You are on page 1of 41

Probabilistic Evaluation of 2-D Steel Moment Frames using

Uncertainty Modeling and Performance based Earthquake Design


for Mumbai City

A thesis submitted to
Indian Institute of Technology, Kharagpur
In partial fulfillment for the award of the degree

Of

Master of Technology in Civil Engineering


With specialization in
Structural Engineering

by
Thontepu Sri Kalyan (10CE31010)

Under the supervision of


Prof. Baidurya Bhattacharya

Department of Civil Engineering


Indian Institute of Technology, Kharagpur
April, 2015
DECLARATION

I certify that
(a) The work contained in this report has been done by me under the guidance of my supervisor.
(b) The work has not been submitted to any other Institute for any degree or diploma.
(c) I have conformed to the norms and guidelines given in the Ethical Code of Conduct of the
Institute.
(d) Whenever I have used materials (data, theoretical analysis, figures, and text) from other
sources, I have given due credit to them by citing them in the text of the thesis and giving their
details in the references. Further, I have taken permission from the copyright owners of the
sources, whenever necessary.

Date: Thontepu Sri Kalyan

Place: Kharagpur 10CE31010

1
DEPARTMENT OF CIVIL ENGINEERING
INDIAN INSTITUTE OF THECHNOLOGY, KHARAGPUR
KHARAGPUR-721302, INDIA

CERTIFICATE

This is to certify that the thesis entitled, “Probabilistic Evaluation of 2-D Steel Moment
Frames using Uncertainty Modeling and Performance based Earthquake Design for
Mumbai City” is a bonafide record of authentic work carried out by Thontepu Sri Kalyan
(10CE31010) under my supervision and guidance for the partial fulfillment of the requirement
for the award of the degree of Master of Technology in Structural engineering in the department
of civil engineering at the Indian Institute of Technology, Kharagpur.

The results embodied in this thesis have not been submitted to any other University or Institute
for the award of any degree or diploma.

Prof. Baidurya Bhattacharya


Date: Department of Civil Engineering
Place: Kharagpur Indian Institute of Technology
Kharagpur, India

2
Acknowledgement

I would like to express my sincere gratitude to Prof. Baidurya Bhattacharya for his
encouragement, guidance, and support and Ms. Radhika Bayya for her help in establishing the
literature review. Without the support and encouragement from these people this project would
not have been possible. Especially, I would like to give my special thanks to my parents and my
family whose patient love enabled me to sustain in a competitive environment.

3
Contents

SYNOPSIS ........................................................................................................................... 6

1.0 INTRODUCTION ............................................................................................................... 7

2.0 OBJECTIVE AND SCOPE ................................................................................................. 8

3.0 LITERATURE REVIEW .................................................................................................... 9

4.0 METHODOLOGY ............................................................................................................... 18

4.1 DEMAND UNCERTAINTIES ........................................................................................... 19

4.2 CAPACITY UNCERTAINTIES ........................................................................................ 20

4.3 MATHEMATICAL FORMULATION .............................................................................. 21

4.4 INCREMENTAL DYNAMIC ANALYSIS (IDA) ............................................................ 22

4.5 COMPUTING THE DRIFT CAPACITY ........................................................................... 22

5.0 MODEL BUILDING AND ANALYSIS.............................................................................. 23

5.1 MULTI STOREY FRAME MODEL.................................................................................. 23

5.2 SAP 2000 MODEL ............................................................................................................. 25

6.0 RESULTS ............................................................................................................................. 27

6.1 EIGENVALUE ANALYSIS............................................................................................... 28

6.2 PUSHOVER ANALYSIS ................................................................................................... 29

6.3 NONLINEAR TIME HISTORY ANALYSIS .................................................................... 30

6.4 EVALUATION OF DEMAND DISTRIBUTION ............................................................. 32

6.5 CAPACITY DISTRIBUTION ............................................................................................ 35

