Chapter 12

A Toolbox and Web Application for the Seismic

Performance Assessment of Buildings

Matjaž Dolšek, Robert Klinc, Matevž Dolenc, Marko Brozovič,

and Iztok Peruš

Abstract A software tool, known as a PBEE toolbox, for the seismic performance
assessment of buildings, which was developed in Matlab in conjunction with the
software framework OpenSees, and a web application for the prediction of approx-
imate IDA curves are presented in this chapter. Although, in its present version, the
PBEE toolbox supports only simple nonlinear models, its capabilities exceed usual
software tools for computational simulation, since it enables seismic performance
assessment of buildings with various procedures and adopts an open-source
philosophy so that it can be easily extended or modified to suit specific user
requirements. The capabilities of the PBEE toolbox and the web application,
which involves the response database of a single-degree-of-freedom system with
a quadrilateral force-displacement relationship, are demonstrated by means of an
assessment of the seismic response parameters of an eight-storey reinforced con-
crete frame, using incremental dynamic analysis, progressive incremental dynamic
analysis, approximate incremental dynamic analysis, and the N2 method. It is
shown that, for the case of the presented example, all the methods produce similar
results, although each method has some advantages and some disadvantages.

Keywords Performance-based earthquake engineering • PBEE toolbox • Web

application • Progressive incremental dynamic analysis • Approximate incremental
dynamic analysis • Reinforced concrete frames

M. Dolšek (*) • R. Klinc • M. Dolenc • M. Brozovič • I. Peruš

Faculty of Civil and Geodetic Engineering, University of Ljubljana,
Jamova 2, Ljubljana SI 1000, Slovenia
e-mail: [email protected]; [email protected]; [email protected];
[email protected]; [email protected]; [email protected]

M. Dolšek (ed.), Protection of Built Environment Against Earthquakes, 233

DOI 10.1007/978-94-007-1448-9_12, # Springer Science+Business Media B.V. 2011
234 M. Dolšek et al.

12.1 Introduction

The response of structures to severe earthquakes is nonlinear, since they are usually
designed to withstand large seismic demands without local or global collapse, but
selected structural elements may be significantly damaged. From this point of view,
the seismic response parameters of a structure, which are estimated in the design or
assessment process, should be determined by using nonlinear methods of analysis.
Unfortunately, in the case of practical applications such an approach is rare, for
many reasons. The main reason probably lies in the fact that nonlinear analysis is
far more complex than linear analysis. This means that the performance assessment
of structure based on nonlinear analysis requires much more knowledge, input data,
and work. The complexity of nonlinear analysis is also probably the reason why the
use of nonlinear methods of analysis is not required in structural codes, but only
defined as an alternative to linear methods of analysis. Consequently, structural
engineers do not check the response of structures by using nonlinear methods of
analysis, since it is not necessary to do so according to the structural codes.
However, they are aware of the fact that structural response will be nonlinear
in the case of the design earthquake.
In order to facilitate the use of nonlinear methods of analysis in practice, without
forcing such use into structural codes, it is necessary to extend the knowledge of
engineers and to develop user-friendly and reliable software tools that are able to
support various nonlinear models and methods for the seismic performance assess-
ment of structures.
Some commercially available software already supports performance-based
methodologies. For example, SAP 2000 (CSI 2008), PERFORM-3D (CSI 2006)
and MIDAS (2008) provide nonlinear structural analysis methods, as well as different
types of simplified procedures for the performance assessment of structures.
In addition to commercial software, free (SEISMOSOFT 2007; Elnashai et al. 2008)
and open-source software (McKenna and Fenves 2007) is also available. This kind
of software has certain advantages in comparison with commercially available soft-
ware, since it provides a comprehensive library of nonlinear elements, material
models, analysis types and solvers. From this point of view OpenSees (McKenna
and Fenves 2007) is probably the most comprehensive software framework in the field
of earthquake engineering. However, it does not usually support the performance-
based assessment methods prescribed by various structural codes, since it is mostly
focused on the research community, in order to support researchers to develop
applications for the simulation of the performance of structural systems subjected
to earthquakes.
In order to further extend the applicability of software for computational simula-
tion, a PBEE toolbox for the seismic performance assessment of reinforced concrete
frames and a web application for the prediction of approximate IDA (incremental
dynamic analysis) curves have been developed. They are briefly presented and their
use is demonstrated by means of the seismic performance assessment of an eight-
storey frame, whose performance has been analysed with various methods.
12 A Toolbox and Web Application for the Seismic Performance. . . 235

12.2 Overview of the Procedures for the Seismic

Performance Assessment of Buildings

Many different procedures for the seismic performance assessment of buildings

have been developed in past decades, not only because of different needs but
also because the fact that different researchers and structural engineers have had
different philosophies for the solving of such problems. Consequently, seismic
performance assessment procedures can be classified according to many different
criteria.
The most general classification of seismic performance assessment procedures
can be based on the type of analysis. In linear elastic analysis it is assumed that
the material remains in the elastic range. Since earthquakes often cause damage to
structures, this assumption is not sufficiently accurate from the theoretical point of
view, but it is often applied in the case of seismic analysis. For example, the
European standard Eurocode 8-3 (CEN 2005) for the assessment and retrofitting
of buildings prescribes that seismic response parameters can also be determined by
lateral force or modal response spectrum analysis, which are both types of linear
elastic analysis. However, the advantage of the linear elastic analysis in comparison
to the nonlinear analysis is its numerical stability, relatively short computational
time, the simple definition of earthquake loading, and the broad availability of user-
friendly software. Nonlinear analysis is, of course, computationally much more
demanding and therefore still rarely suitable for practical use.
Linear elastic and nonlinear analysis can be further classified into static and
dynamic analysis. In the case of static analysis the equilibrium equations are
independent of time, whereas in the case of dynamic analysis the seismic response
parameters are computed for each time step. Since nonlinear dynamic analysis
considers the nonlinear behaviour of structures, and takes into account the dynamic
effects, which are important in the case of earthquake actions on a structure, this
analysis method is the most advanced for the simulation of structural response
under seismic action. However, it is the subject of many uncertainties, and its use is
not trivial. Firstly, nonlinear dynamic analysis is computationally extremely
demanding, and, in the case of complex structural models, can also become
the subject of numerical instability. Furthermore, earthquake action has to be
defined by means of ground motion records. Although simple guidelines for the
selection of ground motion records exist, different engineers/researchers
may select different ground motion records for the analysis. Consequently,
estimated seismic response parameters may vary from case to case. Additionally,
predicted seismic response parameters can differ due to different nonlinear models
used in the analysis. This is because many different nonlinear models exist, and there
is no consensus as to which model is the most appropriate for the seismic analysis.
Last but not least, nonlinear dynamic analysis involves the numerical integration of
the equations of motion. It is well-known that results of nonlinear dynamic analysis
more or less depend on the parameters of integration schemes. Even the level of
critical damping and the corresponding model of viscous damping can significantly
236 M. Dolšek et al.

