hinges at their most critical sections. These sections are subjected to multiple excursions
into the inelastic range, often followed by their gradual mechanical degradation caused by the
cyeles of loading and unloading.
Numerical models for this type of frame structures have basically fallen into two categories: (1)
the distributed plasticity models, where the inelastic behaviour of the whole element is modelled to
automatically compute the spread of plasticity along its length and (2) the concentrated plastieity
models, where the inelastic behaviour is lumped at the critical sections, corrected by a flxed
parameter that assumes an ideal plasticity distribution.
 Nonlinear static and dynamic analyses of reinforced concrete buildings 453

To model the flexural behavior of the cross sections it is often used either a definition of
hysteretic rules or a fibre discretization model, where a uniaxial model of each fibre material is
directly computed.

2.7 Materials

The monotonie behaviour of concrete and steel has been observed since their first employment.
The developed analytical models have mainly focused on reproducing an adequate stiffness and
strength at any strain level as well as the effects of different confinement disposals. More recently,
particular significance has also been given upon the development of reliable models of concrete
under cyclic loadings. As an example, the models of Scott et al. (1982) and Mander et al. (1988)
have been used for modelling the cyclic behaviour of concrete specimens. The fomier was
complemented with the loading and unloading rules proposed by Thompson and Park (1980), and
the latter using a unique expression for the monotonie envelope and specific cyclic rules, later
modified by Martinez-Rueda and Elnashai (1997).
The nonlinear behaviour of reinforced concrete members is highly controlled by the
reinforcement. Therefore, steel models for longitudinal bars are extremely important to accurately
compute the flexural behaviour of a reinforced concrete section, and especially when it is
subjected to load reversals. The Ramberg-Osgood relations (1943) used by Kent and Park (1973)
and the Menegotto and Pinto (1973) model are two examples of modelling the steel cyclic
behaviour, where the characteristic softening of the curves in the reloading branches is
automatically considered. The Menegotto and Pinto (1973) model has also been included in
several studies for its simplicity and efficiency. It has also been implemented in SeismoStruct
program, with the isotropic hardening rules proposed by Filippou et al (1983).

2.2 Sections

The use of hysteretic relationships represents a relatively simple way to model the flexural
behaviour of an element cross section, i.e., a numerical relationship between moment and
curvature. This kind of formulation is generally based on the definition of a monotonie envelope
and on a set of rules for the definition of hysteretic loops (Stojadinovie and Thewalt 1996), which
are calibrated to assess element stiffness, ductility and energy dissipation. Usually when a
moment-eurvature envelope is assigned to model the biaxial bending of a structural element, the
two directions are considered separately and no axial force interaction is taken into account during
analysis. The idealization of these models is hence performed to an average axial force, which is
generally obtained by means of a linear analysis to gravity loads. Another simple model of the
flexural behaviour of a cross section is the fibre model (Taucer et al. 1991). This model can take
into account biaxial bending and axial force interaction. It consists of a discretization of the cross
section into a flnite number of axial springs acting in parallel, by considering the Euler-Bemoulli
beam theory. The section stiffness is computed based on the tangent stiffness of each flbre material,
on its area of influence and on its coordinates within the cross section.
There have been a number of recent research studies that also include the effect of shear in the
flexural model (e.g. Sezen and Chowdhury 2009, Petrangeli et al. 1999). This important factor has
been indicated as the next step to the seismic assessment of existing RC structures, which are
particularly sensitive to shear mechanisms.
454 GONÇALO CAR VALHO, RITA BENTO and CARLOS BHA TT

2.3 Elements

The maximum intemal forces produced in a building structure, subjected to seismic action,
occur at the elements end sections and their distributions may be assumed as linearly varying.
Therefore, the concentrated plasticity models have emerged by simply lumping flexural
nonlinearity at these end sections.
The flrst formulation of concentrated plasticity models was proposed by Clough et al. (1965)
and consisted of an association in parallel of an elastic and an elastoplastic element. The most
common formulation, however, was initially proposed by Giberson (1967) with an association in
series defined by a nonlinear rotational spring at each end of a linear elastic element. The
rotational springs are defined by a moment-rotation relationship M-9 that integrates the inelastic
curvature distribution expected along the nearest sections, which form the plastic hinge, while the
element itself behaves elastically with limited forces. A very common approach is to admit a
uniform distribution of the inelastic curvatures along a plastic hinge length Lp, which is reassigned
by empirical expressions, by defining the plastic rotation a.?, 6 = x Lp^ where % is the plastic
curvature of the cross section.
In a distributed plasticity model, a finite number of cross sections are usually computed
throughout the element to more accurately consider the inelasticity spread along its length. An old
example of this formulation was suggested by Takayanagi and Schnobrich (1979), by placing
several hinges along the element with specified lengths. This approach is also suggested by the
SAP2000 reference manual (SAP2000 1995). However, the most generalized formulation is based
on numerical integration of section quantities at specified sections, e.g. points of
Gauss/Gauss-Lobato quadrature, to determine the element matrices of stiffness or flexibility. This
formulation is thus divided into displacement-based elements, which compute stiffness by
integrating the moment diagrams with linear curvature interpolation, and force-based elements,
which compute flexibility by integrating the curvature diagrams with linear moment interpolation.
To obtain accurate mathematical descripfion, the former need to be discretised in multiple
subelements, since curvature diagrams are nonlinear, but need no more than two Gauss points,
since moment diagrams are nearly linear. On the contrary, the latter will need no subelements, but
multiple Gauss points. Notes on these formulations, as well as related issues, can be found in
Hellesland and Scordelis (1981), Neuenhofer and Filippou (1997), Scott and Fenves (2006) and
Calabrese e? a/. (2010).