7.0 FINAL RELIABILITY EVALUATION .............................................................................. 37

8.0 CONCLUSION ..................................................................................................................... 38

9.0 REFERENCES ..................................................................................................................... 39

4
List of Figures and Tables

Figure 1: Design Response Spectrum according to IS 1893-2002 .............................................. 11

Figure 2: Sample flowchart depicting the logic tree method ........................................................ 17

Figure 3: Computation of drift capacity of structure .................................................................... 23

Figure 4: Four Storey 3 Bay Frame structure used for analysis. .................................................. 24

Figure 5: SAP 2000 model of the 4 storey 3 bay frame structure................................................. 25

Figure 6: Response spectra for different site conditions at Mumbai ............................................ 26

Figure 7: Four Storey 3 Bay Frame structure (a) First Mode of vibration (b) Second mode ...... 27

Figure 8: Pushover Curve obtained from SAP 2000 model ......................................................... 29

Figure 9: Pushover curve obtained from OPENSEES .................................................................. 29

Figure 10: Hinge Failure Sequence during pushover analysis in SAP model ............................. 30

Figure 11: Time history plot of first floor IDR when subjected to San Fernando Earthquake .... 31

Figure 12: Typical variation of IDR (%) over height of frame for 3 different earthquakes ........ 31

Figure 13: Interactive figure depicting the inverse transformation technique ............................. 32

Figure 14: Distribution of  factor values corresponding to each uncertainty ............................ 34

Figure 15: Final demand distribution plots with only record variability and total uncertainty ... 34

Figure 16: Sample IDA Curves of the multi storey frame ........................................................... 36

Table 1: Statistical characteristics of the input random variables for model building ................. 21

Table 2: Time period comparison from eigenvalue analysis ....................................................... 27

Table 3: Comparison of Design base shear and the base shear strength from pushover analysis 28

Table 4: Comparison of base shear strength from pushover analysis of OPENSEES and SAP .. 28

Table 5: Dispersion values obtained for different quantities under consideration ....................... 35

Table 6: Probability of failure for different performance levels…………………………………35

5
Synopsis

This work primarily deals with the study of seismic reliability analysis of steel structures in the
city of Mumbai using the latest SAC/FEMA guidelines. A detailed literature survey is done
covering various aspects of Performance based earthquake engineering and an interactive
flowchart is prepared from the findings. The work consists majorly of two parts. The first part of
work includes the establishment of demand and capacity levels for the 4-storey 3-bay steel
moment frame under consideration. For this a total of 44 real time earthquake ground excitation
records are selected as prescribed by FEMA-695. The Hazard Curve developed for Mumbai city
has been employed here to obtain spectral accelerations corresponding to different hazard levels.
Hazard level is randomly generated and the corresponding spectral acceleration value is recorded
from the hazard curve and the records are then accordingly scaled. Then nonlinear dynamic
analysis is done on the frame to yield the record to record variability of demand (in terms of
maximum inter storey drift ratio (IDRmax). Both material as well geometric non-linearity (to
include P- effects) are accounted for in the OPENSEES code. The model validation is also done
using eigenvalue analysis and Pushover analysis where the results are compared with those
obtained from standardized software (SAP 2000). After the demand distribution is obtained the
capacity distribution is obtained from Incremental Dynamic Analysis procedure. Failure criteria
or capacity value is defined as the point where the slope of the IDA curve becomes less than
0.1*Se (Se = Elastic slope). Thus the record to record variability in the capacity is evaluated.

The second part of the work consists of accounting for various uncertainties arising from
different sources both in demand and capacity estimation. For this various uncertainty modelling
techniques were employed based on existing literature methods. The variation in demand
estimation is accounted by evaluating the dispersion measures (βuD) considering the uncertainty
and randomness present in material properties of steel (yield strength, elastic modulus of steel),
damping ratio, mass and live load on frame structure. Similarly the variation in capacity is
accounted by considering the variation resulting from the modelling approximations and material
properties randomness (βuC). Thus the final distribution properties for demand is used to estimate
the probability of failure at three performance levels defined in FEMA 273 when capacity is
assumed to be deterministic. Capacity uncertainty is also then included by considering the yield

6
criteria capacity distribution obtained from nonlinear time history analysis. Thus in the next step
for the evaluation of failure probability both the developed demand and capacity distributions are
used and failure probability is found analytically.