Fig. 12.1 List of the different types of analyses used for the seismic performance assessment of
buildings

affect the results of nonlinear dynamic analysis. Because of all the issues encountered
in nonlinear dynamic analysis, practice-oriented nonlinear analysis procedures, which
typically involve nonlinear static (pushover) analysis and prediction of the target
displacement, have emerged over the last two decades, and become popular among
both researchers and engineers. Many different procedures have been developed
(Fajfar 2000; Priestley and Kowalsky 2000; Chopra and Goel 2002; Aydinoğlu
2003; Antoniou and Pinho 2004), and some of them have become a part of standards
and guidelines, e.g. CEN (2004a), FEMA (2000), and FEMA (2005).
Seismic performance assessment procedures can be also classified as determin-
istic or probabilistic procedures, which can involve different type of analysis
(Fig. 12.1). In deterministic procedure all the input parameters are assumed
according to the best knowledge and/or often conservatively. For example, the
mean/median or other characteristic value (fractile) is adopted for the strength of a
material, although it can vary in different parts of the structure. The opposite is in
the case, if the seismic performance assessment of building is based on probabilistic
procedure where some or all of the input parameters are considered to be uncertain,
and are most frequently treated by appropriate random variables. In this case, the
uncertainties are usually classified into two principal categories (Ellingwood and
Kinali 2009): the aleatoric uncertainties, which are related to the random nature of
earthquakes and are, by definition, irreducible, and the epistemic uncertainties,
which are knowledge-based and are most often related to the physical properties
of the structure and its modelling parameters.
Most seismic performance assessment procedures were developed to be used for
the assessment of a single building structure, but many procedures have been
simplified, and can be used for the seismic performance assessment of building
stock. Such classification is therefore also viable.
Some of the analyses used in the process of the seismic performance or risk
assessment of buildings are illustrated in Fig. 12.1. In this case all the presented
analyses are nonlinear, starting with the dynamic (SDOF Dynamic, see Fig. 12.1)
and incremental dynamic analysis (IDA) (Vamvatsikos and Cornell 2002) of an
equivalent single-degree-of-freedom (SDOF) system (SDOF-IDA, see Fig. 12.1).
Since this type of analysis involves a simple nonlinear model, which can be arbitrarily
defined based on some structural parameters, it can be used for the seismic perfor-
mance assessment of building stock. The IDA curves of the equivalent SDOF system
can also be used for determination of the target displacement, if combined with
12 A Toolbox and Web Application for the Seismic Performance. . . 237

pushover analysis (Pushover, see Fig. 12.1), which is the first analysis from the list of
analyses presented in Fig. 12.1, which require structural model of the entire building.
The next analysis from the list Fig. 12.1 is the nonlinear dynamic analysis
(Dynamic, see Fig. 12.1), which has already been described, and the last is incre-
mental dynamic analysis, which is a well-known parametric analysis method, and
involves nonlinear dynamic analysis and subjecting a structural model to a number
of ground motion records, each scaled to multiple levels of intensity. IDA is
powerful analysis method, but is computationally extremely demanding. An alter-
native to IDA is progressive IDA (Azarbakht and Dolšek 2011) (Progressive IDA,
see Fig. 12.1). A new element of progressive IDA, when compared to the elements of
IDA, is the precedence list of ground motion records. In general, IDA curves are
calculated for all the ground motion records in a set of such records, whereas in
progressive IDA, the IDA curve is first calculated for the first ground motion record
from the precedence list, and then progressively for the other ground motion records
from the precedence list of records. After several IDA curves have been calculated,
the analysis can be terminated, since the acceptable tolerance is achieved. Such an
approach facilitates practical application, with the aim of selecting the most repre-
sentative ground motion records for incremental dynamic analysis.
The methods of analysis presented on the left hand side of Fig. 12.1, as well as
the corresponding simplified versions of these methods, are not computationally
demanding, and are therefore often used for the seismic performance assessment of
building stock, whereas the methods of analysis on the right hand side of Fig. 12.1
require more computational time and they can be afforded only for the seismic
performance assessment of individual building structures. Additionally, all the
listed analyses can be used for deterministic or probabilistic performance assess-
ment of structures. Deterministic assessment is most often adopted in the case if the
seismic performance assessment is expressed in terms of seismic response
parameters, such as, inter-storey drift, storey acceleration, or similar, whereas
probabilistic assessment is required if seismic performance of a structure is judged
based on the mean annual frequency of exceedance of a given limit state, expected
monetary or other kind of losses.

12.3 Seismic Performance Assessment of Buildings

by Using the PBEE Toolbox and Web Application

An effective approach for the seismic performance assessment of buildings is

introduced. It involves the use of the PBEE toolbox (Dolsek 2010) in conjunction
with a web application for the prediction of approximate IDA curves. The PBEE
toolbox is a software tool, which was developed in order to fill the gap between the
user, who needs to evaluate the seismic performance of a structure, and OpenSees
(McKenna and Fenves 2007), which consists of comprehensive software for
computational simulation in the field of earthquake engineering.
238 M. Dolšek et al.

12.3.1 Overview of the PBEE Toolbox

The PBEE toolbox is a set of Matlab (MathWorks 2007) functions, which can be
used for the seismic performance assessment of reinforced concrete (RC) frames in
conjunction with OpenSees. The aim of the PBEE toolbox is to enable rapid
definition of simple nonlinear structural models of RC frames. In this case
the most time-consuming part of the work involves the determination of the
properties of the plastic hinges. Since the PBEE toolbox automatically generates
the properties of plastic hinges, based on data regarding material strength, reinforce-
ment and section properties, the amount of work which has to be performed in order
to prepare a structural model is reduced significantly. The current version of the
PBEE toolbox combines the Eurocode 8 (CEN 2005) requirements for non-linear
modelling and the non-linear seismic analysis of buildings with some other
approaches which were used elsewhere (Fajfar et al. 2006; Peruš et al. 2006).
However, since the PBEE toolbox is based on open-source philosophy, the user
can simply change the function which, for example, is used to determine the
moment-rotation relationship of the plastic hinges, and still use other functions of
the PBEE toolbox.
Different assumptions which, in general, follow the Eurocode 8 requirements,
are used in order to establish as simple as possible but yet adequate structural
model. These assumptions are as follows:
• the floor diaphragms are assumed to be rigid in their own planes, and the masses and
moments of inertia of each floor are lumped at the corresponding centres of gravity,
• the beam and column flexural behaviour is modelled by one-component lumped
plasticity elements, composed of an elastic beam and two inelastic rotational
hinges, which are located at the ends of elastic element, and defined by a
moment-rotation relationship. The element formulation is based on the assumption
of an inflexion point at the midpoint of the element. For beams, a plastic hinge is
used for major axis bending only. For columns, two independent plastic hinges for
bending about the two principal axes are used.
• the moment-rotation relationship before strength deterioration is modelled by a
bi-linear or tri-linear relationship. Zero axial force and the axial load due to
gravity loads are taken into account when determining the moment-rotation
relationship for beams and columns, respectively. A linear negative post-capping
stiffness is assumed after the maximum moment is achieved.
• the gravity load is represented by a uniformly distributed load acting on the
beams, and/or by concentrated loads acting at the top of the columns.
Work with the PBEE toolbox involves the Matlab script language. First the user
has to define the structural data. In the present version of the PBEE toolbox the data
regarding the structure are organized at the structural and element type level. At the
structural level the user defines the structural gridlines and the storey masses,
12 A Toolbox and Web Application for the Seismic Performance. . . 239