3. Case study

The case study chosen for the current endeavour is an existing five-story reinforced concrete
building (see Fig. 1) located in Turkey. The building was selected from a set of previous studies on
nonlinear static and dynamic analysis for the seismic assessment of torsional sensitive structures
(e.g. Vuran 2007, Ba et al. 2008 and Bhatt and Bento 201 la, b).
The proposed building is asymmetric along the y-axis, and all floors have the same geometry
and height (2.85 m). The columns sections keep the same geometrical and reinforcement features
along the height of the building. The slabs are 0.10 and 0.12 m thick. Beam sections are mainly
0.20 X 0.50 m^ except for the 0.20 x 0.60 m^ located at the centre of the building. Column sections
range from 0.25 x 0.50 m^ to 0.25 x 0.75 m^ and walls from 0.20 x 1.00 m^ to 0.2 x 1.4 m l
Confining stirrups are spaced at 20 cm in both beams and columns. For more structural details, see
 Nonlinear static and dynamic analyses of reinforced concrete buildings 455

(Vuran 2007, Bal et al. 2008). !?• ' - -:

- Z.eo Jf LED -

-iÄ-:f am-
MHow

iMHoar

lMFIeor

fOM

(a) Plan view . (b) Lateral view

Fig. 1 Existing five-story building structure (dimensions in m)

[MPa]

. confined
badly-confined
Z' ^ unconfined

r^T ^ " y

Fig. 2 Concrete capacity curves used. Red points represent conventional failure extensions

Table 1 Concrete parameters

fco [kPal /ccFkPal Ecu
1 0.002 16700 1.0 16700 0.0035
2 0.002 16700 1.1 18370 0.0050
3 0.002 16700 1.2 20040 0.0100

Table 2 Steel parameters'

/.[MPa 371.0 D/fW
J\\\J I 20.0
iSiíMPa 200.0 b 0.005
0.075 al 18.5
a2 0.15
a3 0.025
a4 2.0
The Steel model parameters presented are described in Menegotto and Pinto (1973)
456 GONÇALO CARVALHO, RITA BENTO and CARLOS BHATT

Es

-400

-0.020 -0.015 -0.010 -0.005 0.0 0,005 0.010 0.015 0.020

Fig. 3 Steel capacity curves

3.7 Material models

For materials definition, a mean concrete compressive strength of 16.7 MPa and a steel yield
strength of 371 MPa were considered. The concrete was modelled using the proposal of Mander ei
al (1988), with three straightforward confining ratios kc= 1.0 (unconfined), 1.1 (badly confined)
and 1.2 (confined), which assume a medium-low confinement exploration (see Fig. 2), since the
accurate description of the real structure is not intended in this work. The factor kds defined as the
ratio between the confined and the unconfined concrete compressive strengths, fc and fco
respectively, where^c= kçfo.
The strain £^0, corresponding to fco, was assumed 0.002 and three different ultimate strains £„.
were considered as shown in Table 1.
For the concrete initial modulus of elasticity E^ a value of 19.2 GPa was assigned also
according to Mander et al (1988) expressions.
The reinforcement steel was modelled with the Menegotto and Pinto (1973) relationship, with
an elastic modulus E, of 200 GPa and an ultimate strain e^« of 0.075. Remaining parameters for the
definition ofthe longitudinal reinforcement bars (Menegotto and Pinto 1973) are listed in Table 2.
A cyclic response to a given strain history is presented in Fig. 3.

3.2 Section models

For the nonlinear definition of the cross sections, the two different alternatives mentioned were
considered. The first one is a fibre discretization model applied to each element that automatically
computes section stiffness at every stage, directly from material properties. The second is a direct
definition of fixed moment-curvature relationships M-x, for each axis of each cross section and for
its average axial force (see Section 2.2). This moment-curvature relationships were obtained by a
developed Matlab routine that uses a fibre model and determines a monotonie envelop based on
initial stiffness, moment resistance and deformation energy (see Fig. 4). When used in the
programs SAP2000 and SeismoSirud, different hysteretic rules may be assigned. In this
experiment, the kinematic hysteresis was applied.
 Nonlinear static and dynamic analyses of reinforced concrete buildings 457

M [kNm]
Ultimate moment
_>_ , Frarture

-> X
Fig. 4 SAP2000 hinges for model (a.) (the idealized curve is shown in red)