1. Introduction

Earthquakes can be considered as one of the greatest natural threats on account of its devastating

effects on the economy and lives of people in a country. In wake of the recent Sikkim earthquake

and the devastating effects it had on the structures of that region, questions were raised against

the design codes used in practice and the structural reliability of the existing structures.

Occurrence of any future earthquake of such magnitude in Mumbai city can have severe effects

both in terms of life loss and economic disturbance.

In present days the users expect a certain degree of reliability levels from the knowledge that the

buildings are designed based on modern codes. But it is often not economically feasible to

construct the buildings with such huge conservativeness and a reduction in this conservativeness

leads to an increase in damage levels. The damage which the building can be subjected thus

depends on how the modern codes obtain a trade-off between the conservativeness and damage

levels. Thus we now prefer a solution where the conservativeness is reduced to a certain extent

and are now forced to establish the degree of safety and reliability levels of this new design

method using advanced seismic reliability evaluation techniques.

Reliability analysis of structures often implies an estimation failure probability when the

designed structure is subjected to the force which it is supposed to resist. The evaluation of the

seismic reliability of a structure at a given location is not so straight forward and there is no well-

defined/established procedure for its calculation. The major problems that need to be dealt with

while evaluating this reliability is the randomness and the uncertainty involved in the Capacity

and Demand parameters. Structural response/Demand (moments, stresses or displacements) due

7
to earthquakes are random in nature owing to randomness in the seismic excitation itself. Also

this randomness is often combined with imperfections in Structural modeling procedure adopted,

material property uncertainties, and earthquake orientation uncertainties. Thus the most

important aspect of seismic reliability study is the consideration of these uncertainties

Hence accounting for these uncertainties in the reliability analysis carries utmost importance and

the accuracy of the procedure depends upon the modeling techniques used to include these

uncertainties.

2. Objective and Scope

2.1.Development of 4 Storey 3 Bay Steel Moment Frame model in OPENSEES including


Geometric and Material non-linearity effects.

 Perform eigenvalue analysis for time period evaluation of different modes

 Perform static pushover analysis to estimate the ultimate capacity of the frame structure

2.2. Development of 4 Storey 3 Bay Steel Moment Frame model in SAP 2000.
 Perform eigenvalue and pushover analysis of the frame structure and compare the results
obtained with that of OPENSEES.

2.3. Reliability analysis and uncertainty modeling of 4 Storey 3 Bay Steel Moment Frame

 Evaluation of record to record uncertainty in Demand and Capacity parameters

 Demand and Capacity uncertainty determination arising from various sources

 Monte Carlo simulations for 10 million samples to evaluate the reliability based on the
final distribution properties of demand and capacity parameters.

8
3. Literature Review

3.1 Performance Based Design:

The concept of performance-based earthquake engineering (PBEE) has its roots from the above-

mentioned realization. Vision 2000 report [SEAOC, 1995] is one of the early documents of the

first-generation PBEE in the U.S. Performance-based earthquake design (PBED) is defined as a

design framework which results in the desired system performances at various intensity levels of

seismic hazard. The system performance levels are often classified into three types namely: fully

operational, operational, life safety, and near collapse whereas the hazard levels to be considered

for design are classified as frequent, occasional, rare, and very rare events. The designer consults

with the owner to approach at a conclusion over the hazard level and performance level for

which the structure is to be designed. The intended performance levels corresponding to different

hazard levels are either determined based on the public resiliency requirements, e.g., hospital

buildings, or by the private property owners, e.g., residential or commercial buildings.