whereas at the element level the user separately defines the beams and the columns
of the structure. Once the structural data are defined, the user calls functions in order
to compute the moment-rotation envelopes in the plastic hinges, and after that the
input files for OpenSees can be simply generated by calling functions for the
generation of the tcl code. Depending on the type of analysis, the user has to
prepare some data regarding the analysis, and runs the analysis in OpenSees
through the Matlab environment.
In general there is no need to work with OpenSees since all the results are stored
in Matlab data structures after the analysis has been performed in OpenSees. The
results are organized based on the so-called global results, which are, for example,
the displacements at the mass centre, and the local results, which include the forces
and deformations at the plastic hinges. For some type of analyses, e.g. for incre-
mental dynamic analysis (Vamvatsikos and Cornell 2002), the PBEE toolbox
provides an iterative process between the results of the non-linear dynamic analysis
obtained by OpenSees, the requirements of IDA as defined by the analysis data, and
the automatic tcl code generation for OpenSees.
After the analysis has been performed the user can define different limit states at
the level of the plastic hinge in order to link the damage in the plastic hinge to the
global seismic response parameters. For example, if pushover analysis is performed,
the user can simply link the damage at hinge level to the top displacement or base
shear, or, based on the results of incremental dynamic analysis, link the damage in
the plastic hinges with the intensity measure. Such an approach enables graphical
representation of the damage in the plastic hinges on the pushover or IDA curve.
Additionally, the user can determine the target displacement and the damage to the
structure according to the N2 method (Fajfar 2000), and present the results visually
in AD format. The damage in the plastic hinges can also be visually presented by
plotting it on a structural drawing.
The PBEE toolbox also includes functions which can be used to determine the
IDA curves of equivalent SDOF systems. Such an approach represents an alterna-
tive to seismic performance assessment with the N2 method, which involves
closed-form expressions for determination of the target displacement that can be
developed based on simple parametric studies. However, more sophisticated
parametric studies can lead to more accurate predictions of the target displacement,
and can also provide new information, such as the dispersion in the displacement
demand, or the prediction of collapse capacity, which are needed for the probabi-
listic risk assessment of structures. The advances in information technology (IT), as
well as new software and growing computer power, provide new possibilities for
solving the above-described shortcomings, which are embedded in the simplified
nonlinear methods for seismic performance assessment. Rather than developing
new simplified expressions for inelastic displacement ratios, a web-based applica-
tion for the prediction of approximate fractile IDA curves has been recently
developed, and can be used in conjunction with the PBEE toolbox.
240 M. Dolšek et al.

12.3.2 Brief Description of the Methodology for Prediction

of Approximate IDA Curves

The methodology which was used to develop the web-based application for
determination of approximate IDA curves consists of two independent processes.
The first process involves determination of the response database, whereas the second
process involves the prediction of approximate IDA curves from the response data-
base. The first process is the parametric study, which is performed for the SDOF
model, and involves definition of the input parameters which affect the seismic
response, definition of the discrete values of the input parameters, and computation
of the IDA curves of the defined SDOF models. The usual input parameters of the
parametric study are the period of the system, the parameters of the force-displacement
relationship and the hysteretic behaviour, the damping, and the ground motions. It is
worth emphasizing that definition of the SDOF model depends on expert judgment,
since the selection of a force-displacement relationship or hysteretic behaviour of the
SDOF model, which might be appropriate for simulating the global response of a
specific structural type, is not trivial. Different experts may select different input
parameters or discrete sets of their values. It is therefore important that the process
of the parametric study is independent of the second process, since in the future many
different databases may be created. The results of the parametric study can be used to
create a response database, which can be established for a discrete number of input
parameters of the SDOF model, for a set of ground motion records.
Since the IDA curves, which are stored in the response database, are computed for
the discrete parameters of the SDOF model, the second process involves prediction
of the approximate IDA curves for any input parameter of the SDOF model. Clearly,
this process is trivial if the request for the prediction of approximate IDA curves is
based on the same input parameters as those for which the IDA curves are available
in the response database. However, this is, in general, a rare case. The second process
therefore involves two steps: the query for appropriate IDA curves from the data-
base, and the computation of the approximate IDA curves by the selected interpola-
tion method. If the response database is computed for a sufficient number of discrete
input parameters, then linear interpolation is a suitable method for determining the
approximate IDA curves. Since the IDA curves depends on n input parameters, n-
dimensional linear (also known as multi-linear) interpolation is applied by using
one-dimensional linear interpolation in each separate coordinate dimension.
Such an approach, which is described elsewhere (Peruš et al. 2011), requires the
neighbouring data which are mapped into the unit hypercube [0 1]n. Consequently,
the result of the query from the response database is a set of 2n IDA curves.
It is known that equivalent SDOF models vary depending on the structural system
and the material of the structure. For example, in the case of the prediction of the
approximate IDA curves of a reinforced-concrete building, the force-displacement
relationship can be described by the four dimensionless parameters rv ¼ F1/F2,
rh ¼ u1/u2, mu ¼ u3/u2, a¼ kpc/ki, where points (u1, F1) and (u2, F2) represent the
first and second characteristic points of the idealized force-displacement
12 A Toolbox and Web Application for the Seismic Performance. . . 241

relationship and roughly represent, respectively, the cracking of the concrete and, in
the case of regular structures, yielding of reinforcement at the base of the columns.
The displacement u3 is related to the displacement where the strength of the structure
starts degrading, while kpc and ki are, respectively, the post-capping and initial
stiffness of the idealized force-displacement relationship. With a suitable variation
of the four parameters the idealized force-displacement relationship can be fitted
to almost any pushover curve. The parameters of the force-displacement relationship
are visually explained when the user starts the web application (http://ice4risk.slo-
projekt.info/analysis/ ).
The other two parameters, which describe the SDOF model, are the period T1 and
the critical damping ratio x. Note that the Takeda’s hysteretic rules (Takeda et al.
1970) can be used for the most basic simulation of the hysteretic behaviour of
reinforced concrete buildings. These hysteretic rules were used for determination of
the response database of the web application, presented in the next subsection.
The only practical way is that the response database is filled with automated
computational procedures, for which OpenSees (McKenna and Fenves 2007) can
be used. However, for these computations the SDOF model has to be defined with
dimensional quantities. For this purpose the mass and the force F2 of the SDOF
model can be assumed to be constant, e.g. m ¼ 100 t and F2 ¼ 0:1
m
g, where g
is the acceleration of gravity. Based on these definitions and the known dimension-
less parameters rv, rh, mu and a, the period T1 and the critical damping ratio, it is
trivial to define the SDOF model. This is because the displacement u2 can be
determined from the relationship between the mass, stiffness and the period of
the SDOF model, taking into account the defined parameters rv and rh:

T12 F2
rv T12 g rv
u2 ¼ ¼ : (12.1)
4p2 m
rh 40p2 rh
Other dimensional parameters of the SDOF model can be then obtained from the
definition of the non-dimensional parameters, while the damping constant c can be
calculated from the expression:
4px
c¼ : (12.2)
T1
It is suitable that the IDA curves in the response database are stored for the
relationship between the displacement and the peak ground acceleration, and for the
displacement and spectral acceleration at period T1, which corresponds to the initial
stiffness of the SDOF system. The IDA curves can be also expressed with the
reduction factor R and the ductility m, which are used for the definition of inelastic
spectra (the R-m-T relationship). However, the reduction factor should in this case
be understood as the normalized spectral acceleration. Note that both dimensionless
parameters (R, m) can be defined based on different assumptions. One possibility is
to define the ductility according to the displacement u2, and the yield acceleration
according to the maximum force (F2 ¼ F3) of the SDOF system. However, elastic
spectral acceleration is defined at the period T1, which corresponds to initial
242 M. Dolšek et al.

stiffness of the SDOF system. This means that the ductility and reduction factor can
be written as m ¼ u/u2 and R ¼ Sae(T1)/Say, respectively, where Say ¼ F2 /m.
The IDA curves of the response database are computed only for the SDOF
model, which has the dimensional parameters as described above. Other IDA
curves of user-defined SDOF models, if expressed with dimensional quantities,
can be determined by means of simple transformations of the computed IDA
curves. It can be shown that, in addition to the dimensionless parameters and the
period, only one dimensional quantity of the force-displacement relationship of the
user-defined SDOF model is needed. Knowing that the IDA curves of the two
models are equal if expressed in terms of the reduction factor and ductility, the
following relations can be derived:

m FU uU
SU
ae ðT1 Þ ¼
2
S ae ð T1 Þ; u U
¼ 2
u; (12.3)
mU F2 u2

where index U denotes the quantity of the user-defined system. Similarly, as in the
case of the derivation of u2 (Eq. 12.1), it can be shown that:

T12 rv FU
mU ¼ 2
(12.4)
4p2 rh uU
:
2

Incorporating Eqs. 12.1 and 12.4 into Eq. 12.3 it can be shown, that the
relationship between the IDA curves of the two SDOF models (Eq. 12.3) can be
expressed by knowing only one additional parameter, which is, according to our
definition of the SDOF model displacement uU2:

40p2 rh U 40p2 rh U
SU
ae ðT1 Þ ¼
u 2
S ae ð T1 Þ; u U
¼
u
u: (12.5)
T12 g rv T12 g rv 2

Since SU U
ae ðT1 Þ ¼ b
ag and Sae ðT1 Þ ¼ b
ag the spectral acceleration in Eq. 12.5
can be replaced by the peak ground acceleration

40p2 rh U
aU
g ¼
u
ag : (12.6)
T12 g rv 2

Similar transformations can be used for the case when the approximate IDA
curves need to be predicted directly for the top displacement of the building, and not
for displacement of the equivalent SDOF model. Since, in this case, the characteris-
tic displacement u2U;mdof corresponds to the structural model (multi-degree-of-free-
dom model), the additional parameter, the so-called transformation or modification
factor, which relates spectral displacement of an equivalent SDOF system to the top
displacement of the MDOF system, is required. The top displacement of the building
u2U;mdof is then computed as the product of uU
2 and the transformation factor, which is
usually denoted by G or C0, respectively, in the case of the use of the N2 method
(Fajfar 2000) or FEMA 356 (2000).
12 A Toolbox and Web Application for the Seismic Performance. . . 243

Note that in this case, if the approximate IDA curves are predicted directly for the
MDOF model, the dimensionless input parameters are the same as in the case of the
SDOF model. The period T1 corresponds to the SDOF model, but if the quadrilateral
force-displacement is used for idealization of the pushover curve, the first mode
period of the MDOF model can also be used for T1, since the initial stiffness of the
equivalent SDOF model and that of the MDOF model are practically the same.

12.3.3 Implementation of the Methodology Through

the Web Application

There are several reasons in favour of the implementation of the presented

methodology through a web application. Firstly, the database of the IDA curves
can contain huge amount of data, and as such is not appropriate for distribution.
Although anyone can create his own database, it is unlikely that many researchers or
users will do so, since the determination of the database is computationally and time-
demanding and also requires specific knowledge, which is not usually within the
domain of potential users. It is therefore most appropriate that the database is stored
and maintained in one place, and can be easily accessible through the internet.
The infrastructure of the web application was established by following the
characteristics and benefits of cloud computing for the business environment, in
order to increase the IT capacity in real time, and “on the fly” with no significant
investment in new infrastructure, extensive training of personnel or licensing of the
software. Additionally, the concepts introduced through the Web 2.0 revolution
were adopted (simplicity, adaptability, remixability, openness, etc.). For the inter-
ested reader more on the principles of the newly emerging technologies, which will
shape engineering in the future, can be found in Klinc et al. (2009).
The web application was constructed following the traditional three-tier
client-server architecture, which enforces a general separation into three parts: the
client tier (also named the presentation layer or, more specifically, the user inter-
face), the middle tier (business logic), and the data storage tier. Use of three-tier
architecture has some advantages. Firstly, separation of the tiers makes it easier to
modify or replace any tier (or part of a tier) without affecting the other tiers. Next,
separating the application and the database functionality means better load balanc-
ing. For example, the database can reside on a completely different server. Last but
not least, adequate security policies can be enforced within the server tiers without
significant effects on clients.
The top-most level of the application is the user interface, or the presentation tier
of the system. Its main function is to translate tasks and results to something which
the user can understand. In the web application the presentation is disseminated
through a web browser which handles web pages encoded in (X)HTML language
and is generated by a web server on the layer of business logic. Calls between the
user interface and the web server are both synchronous and asynchronous.
For asynchronous calls, the Ajax web programming approach has been applied.
244 M. Dolšek et al.