3.3 Element models

For this case study, six three-dimensional models were considered (Fig. 5), eaeh composed by a
different finite element formulation, according to each program availabilities:
(a) SAP2000 elastic element coupled with two frame hinge elements modelled with the extemally
computed moment-curvature relationships (Fig. 5(a));
(b) SAP2000 elastic element coupled with two frame hinge elements, with the use of the fibre
models of the cross sections (Fig. 5(b));
(e) SeismoStruct elastic element coupled with two nonlinear link elements modelled with the
extemally computed moment-eurvature relationships (Fig. 5(e));
(d) SeismoStruct plastic hinge elements (Fig. 5(d));
(e) SeismoStruct elastic element eoupled with two distributed plasticity elements (Fig. 5(e));
(f) SeismoStruct distributed plastieity elements (Fig. 5(f)).
The element models a., b., c, d. and e., are models of concentrated plasticity and therefore a
plastic hinge length dependent on a fixed ratio A, of the cross sections height Hs is used, i.e.,
Lp = ÄH^. The parameter X was considered as 0.25, 0.5, 0.75, 1.0 and 1.25 for parametric study.
The model a. of SAP2000 is described for beams by an association in series of an elastic
element with one nonlinear rotational spring at each end. These springs are characterized by the
flxed moment-curvature relation of the cross section, detennined by the aforementioned Matlab
routine with zero axial force. For the columns, two moment-curvature relations, one for each axis,
were assigned to each end of the element for the axial force determined at each level. Curvatures
are multiplied by the plastic hinge length to compute element stiffness. Kinematic hysteretie rules
are applied.
The model of auto-computed fibre hinge elements of SAP2000 (model b.) is similar to model a.,
except that each hinge is described by fibre models, being stiffness evaluated directly from
material nonlinearity. In these models a single hinge at each end is enough to model biaxial
bending. Materials' curves in SAP2000 were deflned manually by a number of points, with
kinematic hysteresis. The same association of the hinge length is applied.
The model c. of SeismoStruct is equivalent to model a., being the link elements exactly deflned
as the plastie hinges of SAP2000.
The plastie hinge model of SeismoStruct (model d.) is a model proposed by Scott and Fenves
(2006) that uses a Gauss-Lobato integration method, dividing the element into three regions: a
458 GONÇALO CARVALHO, RITA BENTO and CARLOS BHATT

!*«%• u^

(a) (b) (c) (d) (e) (f)

Fig. 5 Nonlinear finite element models used

central region that remains elastic and two end regions where nonlinearity occurs. The integration
is force-based and it is computed directly using the fibre model of the cross sections.
Models e. and f. of SeismoStruct are based on distributed plasticity elements with a
displacement-based integration method. The formulation of these models assumes two integration
Gauss points (element sections where equilibrium is established). Since nonlinear curvatures are
not reasonably approached by a single linear interpolation, it needs discretization of the element
itself. Thus, the elements of model f were divided into three up to six distributed plasticity
elements depending on their total length. Fibrerized cross-sections - representing sectional details
such as cover and core concrete and longitudinal reinforcements - were then defined at the
respective integration points, whereby every fibre was assigned to an appropriate material
constitutive relationship, as described in Section 3.1. The sectional moment-curvature state of
beam-column elements is obtained by integrating the nonlinear uniaxial stress-strain response of
the individual fibres into which the section has been subdivided. Each section was defined with
300 fibres. In model e., the authors converted the central elements of the beams/columns of model
f. into a linear element so as to limit nonlinearity progression through the element length. Its
purpose was to study the influence of this limitation constituting a concentrated plasticity element
to be compared with its equivalents, and thus with the same procedure regarding the plastic hinge
length. The model e. was deñned because the plastic hinge models c. and d. available in
SeismoStruct were not able to operate sueeessfuUy, having significant convergence problems. This
model was 'created' by the authors as an alternative, since comparison of lumped plasticity models
from different structural programs was one of the objectives.
It is also worth noticing that the rigid diaphragm effect of the slabs was taken into account in all
models at each floor. This assumption, with the use of flbre elements, may lead to an artificial
stiffening/strengthening of the beams, since they become prevented from deforming axially, thus
interacting with the moment-curvature relationship (it is reminded that unrestrained RC elements
subjected to flexure will deform axially, since the neutral axis is displaced from the center of
gravity of the cross section). Being aware that this effect is actually present in real buildings, little
is still known on how it can be modelled with a more realistic stiffness, without using shell
elements. There are, however, experimental tests that have shown good results with using both
rigid diaphragms and distributed plasticity element models (e.g. Bento et al. 2010) and, for its
common use, it has been introduced on our models. Regarding the fixed M-x relationship models,
the diaphragm constraint will not interfere, by definition, with the flexural behaviour of the
elements.
 Nonlinear static and dynamic analyses of reinforced concrete buildings 459

Finally, mass was linearly distributed along the beams and rotations at the base of the vertical
elements were fully restricted.

4. Numerical analyses-description

To firstly study the dynamic characteristics and the linear response of the building structure,
linear modal and response spectrum analyses were carried out. To validate each model, linear static
analyses were also perfonned to check equilibrium of vertical gravity loads.
Pushover analyses were performed in each direction and for all nonlinear models with a lateral
load distribution proportional to the relevant mode. With the distributed plasticity model, a
uniform load was also applied, aiming to assess the structural response differences. To evaluate the
influence of the plastic hinge length on the results, pushover analyses were also carried out to all
variants employed. The N2 method proposed by Fajfar (2000 and 2005) and prescribed by EC8
(CEN 2010) was applied to the different capacity curves obtained by means of pushover analysis,
aiming to evaluate the global structural response to a peak ground acceleration of 0.4 g. Only the
value of A. equal to 0.75 was used to perform this procedure in the concentrated plasticity models
(see Section 3.3).
Finally, nonlinear dynamic time-history analyses were carried out for all models (again only À, =
0.75 was adopted for the concentrated plasticity models) to the three intensities considered in this
study (peak ground acceleration of 0.2 g, 0.3 g and 0.4 g). Regarding the seismic action definition,
three real records from the PEER database (PEER 2010) were considered. The records were fitted
to the EC8 (CEN 2010) elastic response spectrum, using the software RSPMatch2005 (Hancock et
al. 2006) for all the seismic intensities. Each of the three semi-artificial pair of records was applied
twice in the structure changing the direction of the components and thus forming a set of six
analyses: NRl, NR2, TBl, TB2, WNl and WN2.