Subsequent documents of the first-generation PBEE namely, ATC-40 [ATC, 1996], FEMA-273

[BSSC, 1997], and FEMA-356 [ASCE, 2000] express the design objectives using a similar

framework, with slightly different performance descriptions and hazard levels. The procedure

prescribed in each of these documents is compiled and expressed in the form of a flowchart

given below.

The element deformation and force acceptability criteria corresponding to the performance are

specified for different structural and non-structural elements for linear, nonlinear, static, and/or

dynamic analyses. These criteria do not possess probability distributions, i.e., the element

performance evaluation is deterministic. The defined relationships between engineering demands

and component performance criteria are based on relationships measured in laboratory tests, for

which an analytical model is then fitted to obtain an equation form [Wen, 2003]. In addition, the

element performance evaluation is not tied to a global system performance. Considering the

9
shortcomings of the first-generation procedures, a more robust PBEE methodology was

developed in the Pacific Earthquake Engineering Research (PEER) Centre. A key feature of the

methodology is the calculation of performance in a rigorous probabilistic manner without relying

on expert opinion. Accordingly, uncertainties in earthquake intensity, ground motion

characteristics, structural response, physical damage, and economic and human losses are

explicitly considered in the method. The increasing trend to use probabilistic PBED as a design

method is justified by the consequences of recent earthquakes, where the traditional earthquake

design philosophy has occasionally fallen short of meeting the requirements of sustainability and

resiliency.

3.2 Indian Earthquake Resistant Design Scenario:

The design base shear prescribed by the Indian Standard [6] is given by:
Vb  Ah w
(1)

Where ‗w‘ is the seismic weight of the structure, ‗Ah‘ is a horizontal acceleration response
spectrum factor, given by:
ZIS a
Ah 
2 Rg (2)

Where ‗Z‘ is the zone factor, ‗I‘ is the importance factor (taken as 1.0), ‗R‘ is the response
reduction factor to consider ductile response effects (taken as 5.0 which is prescribed for
Moment Resisting Frames), ‗g‘ is the acceleration due to gravity and ‗Sa‘ is the value obtained at
the natural time period from the response spectrum.

The response spectrum for the Indian Standard is a function of Tn alone and is thus independent
of hazard levels for a region. There are 3 spectra defined for different soil types and a sample
design response spectrum is plotted in Fig

10
25

Spectral Acceleration (m/s/s)


20

15

10

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Time period (s)

Fig 1: Design Response Spectrum according to IS 1893-2002. [6]

The natural time period of a structure of height ‗h’ is given by:

Tn  0.085h 0.75
(3)

The base shear obtained is distributed as lateral forces among the different storeys of the
structure according to the following equation.

Vb hi2 mi
Fi 
h m i
2
j
(4)

Drawbacks: It can be observed that nowhere in the design procedure followed by the Indian
code are the uncertainties in various modeling techniques and time period evaluation is
considered. Also the hazard levels are quantified by rather a simple parameter called zone factor
which covers vast regions of India without providing much region specific hazard specific data.
Also a standard linear static analysis method is used to design the frame instead of relying on
advanced nonlinear analysis techniques. And the material properties and other design parameters
are assumed to be deterministic when opposed to real world scenarios where they are random in
nature. Thus in order to account for all such drawbacks one should consider probabilistic based
earthquake design with emphasis of seismic reliability analysis.

11
Interactive Flowchart summarizing the various steps involved in the Performance based earthquake design procedure
All previous
earthquakes data
1. Location
2. Magnitudes
3. Annual rate of
occurrence

Develop Attenuation Relationships


for different magnitudes
(PGA/Sa vs Distance)