Fig. 12.2 The sequence diagram of the web application

The business logic layer is based on an Apache web server running on a Linux
platform. Requests are handled by scripts written by using the PHP programming
language, which process the input parameters, interact with the relational database,
parse the results and prepare the output (X)HTML pages. For mathematically
advanced processing, Fortran script is used.
The data storage layer consists of a relational database, where the IDA curves are
stored. On the basis of the excellent connectivity with the PHP programming
language and the Apache web server, it was decided to use the MySQL relational
database. Another reason for this decision is the fact that this is open-source
software, which is freely available. In the test environment, the whole system resides
on one server. However, the architecture is scaleable so that the business logic and
the data layer can be distributed to different physical servers if such requirements
emerge. In the first iteration, the database of the analysis results had approximately
five million records spread over two relational tables, and occupied almost 400 MB
of space. However, after normalization and optimization the database now has
approximately 450,000 records in two relational tables and occupies roughly
170 MB of space, which is enough to store about a quarter of a million IDA curves.
As a consequence, the calculation and response time dropped drastically from the
initial 30 s to less than 3 s (for combined input and output processing).
Details of the activities and the information passing between and within all the
tiers are presented in Fig. 12.2. The events are triggered in the following
(simplified) sequence:
1. The engineer navigates to the web site of the developed application and inserts
the input parameters (e.g. period, available ductility, damping, etc.) of his
structure into the HTML form.
12 A Toolbox and Web Application for the Seismic Performance. . . 245

2. The parameters are passed to the PHP script, which builds an appropriate
MySQL query and queries the MySQL database. The results of these queries
are IDA curves from the database for all the ground motion records and for a
combination of the input parameters, which are in the vicinity of the input
parameters defined in step 1.
3. After receiving the resulting data, the data is processed and the requested
parameters are saved to the file.
4. The Fortran program is called and started, using the saved file as an input. The
result is an output file. In this process approximate IDA curves are computed by
using n-dimensional linear interpolation.
5. PHP script uses the result data passed through the output file, and processes it.
6. The result is sent to the browser and stored to the result file, which is also
available. This file contains approximate IDA curves for all ground motion
records, and the 16th, 50th and 84th fractile IDA curves.
7. In parallel, PHP script is used to prepare graphs of the results and passes them to
the browser, so that the user can immediately see graphic presentations of the
approximate IDA curves.

12.4 Example: Seismic Performance Assessment

of an Eight-Storey Reinforced Concrete Frame
by Using Different Methods

The use of the PBEE toolbox and the web-application are demonstrated by means
of a seismic performance assessment of an eight-storey reinforced concrete frame,
which was performed by employing four methods, i.e. the N2 method, web-based
approximate incremental dynamic analysis, progressive incremental dynamic
analysis, and incremental dynamic analysis. Structural performance was assessed
for two limit states, which are defined on the basis of the damage observed in
the beams and columns. The top displacement and the peak ground acceleration are
adopted as the engineering demand parameter and the intensity measure, respectively.

12.4.1 Description of the Structure, the Structural Model

and the Seismic Loading

The structure under consideration is an eight-story reinforced concrete frame

building, which was initially designed according to the European standard:
Eurocode 8 (Causevic and Mitrovic 2011). After additional design checks, the
strength class of the concrete and the amount of reinforcement was slightly changed
(Kosič 2010) in order to be in full compliance with Eurocode 8, and because the
structure was designed for a peak ground acceleration 0.2 g, soil class B and a
behaviour factor of 3.9.
246 M. Dolšek et al.

Fig. 12.3 Elevation and plan views of the eight-storey RC frame and the reinforcement in typical
cross-sections of the columns and beams

Elevation and plan views of the structure are shown, together with the reinforce-
ment in the columns and beams, in Fig. 12.3, where it can be observed that the height
of the first (lowest) and second storeys is 5 m, whereas the other storeys are 3.1 m
high. The building has three bays in each horizontal direction, with a distance of 8 m
between the centrelines of the columns. All the sections of the columns and beams
in the structure have dimensions 60/60 cm and 40/60 cm, respectively. For the
columns, the steel reinforcement is the same for all sections, except for the sections
at the base, where the density of the stirrups is greater (F8/5 cm and F10/5 cm).
The steel reinforcement for the beams is the same for all cross sections, except
for the beams in first two storeys, where at the top of the beams there are 6 instead
of 4 F20 bars. The concrete cover of the longitudinal reinforcement is 5 cm.
The slabs are 20 cm thick, with steel mesh reinforcement Q226 on top and Q524
underneath. The concrete strength of the building is C30/37, and the steel
strength is B500.
A structural model of the building was prepared in the PBEE toolbox. Since the
height of the stories was determined by the distance between the centrelines of the
12 A Toolbox and Web Application for the Seismic Performance. . . 247

Fig. 12.4 Acceleration

spectra for all 30 ground
motion records, mean
spectrum, 16th, 84th fractile
spectrum and the Eurocode-
based spectrum

beams, the height of the first storey amounted to 4.7 m. The masses were
concentrated at the storey levels, at the centre of gravity. For the vertical loading
a uniformly distributed self-weight (gc ¼ 25 kN/m3) of the structure and a live load
(q ¼ 0.6 2kN/m2) were assumed.
The effective width of the beams was modelled as described in Eurocode 2 (CEN
2004b), assuming zero moment points at the midpoint of the beams. The effective
width of beams in the exterior and interior frames was therefore 1.2 m and 2 m,
respectively. Beam and column flexural behaviour was modelled by one-component
lumped plasticity elements, consisting of an elastic beam and two inelastic rotational
hinges. The moment-rotation relationship before strength deterioration was
modelled by a bi-linear relationship, whereas the post-capping stiffness was
assumed to be linear, with a descending branch. The parameter bu, which controls
the unloading stiffness in the plastic hinges, was assumed to have a value of 0.8.
The seismic loading was defined in two ways, depending on whether the
building’s performance was assessed by using the N2 method or by nonlinear
dynamic analysis. In the later case, a set of 30 ground motion records (Vamvatsikos
and Cornell 2006) was used. The peak ground acceleration of the 30 ground motion
records varied between 0.05 and 0.52 g. The records were selected within events
having a magnitude of between 6.5 and 6.9. All the ground motion records were
recorded on firm soil, with a distance range from the epicentre of 12–55 km. Note
that the same ground motion records were used to establish the response database
for the web-based application for determination of the approximate IDA curves.
The acceleration spectra for each of the 30 ground motion records, and the
corresponding mean, 16th and 84th fractile spectra are presented in Fig. 12.4.
A Eurocode-based acceleration spectrum was defined for estimating the seismic
response parameters when using the N2 method. The estimated parameters of
the acceleration spectrum are S ¼ 1, TB ¼ 0.15 s, TC ¼ 0.47 s and TD ¼ 2 s.
248 M. Dolšek et al.

These parameters were selected in such a way that the mean spectrum of 30 ground
motion records and of the Eurocode-based spectrum are compatible in the range
near the first period of the structure (T1 ¼ 1.76 s), as can be seen in Fig. 12.4.