5. Numerical analyses-results

The most relevant results obtained in this work are summarized in this section. It is important
to emphasise that not all models were able to operate successfully, namely the two concentrated
plasticity models in SeismoStruct (c. and d.). Thus, their results could not be shown in this paper.
Convergence difficulties were also observed during the analyses on the models of SAP2000,
especially on the fibre models that were recently introduced in the program. In both of the
concentrated plasticity models used in SAP2000, convergence failure occurred at a given time step,
which required reductions of the seismic intensity in order to complete analyses.
In general, a structure collapses when the plastic behaviour takes place in a sufficient number
of sections to create a collapse mechanism. This is experienced, with most computer programs, by
observing numerical convergence problems, which can either evidence equilibrium failure or
software issues. In fact, element unloading methods are difficult to describe, and several
convergence complications may lead to unreliable results, sensible to the type of model used,
methods and thus to different software. For instance, SeismoStruct does not consider the loss of
clement strength when ultimate strains occur in materials.
Moreover, since users do not usually have complete access to the internal variables that govern
numerical issues, when using commercial software, it becomes difficult to identify its cause, even
460 GONÇALO CARVALHO, RITA BENTO and CARLOS BHATT

with direct analysis of the output results. Therefore, only a rather superficial analysis is given to
compare these different alternatives and to show which results can be expected from each one
(although not encouraging generalization).
Finally, it is worth to mention that SAP2000 program is able to show the evolution of the
plastic hinges, when described with the frame hinge element property (model a.), during both
non-linear static (pushover) and dynamic analyses. SeismoStruct does not have yet this capability
available nor has SAP2000 with the fiber hinge models. A detailed analysis and comparison of
damage spread at this level would bring substantial value to this study, adding to the global
variables presented herein.

5.1 Dynamic properties

The first three modes of vibration were obtained by a linear model in SeismoStruct. The first
mode (0.615 sec) is basically characterized by a global translation motion in the x direction, with a
slight rotation about the z-axis due to the asymmetry of the structure along the x-axis. The second
mode (0.592 sec), on the other hand, is described as a pure translational motion, derived from the
symmetry along the y-axis, composed by four shear walls oriented in this direction. In the third
mode (0.508 sec), almost pure torsional motion of the structure is observed. Thus the structure is
classified as torsionally stiff.

5.2 Nonlinear static analysis

The capacity curves (top displacement dtopvs total base shear Vb) obtained with the distributed
plasticity model in SeismoStruct (f ) are shown in Fig. 6, for modal and uniform lateral load
distributions, and for x and y directions. The structure presents a slightly greater resistance along
the y-axis, characterized by a hardening phase, whereas along the x-axis an evidenced softening
behaviour is demonstrated. Regarding the two different load distributions, it is shown that the
uniform load is distinguished by higher resisting forces and stiffness. This fact is due to the
presence of stronger forces in superior levels in the modal distribution, which increases the values
of the total shear force of each floor for the same base shear, leading to higher deformations,
affecting resistance itself. Since SeismoStruct does not consider the loss of element strength when
ultimate strains occur in materials, a (red) point was placed on the curves when ten SeismoStruct
strain warnings are registered in different elements, after which the curve should not be taken as
accurately representing the actual capacity. These strain warnings are given as a single fibre
reaches its conventional strain limits (see Section 3.1, Tables 1 and 2) and they are particularly
relevant since the program does not account for these limits.

(a) X-direction (b) Y-direction

Fig. 6 Capacity curves obtained with SeismoStruct distributed plasticity model (f.)
 Noniinear siaiic and dynamic anaiyses of reinforced concrete buiidings 461

i Vb(ieNl

zm~

• ' / " • • '

•~r~i

f
/ _ . e » dbMIMKliitat.

ssw-J
1 s

[m] * dttt» [m)

(a) X-direction (b) Y-direction

Fig. 7 Capacity curves obtain with SeismoStruci limited distributed plasticity model (e.)

t Vt. [kN| A Vi>[kN]

(a) X-direction (b) Y-direction

Fig. 8 Capacity curves obtained with SAP2000 concentrated plasticity model with fibre models (b.)

In terms of the capacity curves obtained with limited distributed plasticity model in
SeismoStruct (e.) for different values of the factor X, the results were considerably close, even for
very small values of X, see Fig. 7. However as the plastic hinge length was reduced, more
convergence difficulties were observed and more computational effort was required, therefore
increasing significantly the duration of the analyses. In fact, when inelastic progress is limited by
the length of the distributed plasticity elements, greater values of curvature are concentrated in the
element critical sections, generating higher section internal forces. An increase of structural
stiffness and a slight decrease of strength are then observed.
Fig. 8 shows the capacity curves obtained with the concentrated plasticity model in SAP2000
using the fibre models ofthe cross sections (b.), for different values of X.
Firstly, a considerably reduced maximum top displacement is verified, which is consistent with
the red dots presented in Fig. 6, as ultimate strain values were directly described in the program
materials definition. This formulation leads to great convergence difficulties as element forces
abruptly drop to zero and stresses are constantly redistributed throughout the structure. The same
result in stiffness is observed when A, is modified.
Different results were obtained for the capacity curves (see Fig. 9) when a hysteretic rule was
imposed in SAP2000 to model the behaviour of each section (a.). In both directions, a lower value
of the maximum shear force is reached, and a considerably high value of ductility factor is
demonstrated. The higher values of strength obtained with model f. could be a result of different
effects, for instance: (1) different failure criteria adopted in SAP2000 and SeismoStruci or (2) the
effect of using distributed plasticity elements with rigid diaphragms - model f As observed in the
462 GONÇALO CARVALHO, RITA BENTO and CARLOS BHATT

i i Vh [kN]

dm [m]
s.» a* «Ä iu« êM 6.m Am
(a) X-direction (b) Y-direction
Fig. 9 Capacity curves obtained with SAP2000 concentrated plasticity model with hysteretie models (a.)