Obtain the lognormal distribution


properties of Intensity measure
considered (PGA/Sa) 1. Frame Member
Section Properties
2. Nonlinear Construct Loss functions which
Material Properties represents the probability that loss will
3. Pre-Loading/ be less than or equal to an indicated
Mass conditions amount, given that the building is
Based on Lognormal properties damaged to a given level
estimate the Probability of
exceedance in T years and Three types of losses are typically
If Economic loss
generate Hazard curve considered:
estimation is the final
Rn (>lnPGA) = rnPn(>lnPGA) Direct loss to the building
Decide the analysis type to be objective
Ppois (> lnPGA,T) = 1- e-R T Indirect loss due to time lost in
implemented restoring the building to its original
1. Linear Static Analysis Damage Estimation based Quantification of direct damage state
2. Nonlinear Static Analysis on the final objective level sustained by the structure Casualities/No of deaths and injured
3. Linear Dynamic Analysis
If Matching method is 4. Nonlinear Dynamic Analysis
to be used If Structural Design is
Match the response spectrum
of the structure to that of design Matching/Scaling Method the final objective
spectrum corresponding to the for earthquake record
hazard level required generation Construct loss functions by
Develop Fragility Curves 1. Using sufficient historical data on
Structural Capacity Estimation
(Prob of exceedance vs EDP) actual losses incurred in past
Decide the Engineering Design through
If Scaling method is to corresponding to different earthquake
parameter to be considered. 1. Incremental Dynamic Analysis
be used damage levels of structure based 2. Estimating loss associated with a
Prominent ones are: 2. Pushover Analysis.
on experimental results or damage level given we know the
Scale the present earthquake records 1. Interstorey Drift ratio 3. Direct Capacity values
previous data from past fragility curve corresponding to that
based on the Spectral acceleration value 2. Column Shear suggested by codes (FEMA 350)
earthquakes damage level.
from hazard curve corresponding to a 3. Floor Acceleration
hazard level of the site

If Economic loss
Inputs estimation is the final
Estimate the economic
1. Location Hazard Specific Demand estimation objective
2. Type of Structure to be concerned ground motions from from the model and Capacity/ Damage loss associated with
3. Soil properties occurence of earthquake
4. Structural Design or Economic loss hazard analysis the ground motions Analysis
estimation based on the fragility
generated previously curve and loss functions
If Structural Design is
the final objective

Check Demand vs Capacity and


IS 1893, 2002 redo the design process if the
prescribed limits on confidence
levels or reliability are not met
FEMA 273
FEMA 350
ATC 58

Fig 1: Interactive flowchart summarizing the various steps in the performance based earthquake
design procedure

12
3.3 Seismic Reliability analysis techniques:

The concept of seismic reliability analysis of steel moment frame structures was developed in

1980‘s as a need to assess the reliability and performance levels of significant structures. The

design procedure slowly evolved from a simple elastic static pushover analysis to a non-linear

dynamic analysis. In the recent SAC/FEMA effort (Somerville et al 1997), the selection of the

ground excitations to match the spectral acceleration with a given probability of exceedance is

extended for all periods. Probabilistic Basis for the SAC/FEMA Steel Moment Frame Guidelines

were well established recently (Cornell et al. 2002). The seismic reliability estimation method

performance evaluation for steel moment frames based on the developed probabilistic basis is

also developed (Yun et al. 2002).

The scaling factors required to match the earthquake records to the required spectral acceleration

are to be generated from the Uniform Hazard Spectrum (UHS) of the given site. UHS for the city

of Mumbai and the attenuation relationship was developed recently using the various past

earthquake records/data (Raghukanth et al. 2006). Incremental Dynamic analysis method was

developed (Vamvatsikos and Collins 2002) as an extension to static pushover analysis to

calculate the values of required parameters by non-linear dynamic analysis method.

FEMA-273 proposed a set of design guidelines for different performance levels for seismic

response of steel moment frames. Three performance levels are proposed, namely Immediate

Occupancy (IO) level, Life Safety (LS) level and Collapse Prevention (CP) level. Immediate

Occupancy level is the situation when minor local yielding occurs at a few places and no

fractures has been detected. Life Safety level is the situation when plastic hinges form at the

frame members. Collapse Prevention level is defined as the situation when extensive distortion

of beams and columns panels is observed with a large number of fracture regions. FEMA-273

suggests an inter storey drift limit values of 0.7% for IO level, 2.5% for LS and 5% for CP.