12.4.2 Seismic Performance Assessment Using

Different Methods

The seismic response parameters, the top displacement and the corresponding peak
ground acceleration, were estimated for the significant damage (SD) and near
collapse (NC) limit states (LS) at the structural level. It is considered that the SD
limit state is violated when the rotation in the plastic hinge of the first column or in
all the beams in one of the storeys exceeds the rotation which corresponds to the
maximum moment in the columns or beams, respectively. Similarly, it is consid-
ered that the NC limit state is violated when the rotation in the plastic hinge of first
column or all the beams in one of the storeys exceeds the ultimate rotation in the
columns or beams, respectively. In dynamic analyses there are two additional
criteria, which control the near collapse limit state at the structural level. The
near collapse limit state is also violated if the average residual top displacement
in last 5 seconds of the analysis exceeds the height of a building by more than 1%,
or if global dynamic instability is reached before the near collapse limit state, which
is defined based on the ultimate rotations in the columns and beams.
Note that the ultimate rotation Yu in the columns corresponds to 80% of
the maximum moment measured in the post-capping range of the moment-rotation
relationship. It was estimated by means of the Conditional Average Estimate (CAE)
method (Peruš et al. 2006). For the beams, the EC8-3 (CEN 2005) formulas were
used to compute the ultimate rotations in the plastic hinges, where the parameter
gel is assumed to be equal to 1.5. Note, that computation of the described limit states
at the level of the plastic hinge is embedded in the PBEE toolbox, which considers
also the P-D effects.

12.4.2.1 Pushover Analysis

Nonlinear static (pushover) analysis was performed by assuming a modal distribution

of the horizontal forces. Since the structure is symmetric, pushover analysis was
performed in one direction only. The pushover curve, together with the points which
indicate the defined limit states, is presented in Fig. 12.5. As described in the
previous section, in the case of the pushover analysis each limit state is defined
by two conditions, which are indicated on the pushover curve. The square indicates
the case if the limit state corresponds to the damage in the columns, while the
cross indicates the case if the limit state is violated on the basis of damage in
the beams. Consequently, the SD limit state at the structural level is controlled by
12 A Toolbox and Web Application for the Seismic Performance. . . 249

Fig. 12.5 A pushover curve, showing the displacements which correspond to the defined limit
states, and the idealized force-displacement relationship

the damage in the columns, whereas the NC limit state is controlled by the damage in
the beams (Fig. 12.5).
The damage at the plastic hinges of the beams and columns for the points
indicated on the pushover curve are presented in Fig. 12.6. Note that, green, yellow
and red colours represent, respectively, yielding of reinforcement, the state of
exceeding the maximum moment, and the state of exceeding the ultimate rotation,
whereas the grey part of the structure does not suffer any damage. Figures 12.6a, b
correspond to the SD limit state, respectively, to the case when the maximum
moment is exceeded in the first column, and when the maximum moment is
exceeded in all the beams in the first storey, which are oriented in the direction of
loading. In the other two figures (Figs. 12.6c, d), the damage in the elements is
presented for the NC limit state. More specifically, Fig. 12.6c presents the case
when the ultimate rotation is exceeded in all the beams in the first storey, whereas in
Fig. 12.6d the ultimate rotation is exceeded in the columns for the first time.
The strength of the structure is, in the case of the NC limit state, reduced by 33%
with respect to the maximum strength (Fig. 12.6c). The majority of the damage is
concentrated in the first and second storeys. The upper stories remain almost
undamaged.

12.4.2.2 The Incremental N2 Method

The first method used to determine seismic demand (maximum top displacement)
for different levels of seismic intensity (peak ground acceleration) was the N2
method (Fajfar 2000), which results in an IN2 curve (Dolšek and Fajfar 2007).
250 M. Dolšek et al.

Fig. 12.6 The distribution of damage at different limit states as indicated on the pushover curve in
Fig. 12.5 for the case (a) “SD columns”, (b) “SD beams”, (c) “NC beams” and (d) “NC columns”

This method is well-known, and involves the results of pushover analysis and the
inelastic response spectrum, which can be easily computed on the basis of
the defined Eurocode-based acceleration spectrum shown in Fig. 12.4.
In order to determine the IN2 curve, the pushover curve was idealized as
presented in Fig. 12.5. The maximum strength of the idealized force-displacement
relationship was set equal to the maximum strength determined by the pushover
analysis. Note, that in the case that the structure’s performance is assessed by the N2
method, only the ideal elasto-plastic force-displacement relationship of the idealized
force-displacement relationship is needed, whereas in the case of the web-based
approximate IDA (WIDA) curves and progressive incremental dynamic analysis
estimation of the degrading part of the idealized force-displacement relationship is
also required. In order to be consistent, the idealized force-displacement relationship
is equal for all the analyses, whereas the mass of the equivalent SDOF model m* and
the transformation factor G, amounted to 2,860 t and 1.21, respectively.
Since the period of the equivalent SDOF model T * ¼ 1.76 s is longer than the
period TC ¼ 0.47 s corresponding to the acceleration spectrum (Fig. 12.4), it can be
assumed, according to the N2 method, that the elastic and inelastic displacements
are equal, and consequently the IN2 curve is a straight line up to the top displace-
ment, which corresponds to the NC limit state. Definition of the ultimate point on
12 A Toolbox and Web Application for the Seismic Performance. . . 251

Fig. 12.7 IN2 curve with

marked significant damage
and near collapse limit states

the IN2 curve is required, since in the case of the N2 method the inelastic spectra are
defined on the assumption of unlimited ductility of the equivalent SDOF model.
The IN2 curve is presented in Fig. 12.7. It can be seen that the SD and NC limit
states are violated if the peak ground acceleration exceeds 0.57 g and 0.76 g,
respectively.

12.4.2.3 Web-Based Approximate Incremental Dynamic Analysis (WIDA)

The approximate IDA curves can be determined based on the results of pushover
analysis in conjunction with the developed web application (http://ice4risk.
slo-projekt.info/analysis). The engineering demand parameter of the approximate
IDA curves can be the top displacement of the structure or the top displacement of
the equivalent SDOF model, whereas the intensity measure can be the peak ground
acceleration and the spectral acceleration corresponding to the first mode period.
The input parameters describing the tri-linear (bi-linear with softening) force-
displacement relationship of the pushover curve are the available ductility
ma ¼ 3.92, and the post-capping stiffness parameter a ¼ 0.36, which are dimen-
sionless parameters, and the displacement at yielding uy for the structural model
(0.097 m). The other parameters which are needed to derive the approximate
fractile IDA curves are the transformation factor G ¼ 1.21, the period T1 ¼ T *
¼ 1.76 s and the critical damping ratio, which, in the case of presented example, is
assumed to be 0.05.
The approximate IDA curves and the corresponding 16th, 50th and 84th fractile
IDA curves are presented in Fig. 12.8 together with highlighted points which
indicate the SD and NC limit states. Note, that the limit-state top displacement of
the approximate IDA curves (i.e. the WIDA curves), corresponds to the limit-state
top displacement presented on the pushover curve (Fig. 12.5).
252 M. Dolšek et al.