other models, the greater the plastie hinge length is deflned, the higher deformation is achieved
and the higher resistance is expected. In faet, when the plastic hinge length is increased, more
rotation capacity is given to the hysteretie models and thus more deformation capacity is acquired
by the structure. Occasionally, e.g. the y-axis in this case, as the structural displacements grow
higher, resisting forces on the linear elastic elements increase, thus experiencing a greater value of
Vb. Note that until yielding is reached, structural response is kept the same due to the rigid
behaviour of plastic hinges.
The duration of the different pushover analyses are listed in Table 3. It is primarily seen that
concentrated plasticity models with hysteretie rules in SAP2000 (a.) were considerably faster. It is
important to notice however that the durations displayed are highly dependent on the convergence
difflculties of the models and thus the duration of model b. shall be considered cautiously as it
goes less than 1/3 of the top displacement of the other models.
The target displacements determined by means of the N2 method are presented in Table 4. It is
shown that with the concentrated plasticity model with fibre hinges in SAP2000 (b.) it was not
possible to obtain the target displacement values, as the model seems not to have necessary
ductility to reach the deformation imposed by the seismic action applied, which is evidenced by
the clear drop of the curves before encountering convergence difflculties. For the remaining
models, the target displacements were very similar, for both directions.
In Fig. 10, the interstory drifts obtained with this procedure show that models a., e. and f
conducted to considerably consistent results. Compared to the results obtained with the modal
response spectrum analysis, it is concluded that, in this case study, nonlinear models led to a
greater deformation, concentrated in the flrst three stories in the x-direction and more uniformly
distributed along the height in the y-direction.

Table 3 Duration of the pushover analyses

Modal load Uniform load
x-dir y-dir x-dir y-dir
a. 30 m 32 m — -
b. Ih24m lhO5m — —
e. 2 h 30 m 2 h 44 m — —
f. lh30m Ih21m lhOlm Ihl2m
 Nonlinear static and dynamic analyses of reinforced concrete buildings 463

Table 4 Roof target displacements (N2 method)

Modal load
dAm\ dAra\
a. 0.118 0.123
b. — —
e. 0.122 0.125
f. 0.126 0.126

distributed plast., Seismosiruct (f.)

limitBd distr. plast., 5e/smosiruct(e.)
- • — hister. cone, plast., SAP2000 (a.)
—•— elastic spectrum,

(a) X-direction (b) Y-direction

Fig. 10 Interstory drifts obtained with the N2 method and with the elastic response spectrum analysis

In fact, prevailing wall systems (y-direction) tend to form plastic hinges at the base of the walls
due to the greater stiffness of the walls compared to beams, homogenizing the spread of story
drifting in height. On the other hand, in a frame system (x-direction), the exceeding drift at the
base floor caused by plastic excursion does not affect the upper floors.
Torsional effects are also compared in Fig. 11, where the top displacements of the comer
frames PI and P23 (Fig. 1) are normalized with respect to the roof centre of mass displacement in
each direction. While in the y-direction, due to symmetry, no rotation is observed, in the x-axis
normalized top displacements of the comer frames are verifled, which are considerably smaller
compared to the ones experienced in the elastic response spectrum analysis. These results are
according to what was expected as the torsional effects are reduced in the structure when nonlinear
behaviour excursions occur and the original version of the N2 method, herein used, does not take
into account the torsional behaviour of the building.

distributed plast., Sefsmostrvct (f.)

i limited distr. fisst, Seismostfuct{e.)
hister. cona plast., SAP2000 (a.)

•4.2-
1—r elastic spectrum, SAP2000

(a) X-direction (b) Y-direction

Fig. 11 Normalized roof displacements obtained with the N2 method, and with the response spectrum
analysis
464 GONÇALO CARVALHO, RITA BENTO and CARLOS BHATT

5.3 Nonlinear dynamic analysis

The roof displacements obtained with the distributed plasticity model in SeismoStruct (f.), for
the three values of the peak ground acceleration in the first combination of the Northridge record,
arc shown in Fig. 12. On the top of this graphic, as in the following ones (Figs. 13 and 14), a
response period chart is displayed. This chart simply measures the time between maximums and
minimums and multiplies them by two, i.e., it represents the response periods of the structure
through time.
It is observed that increasing the seismic intensity, maximum roof displacements are
near-proportionally increased, and the response periods slightly increased. The latter is caused by
the greater nonlinear excursions and consequent loss of stiffness.
When compared to the equivalent linear behaviour (see Fig. 13), the difference between periods
of vibration is more significant. It is also shown that maximum displacements are very similar, fact
that was confirmed in all remaining records and intensities.

. distr. plast. {ag=0.4 g) SeismoStruct

. distr plast iag=O 3 g) SeismoStruct
. distr. plast. {ag=0.2 g) SeismoStruct

Fig. 12 Top displacements in the x-direction obtained with the distributed plasticity model (f.) in the first
combination of the Northridge record, for different ground peak accelerations

Fig. 13 Top displacements in the x-direction obtained with the distributed plasticity model (f.) in the first
combination of the Northridge record, and with a linear dynamic time-history analysis
 Nonlinear static and dynamic analyses of reinforced concrete buildings 465

Regarding the limited distributed plasticity model built in SeismoStruct (e.), it is observed in
Fig. 14 that no major difference was detected, as expected by the capacity curves shown in Fig. 7.
A very small increase in periods of vibration and amplitudes is verified between the two curves.
To evaluate nonlinear excursions to which the structure is subjected for the different seismic
intensities considered, the total base shear force was compared with and without considering
inelastic behaviour. Fig. 15 shows the values of base shear in the x-direction obtained with the
distributed plasticity model in SeismoStruct (f.), and the corresponding values obtained with a
linear dynamic time-history analysis for the different intensity levels. For this comparison, only
the first seismic combination of the Tabas record is shown. Each pair of the shear forces is very
close during the first 3sec, evidencing a linear elastic response of the structure. From that instant,
the three curves representing the distributed plasticity models nearly follow the same course while
linear elastic curves experience high peak values keeping the same proportion. Therefore, these
results indicate that even though the seismic intensity is reduced by half, a strong nonlinear
behaviour still affects the structure.

dtop [mm]

mtted distr. plast. (ag=0.4 g) SeismoStnict

distributed piast. {ai=0.4 g) SeismaStnict

Fig. 14 Top displacements in the x-direction obtain with the limited distributed plasticity model (e.) and
with the distributed plasticity model (f,) in the first combination of the Northridge record

Vb[kN]

- linear, MP2000 {ag=0,4 g]

- linear, SAFIOS (ag=0,3 g)
- linear, MP2000 {ag=0,2 g)
- plast. dIstr., SeiitnaStrvct {ag=D,4 g)
plast. dIstr., SeismoStruct (aB=0,3 %]
plast. distr., SeismoStruct (ag=0,2 g)

t[sl

Fig. 15 Base shear force in the x-direction, obtained with the distributed plasticity model (f.) for the first
combination of the Tabas record, and with a linear dynamic time-history analysis
466 GONÇALO CARVALHO, RITA BENTO and CARLOS BHATT

[mm]

linear, SAP2000
fibre conc plast., Í1P2000 (b.|
hyster. conc. plast., SAP2000 (a.)
distributed piast., SeismoStruct (f.)

5.0 75 .!.üD 15.S IS.O VS

(a) Peak ground acceleration of 0.4 g

1.5 T[s]

'[mm] rf

àr,»IP2000
fibre conc. piast., S/1P2000 (b.)
- hyster. conc piast., SAP2000 (a.)
St., ie/smoSirucf (f.)

10.0 1X5

(b) Peak ground acceleration of 0.3 g

dtop [mm]

linear, SAP2000
fibre conc. piast., SAP2000 (b.)
. hyster. conc. plast, SAP2000 (a.|
distributed plast., SeismoStruct (f.)

(c) Peak ground acceleration of 0.2 g

Fig. 16 Top displacements in the x-direction obtained with different models (a., b. and e.), in the first
combination ofthe Tabas record and with a linear dynamic time-history analysis

Such conclusion was very important to the analysis of the remaining concentrated plasticity
models (a. and b.) in SAP2000, with which it was found great convergence difficulties. By
knowing the aforementioned conclusion, the peak ground acceleration could be reduced and it
could still be possible to compare models in the inelastic range. Fig. 16 shows the top
displacement obtained with these two concentrated plasticity models compared with the
distributed plasticity model (f.) and with the linear dynamic time-history analysis.
 Nonlinear static and dynamic analyses of reinforced concrete buildings 467

lable 5 Duration of the nonlinear dynamic analyses

A8[gl NRl NR2 TBl TB2 WNl WN2 Average
0.2 55 m 50 m lhO5m 1 hlOm 09 m 1hl6m 54 m
a. 0.3 36 m 36 m 2 h 02 m 24 m 19m 2 h 28 m 1 h 04 m
0.4 54 m 2 h 20 m Ih29m 26 m 14 m 50 m 1 h 02 m
0.2 2 h 26 m 1 h 40 m 3 h 05 m 2 h 58 m 3 hl6m 1 h 40 m 2h31 m
b. 0.3 1 h 06 m 2 h 03 m 33 m 30 m 1h l 6 m 1 h 20 m lhO8m
0.4 2 h 57 m Ih34m 40 m 33 m 1 h 32 m 1 h 29 m 1 h 28 m
e. 0.4 5h31 m 7 h 38 m 6h 16m 6 h 07 m 7 hlOm 6 h 38 m 6 h 33 m
0.2 2 h 23 m 2 h 25 m 2h21m 2 h 33 m 2 h 55 m 2 h 39 m 2 h 33 m
f. 0.3 2 h 37 m 2 h 33 m 2 h 35 m 2 h 50 m 3hl7m 2 h 25 m 2 h 43 m
0.4 3 h 03 m 2 h 58 m 3hl6m 2 h43m 1 h 13 m 1 h 15 m 2 h 25 m

It is concluded that the results obtained with models a. and f are reasonably precise in the first
steps of the analysis, particularly when Fig. 15 predicts higher nonlinear behaviour up to the flrst 7
sec. Model a., presents however lower periods of vibration. The deformability exhibited with the
fibre concentrated plasticity model in SAP2000 (b.) is considerably higher compared to the other
models, which leads to higher values of displacement. It was not possible to detect a fair reason for
this problem, although the capacity curves (Fig. 8) evidenced lack of ductility and consequently
low energy dissipation capacity.
In Table 5, a list of the duration of each nonlinear dynamie analysis for the four models a., b., e.
and f, with the three peak ground accelerations for the six semi-artiflcial accelerograms is
presented. The last column of Table 5 depicts the average time duration for each model and for
each peak ground acceleration considered in this work.
It is pointed again, that the durations of models a. and b. shall be examined carefully as the
analyses could not be completed in some cases due to convergence difficulties.