Fig. 12.8 Approximate IDA

(i.e. WIDA) curves and
corresponding 16th, 50th and
84th fractile IDA curves
determined by using the web
application, with indication of
the significant damage (SD)
and near collapse (NC) limit
states

12.4.2.4 Progressive Incremental Dynamic Analysis

Progressive IDA involves a precedence list of ground motion records, which was,
for the presented example, determined from the IDA curves for the equivalent
SDOF model and the simple optimization procedure (Azarbakht and Dolšek 2011).
The same equivalent SDOF model as in the case of the N2 method was used. By
using the PBEE toolbox the precedence list was determined and then IDA was
performed, firstly, only for the first six records (two subsets (s ¼ 2) of three
records) from the precedence list. Since the tolerance in this case exceeded the
acceptable tolerance, which is 8% in this case, IDA had to be performed for
additional subsets of records from the precedence list of records. Finally, the
progressive IDA was terminated, since the tolerance became acceptable if IDA
was performed for the first four subsets (s ¼ 4) of records (12 records). In this case
(s ¼ 4) the difference between the 16th, 50th and 84th fractile IDA curves with
respect to the fractile IDA curves from the previous case (s ¼ 3) was minor, and
amounted to less than 8%. The IDA curves for the first 12 ground motion records
from the precedence list of records and the corresponding 16th, 50th and 84th
fractile IDA curves, with indication of the significant damage (SD) and near
collapse (NC) limit states, are presented in Fig. 12.9. Note that the top
displacements corresponding to the SD or NC limit state, and for a given set of
ground motion records, are not equal to those obtained in the case of the WIDA
curves, since they are determined directly from the results of dynamic analysis, and
not from pushover analysis as in the case of WIDA.

12.4.2.5 Incremental Dynamic Analysis

IDA is the most accurate analysis method used in the case of the presented example.
The global dynamic instability of each of the 30 IDA curves was estimated
12 A Toolbox and Web Application for the Seismic Performance. . . 253

Fig. 12.9 The IDA curves

for first 12 ground motion
records (s ¼ 4) from the
precedence list of records and
the corresponding 16th, 50th
and 84th fractile IDA curves,
with indication of the
significant damage (SD) and
near collapse (NC) limit states

with a precision of 0.02 g. The largest interval between the peak ground accelera-
tion, for which the seismic response parameters were computed, was defined as
being equal to 0.05 g. However, if the peak ground acceleration, which corresponds
to global dynamic instability, was large, the IDA curves were computed for only 20
points, since the accuracy of predicting these IDA curves is not very important. The
same procedure was used in the case of progressive IDA, as well as for determina-
tion of the precedence list of ground motion records.
Each nonlinear dynamic analysis within the IDA was calculated by employing
the Newmark integration scheme, assuming gn ¼ 0.5, bn ¼ 0.25 and an integration
time interval of 0.005 s.
The IDA curves for a set of 30 records and the corresponding 16th, 50th and 84th
fractile IDA curves, with indication of the significant damage (SD) and near
collapse (NC) limit states, are presented in Fig. 12.10.

12.4.2.6 Discussion

The fractile IDA curves, determined by three different approaches and the IN2
curve, are compared in Fig. 12.11. The difference between the IDA curves is minor,
even in the case of the NC limit state, where the structure is already severely
damaged. This is because the structural damage is concentrated only in the bottom
two storeys, which are significantly higher if compared to the height of other
stories. However, such mechanism should not occur, since the structure has been
designed according to capacity design rules defined in Eurocode 8. The largest
difference between the IDA curves is observed for the 16th fractile IDA curves,
where the difference between the results of IDA and approximate IDA expressed in
terms of peak ground acceleration corresponding to the NC limit state is less than
15%. The points in Fig. 12.11 presents the 16th, 50th and 84th fractiles of the peak
ground acceleration, which cause the median top displacement corresponding to the
254 M. Dolšek et al.

Fig. 12.10 The IDA curves

for a set of 30 records and the
corresponding 16th, 50th and
84th fractile IDA curves, with
indication of the significant
damage (SD) and near
collapse (NC) limit states

Fig. 12.11 The 16th, 50th

and 84th fractile IDA curves
determined by IDA,
progressive IDA, web-based
approximate IDA, and the
IN2 curve. The highlighted
points on the curves indicate
the SD and NC limit states

SD or NC limit state, and are also presented in Table 12.1. It can be observed that
the difference between the median estimates based on the simplified methods and
that of IDA rarely exceeds 10%. A somewhat larger error is observed in the case of
the 16th and 84th fractiles if these are estimated by the web-application, whereas
progressive IDA practically produces same results as IDA.
However, the largest error is observed in the case of the prediction of global
dynamic instability (Fig. 12.11). On the other hand, the seismic performance
assessment by using simplified methods requires significantly less computational
time. It is estimated that the time required for assessing the structural performance
in the case of the IN2 or WIDA is only about 1% if compared to the computational
time needed for IDA.
12 A Toolbox and Web Application for the Seismic Performance. . . 255

Table 12.1 The 16th, 50th and 84th fractiles of peak ground acceleration (PGA) causing the SD
and NC limit states, and the corresponding median top displacement (MTD)
Comp. SD limit state NC limit state
Method time (%) Fractile MTD (m) PGA (g) MTD (m) PGA (g)
IN2 1 50 0.36 (11%) 0.57 (6%) 0.47 (5%) 0.76 (3%)
Approximate IDA 1 16 0.36 (11%) 1.43 (16%) 0.47 (5%) 1.61 (14%)
50 0.57 (5%) 0.69 (6%)
84 0.27 (13%) 0.31 (3%)
Progressive IDA 50 16 0.33 (4%) 1.26 (2%) 0.44 (2%) 1.82 (3%)
50 0.55 (2%) 0.72 (2%)
84 0.26 (6%) 0.33 (3%)
IDA 100 16 0.32 1.23 0.45 1.87
50 0.54 0.74
84 0.24 0.32
The values in brackets represent an error with respect to results of the IDA.

The most accurate method among the employed methods presented in this
chapter is incremental dynamic analysis, which is widely used for the performance
assessment of structures under seismic loading. It is shown that, by using progres-
sive incremental dynamic analysis, computational time can be reduced for about
50% compared to that needed to perform IDA. It was proved, at least for the
presented example, that the seismic response parameters are sufficiently accurate
when determined by progressive IDA, which requires only 12 IDA curves
instead of 30, which was the number of records used in the IDA. Nevertheless
there is some additional work with the preparation of the precedence list.
The web-based approximate IDA curves are easily determined based on the
parameters of the idealized pushover curve, and can be obtained in a few seconds.
However, the use of the web-application is currently limited, since the response
database is established for only 30 records.
The IN2 method is the simplest method, which is presented in this chapter.
It cannot estimate the global dynamic instability, nor consider record-to-record
variability. However, all the simplified methods provided sufficiently accurate
results, although they are based on response of an equivalent SDOF model.