6. Conclusions

The work developed in this study is able to evidence the difflculties, which may be found in
using both nonlinear static and dynamic analyses. One can flnd in many cases unreliable results
and its strong dependence on the plastic model used, in the analysis method chosen or in the
structural program adopted.
In this research, during the process of creating each model and performing each analysis, it was
concluded that the versions of both programs used in this endeavour {SAP2000 2008, SeismoStruct
2010) have still several issues in terms of practical applicability in the design activity of real
structures.
When modelling a building structure, each program has different limitations:
• In SeismoStruct, although considering nonlinear behaviour is almost automatic due to the
section fibre models, building the complex model itself becomes heavy;
• In SAP2000, despite the case of defining the structure geometry by using its intuitive
graphical interface, the nonlinear behaviour modelling requires the use of external applications to
compute the large amount of hysteretie relationships (if the option of using the fibre models of the
cross sections is not adopted)
468 GONÇALO CAR VALHO, RITA BENTO and CARLOS BHA TT

As concentrated plasticity models in SeismoStruct (c. and d.) were not able to work in any type
of analysis, their use should be cautious and it seems not advisable for building structures of
equivalent or higher complexity, at the moment. The limited distributed plasticity model (e.) is
also not recommended as its results were very similar to those obtained with the distributed
plasticity models, with the additional convergence difficulties and duration increase in both
nonlinear analyses performed. The concentrated plasticity model in SAP2000, with fibre models
(b.), has also given convergence problems (in this version of SAP2000), as in the capacity curves
deflnition it lead to a relatively low ductility factor and in the nonlinear dynamic analysis it
showed convergence failure at early time steps, with relatively high displacements. Convergence
difficulties could not be avoided reducing time steps, but arised instead from the fact that the
elements itself could not withstand the imposed displacements. A careful analysis of single
elements showed that materials were lead to the strain limits. This is actually cleared by the final
drop in resistance presented in the capacity curves of Fig. 8(a).
Both the concentrated plasticity model of SAP2000 (a.) and the distributed plasticity model of
SeismoStruct (f.) have shown fast applicability (durations of analysis) and little convergence
difficulties. The results obtained with these models were also comparable in the time-history
analysis, although pushover has evidenced differences in the base shear force estimation. Although
the time-consuming is reduced by half using model a., the nonlinear modelling is straightforward
with model e.
What is finally evidenced in this study is that inexperienced users should not rely their studies
on one single model nor on one single analysis. Reliable models will only exist when researchers
can gather a considerable amount of real scale tests to confront with computer models. As an
example of this type, there is the SPEAR building (Bhatt and Bento 201 la, 201 lb), a real scale test
that was confronted with analytical models, including the distributed plasticity models of
SeismoStruct, with adequate comparison.
Regardless of nonlinear analysis being, beyond any doubt, a powerful tool for the seismic
assessment and design of structures, its use in design engineering offlces has not been widespread.
Based on the authors' knowledge and on the set of issues herein presented related to such kind of
analysis, it is recommended that designers avoid nonlinear analyses, unless they are well
experienced in their use and adopt a strong critical stance towards the obtained results. In
particular, the use of nonlinear dynamic analysis, which cannot attract design offices due to its
time-consuming nature (long computation times) and complexity (when compared with nonlinear
static analyses), namely in terms of record selection. In fact, this is a topic for which the scientiflc
community has not yet found widely accepted and deflnitive answers.

Acknowledgments

The authors would like to acknowledge the financial support of the Portuguese Foundation for
Science and Technology (Ministry of Science and Technology of the Republic of Portugal)
through the research project PTDC/ECM/100299/2008 and through the PhD scholarship
SFRH/BD/28447/2006 granted to Carlos Bhatt. The authors acknowledge with thanks the support
by FEDER and the Fundaçâo para a Ciencia e Tecnologia through the funding of the research unit,
ICIST, Instituto de Engenharia de Estruturas Territorio e Constmçâo.
 Nonlinear static and dynamic analyses of reinforced concrete buildings 469