12.5 Conclusions

A PBEE toolbox for the seismic performance assessment of RC frames, which

consists of Matlab functions and works in conjunction with OpenSees, and a web-
application for the prediction of approximate IDA curves, have been briefly
presented. The advantage of the web-application in comparison with the traditional
approaches which are used for the determination of the approximate seismic response
of structures can be found in the systematic separation of the process of the
parametric study and the process of the prediction of the approximate IDA curves.
256 M. Dolšek et al.

Such an approach enables the creation of many response databases, which would
eventually result in a global component for the prediction of approximate IDA curves
at any location in the world and for any kind of structural system. On the other hand,
the PBEE toolbox can serve for code calibration, as a basis for new applications, e.g.
for the design of frames based on nonlinear analysis, or for parametric studies, since it
enables rapid definition of simple nonlinear models of RC frames.
In the presented examples it was shown that it is possible to predict the 16th,
50th and 84th fractiles of the seismic response parameters for the SD and NC limit
states with sufficient accuracy by using WIDA, which is based on pushover analysis
and requires only about 1% of computational time with respect to the computational
time needed for IDA. In this case the median peak ground acceleration which
causes the defined limit states was estimated with an error of less than 10%,
whereas the error in predicting the corresponding 16th and 84th fractiles was
slightly greater. However, if such an error is too large, it is possible to use
progressive IDA, which predicts the seismic response parameters with minor errors
and with only about 50% of the computational time required for IDA.

Acknowledgements The results presented in this paper are based on work supported by the
Slovenian Research Agency within the framework of the project High-throughput computing
environment for seismic risk assessment (http://ice4risk.slo-projekt.info/) (J2-0845-0792-08).
This support is gratefully acknowledged.

References

Antoniou S, Pinho R (2004) Development and verification of a displacement-based adaptive

pushover procedure. J Earthquake Eng 8:643–661
Aydinoğlu MN (2003) An incremental response spectrum analysis based on inelastic spectral
displacement for multi-mode seismic performance evaluation. Bull Earthquake Eng 1:3–36
Azarbakht A, Dolšek M (2011) Progressive incremental dynamic analysis for first-mode
dominated structures. J Struct Eng 137:445–455
Causevic M, Mitrovic S (2011) Comparison between non-linear dynamic and static seismic
analysis of structures according to European and US provisions. Bull Earthquake Eng
9:467–489
CEN (2004a) Eurocode 8: Design of structures for earthquake resistance. Part 1: General rules,
seismic action and rules for buildings. EN 1998-1, European Committee for Standards,
Brussels, Dec 2004
CEN (2004b) Eurocode 2: Design of concrete structures. Part 1-1: General rules and rules for
buildings. EN 1992-1-1, European Committee for Standards, Brussels, Dec 2004
CEN (2005) Eurocode 8: Design of structures for earthquake resistance. Part 3: Strengthening and
repair of buildings. EN 1998-3, European Committee for Standards, Brussels, Mar 2005
Chopra AK, Goel RK (2002) A modal pushover analysis for estimating seismic demands for
buildings. Earthquake Eng Struct Dyn 31:561–582
CSI (2006) PERFORM-3D User guide, Version 4. Computers & Structures Inc., Berkeley
CSI (2008) SAP 2000 User manual, Version 12. Computers & Structures Inc., Berkeley
Dolsek M (2010) Development of computing environment for the seismic performance assessment
of reinforced concrete frames by using simplified nonlinear models. Bull Earthquake Eng
8:1309–1329. doi:10.1007/s10518-010-9184-8
12 A Toolbox and Web Application for the Seismic Performance. . . 257

Dolšek M, Fajfar P (2007) Simplified probabilistic seismic performance assessment of plan-

asymmetric buildings. Earthquake Eng Struct Dyn 36:2021–2041
Ellingwood BR, Kinali K (2009) Quantifying and communicating uncertainty in seismic risk
assessment. Struct Saf 31:179–187
Elnashai AS, Papanikolaou VK, Lee DH (2008) Zeus NL – A System for Inelastic Analysis of
Structures, User Manual, Version 1.8.7
Fajfar P (2000) A nonlinear analysis method for performance-based seismic design. Earthquake
Spectra 16:573–592
Fajfar P, Dolšek M, Marušić D, Stratan A (2006) Pre- and post-test mathematical modelling of a
plan-asymmetric reinforced concrete frame building. Earthquake Eng Struct Dyn
35:1359–1379
FEMA (2000) Prestandard and commentary for the seismic rehabilitation of buildings. FEMA
publication 356, Washington DC, Nov 2000
FEMA (2005) Improvement of nonlinear static seismic analysis procedures. FEMA publication
440, Washington, DC, June 2005
Klinc R, Dolenc M, Turk Ž (2009) Engineering collaboration 2.0: requirements and expectations.
ITcon 14:473–488
Kosič M (2010) Design of eight-storey reinforced concrete frame building according to European
standard Eurocode 8. Personal communication, 15 Dec 2010
MathWorks (2007) MATLAB the Language of Technical Computing. http://www.mathworks.
com/
McKenna F, Fenves GL (2007) Open system for earthquake engineering simulation, Pacific
Earthquake Engineering Research Center, Berkeley. http://opensees.berkeley.edu
MIDAS (2008) Pushover Analysis of RC structures as per EC8:2004, MidasGen Version 7.4.1,
MIDAS Information Technologies Co, Ltd, Korea
Peruš I, Poljanšek K, Fajfar P (2006) Flexural deformation capacity of rectangular RC columns
determined by the CAE method. Earthquake Eng Struct Dyn 35:1453–1470
Peruš I, Klinc R, Dolenc M, Dolšek M (2011) A web-based methodology for the prediction of
approximate IDA curves. Earthquake Eng Struct Dyn (in review).
Priestley MJN, Kowalsky MJ (2000) Direct displacement-based seismic design of concrete
buildings. Bull New Zeal Soc Earthquake Eng 33:421–444
SEISMOSOFT (2007) Seismostruct version 4. SeismoSoft Ltd., Messina
Takeda T, Sozen MA, Nielsen NN (1970) Reinforced concrete response to simulated earthquakes.
J Struct Division 96:2557–2573
Vamvatsikos D, Cornell CA (2002) Incremental dynamic analysis. Earthquake Eng Struct Dyn
31:491–514
Vamvatsikos D, Cornell CA (2006) Direct estimation of the seismic demand and capacity of
oscillators with multi-linear static pushovers through IDA. Earthquake Eng Struct Dyn
35:1097–1117