References ,. j •

Bal, I., Crowley, H., Pinho, R. and Gulay, G. (2008), "Detailed assessment of structural characteristics of
Turkish RC building stock for loss assessment models", Soil Dyn. Earthq. Eng., 28(10-11), 914-932.
Bento, R., Bhatt, C. and Pinho, R. (2010), "Using nonlinear static procedures for seismic assessment of the
3D irregular SPEAR building", Earthq. Struct., 1(2), 177-195.
Bhatt, C. and Bento, R. (2011a), "Assessing the seismic response of existing RC buildings using the
extended N2 method", B. Earthq. Eng., 9(4), 1183-1201.
Bhatt, C. and Bento, R. (2011b), "Extension of the CSM-FEMA440 to plan-asymmetric real building
structures", Earthq. Eng. Struct. D., 40(11), 1263-1282.
Calabrese, A., Almeida, J.P. and Pinho, R. (2010), "Numerical issues in distributed inelasticity modeling of
RC frame elements for seismic analysis", J. Struct. Eng., 14(S1), 38-68.
Carvalho, G. (2011), Análise Sismica de Edificios de Betào Armado - Estudo de Alternativas de Modelacäo
e Análise Näo-Linear, Master'sthesis, Instituto Superior Técnico, Universidade Técnica de Lisboa,
Portugal.
CEN (2010), Eurocode 8: Design of structures for earthquake resistance - Part 1: General rules, seismic
actions and rules for buildings, Brussels, Belgium.
Clough, R., Benuska, K. and Wilson, E. (1965), "Inelastic earthquake response of tall buildings".
Proceeding of Third World Conference on Earthquake Engineering, New Zealand 11.
Fajfar, P. (2000), "A nonlinear analysis method for performance based seismic design", Earthq. Spectra,
16(3), 573-592.
Fajfar, P., Marusic, D. and Perus, I. (2005), "The extension of the N2 method to asymmetric buildings",
Proc. of the 4th European Workshop on the Seismic Behaviour of Irregular and Complex Structures,
Thessaloniki, Greece.
Filippou, F.C., Popov, E.P. and Bertero, V.V. (1983), "Modelling of r/c joints under cyclic excitations",
ASCEJ. Struct. Eng., 109(11), 2666-2684.
Giberson, M. (1967), The response of nonlinear multi-story structures subjected to earthquake excitation,
Ph.D. thesis, California Institute of Technology.
Hancock, J., Watson-Lamprey, J., Abrahamson, N.A., Bommer, J.J., Markatis, A., McCoy, E. and Mendis,
R. (2006), "An improved method of matching response spectra of recorded earthquake ground motion
using wavelets", J. Earthq. Eng., 10(1), 67-89.
Hellesland, J. and Scordelis, A. (1981), Analysis ofRC bridge columns under imposed deformations, LABSE
Colloquium, Delft, Netherlands.
Kent, D.C. and Park, R. (1973), "Cyclic load behaviour of reinforcing steel". Strain J. British Soc. Strain
Meoj., 9(3), 98-103.
Mander, J., Priestley, M. and Park, R. (1988), "Theoretical stress-strain model for confined concrete", J.
Struct. Eng., 114(8), 1804-1826.
Martinez-Rueda, J. and Elnashai, A. (1997), "Confined concrete model under cyclic load". Mater. Struct.,
30(3), 139-147.
Menegotto, M. and Pinto, P. (1973), "Method of analysis for cyclically loaded RC plane frames including
changes in geometry and non-elastic behaviour of elements under combined normal force and bending".
Symposium on the Resistance and Ultimate Deformability of Structures Acted on by Well Defined
Repeated Loads, International Association for Bridge and Structural Engineering, Zurich, Switzerland
15-22.
Neuenhofer, A. and Filippou, F.C. (1997), "Evaluation of nonlinear fi-ame finite element models", J. Struct
Eng., 123(7), 958-966.
PEER (Pacific Earthquake Engineering Research Center) (2010), Strong ground motion database,
http://peer.berkeley.edu.
Petrangeli, M., Pinto, P.E. and Ciampi, V. (1999), "Fiber element for cyclic bending and shear of RC
structures",/. Eng. Mech.-ASCE, 125(9), 994-1009.
470 GONÇALO CAR VALHO, RITA BENTO and CARLOS BHA TT

Priestley, M.J.N. (2003), Myths andfallacies in earthquake engineering (Revisited), Pavia : IUSS Press.
Ramberg, W. and Osgood, W.R. (1943), Description of stress-strain curves by three parameters. Technical
Note 902.
SAP2000 (1995), Analysis reference manual. For SAP2000 R, ET ABS R and SAFETM.CSI.
SAP2000 (2008), vl2.0.0 Advanced, Computers and Structures, Inc.
Scott, B., Park, R. and Priestley, M. (1982), "Stress-strain behaviour of concrete confmed by overlapping
hoops at low and high stain rates", J. Am. Concrete Inst.
Scott, M.H. and Fenves, G.L. (2006), "Plastic hinge integration methods for force-based beam-column
elements", J. Struct. Eng.-ASCE, 132(2), 244-252.
SeismoStruct (2010), v5.0.5. A computer program for static and dynamic nonlinear analysis of framed
structures, SeismoSoft, Ltd. Available from: www.seismosoft.com.
Sezen, H. and Chowdhury, T. (2009), Hysteretic model for reinforced concrete columns including the effect
of shear and axial loadfailure, J. Struct. Eng.-ASCE, 135(2), 139-146.
Stojadinovic, B. and Thewalt, CR. (1996), "Energy balanced hysteresis models". Eleventh World
Conference on Earthquake Engineering, Earthquake Engineering Research at Berkeley, College of
Engineering, University of California at Berkeley.
Takayanagi, T. and Schnobrich, W. (1979), "Nonlinear analysis of coupled wall systems", Earthq. Eng.
Struct. D., 7(1), 1-22.
Taucer, F.F., Spacone, E. and Filippou, F.C. (1991), "A fiber beam-column element for seismic response
analysis of reinforced concrete structures". Earthquake Engineering Research Center, College of
Engineering, University of California, Berkeley.
Thompson, K.J. and Park, R. (1980), "Moment-curvature behaviour of cyclically loaded structural concrete
members", ICE Proceedings.
Vuran, E. (2007), Comparison of nonlinear static and dynamic analysis results for 3D dual structures.
Master's thesis, UniversitàdegliStudi di Pavia.

SA
Copyright of Earthquakes & Structures is the property of Techno-Press and its content may
not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for
individual use.