13
3.4 Challenges in Reliability estimation:

The major problems that need to be dealt with while evaluating this reliability is the randomness

and the uncertainty involved in the Capacity and Demand parameters. Structural

response/Demand (moments, stresses or displacements) due to earthquakes are random in nature

owing to randomness in the seismic excitation itself. Also this randomness is often combined

with imperfections in Structural modeling procedure adopted, material property uncertainties,

and earthquake orientation uncertainties. Thus the most important aspect of seismic reliability

study is the consideration of these uncertainties.

But the uncertainty treatment is only applied in the selection of design earthquake ground motion

parameters based on return period until now. But as risk analysis is gathering a much rather

importance in recent period there is an increasing need to include other uncertainties as well

since the loss estimate and strategy development depends on the results from the risk analysis.

Specifically the decision on retrofit and rehabilitation is one the often requested end product of

the uncertainty analysis. [12]

3.4.1. Randomness vs. Uncertainty:

In order to proceed with the uncertainty modeling it is important in earthquake risk analysis to

distinguish between uncertainty that reflects the inherent randomness in the result(aleatory

uncertainty) and uncertainty due to lack of knowledge (epistemic uncertainty). [13]

Epistemic thus represents those that are originating from inherent randomness and modeling

errors whereas Aleatory uncertainty is due to the randomness of the model parameters.

The randomness associated with the prediction of a parameter from a specific model used in the

analysis is accounted by the Aleatory variability

14
Epistemic uncertainty arises from the modeling assumptions, unknown properties and parameters

and extrapolation of data beyond their observed range.

Examples:

In earthquake engineering, examples of aleatory random variables are the earthquake's

occurrence time, magnitude and distance of the next severe earthquake. And the unsurety in the

assumed distribution properties or the assumed model behavior like the attenuation equation

development are examples of epistemic uncertainty.

Similar observation can be in case of capacity as well. The variability in yield strength of steel,

member stiffness properties, damping ratio are aleatory; and the errors in the structural analysis

models used in describing these parameters like the Incremental dynamic analysis are epistemic.

As our knowledge improves with observations and new data better models can be developed

with more certainty. For simplicity, aleatory uncertainty is referred to as "randomness" and

epistemic uncertainty as "uncertainty."

3.4.2 Various sources of uncertainties in demand estimation:

1) Occurrence Time of earthquake: A simple Bernoulli sequence, is often used to model the

random occurrence time of the earthquake and its limiting form becomes the Poisson

process. The annual probability of occurrence, p, for the Bernoulli sequence or mean

occurrence rate, for the Poisson process is the only parameter that needs to be estimated. [14]

2) Epicenter Location of earthquake: The randomness in the exact location of future

earthquake epicenter is accounted through random spatial distribution models like a line and

areal source models. Here the epicenter is assumed to follow certain distribution on the line

or area under consideration. [15]

15
3) Magnitude: The Gutenberg-Richter equation is often used to estimate the magnitude of the

earthquake, It is a linear equation between logarithmic frequency N and magnitude M over a

certain range [15,16]

LogN = a - b Log M for mL < M < mU (5)

Uncertainty can creep through while estimating the parameters a, b, mU, and mL.

4) Path and Site: The seismic waves are attenuated or amplified while propagating from the

source through the soil to the site, and many factors contribute to the uncertainty in the

modeling of the inherent attenuation and amplification processes. The most important

independent variables considered for modeling the attenuation equations are the magnitude

(M), distance (R), and site soil classification (S).

The attenuation equation A(M, R, S) generally describes the central value and the scatter is

modeled by a random variable (lognormal distribution). Therefore given M, R, and S, the

intensity measure, e.g. spectral acceleration Sa at the site is approximately a lognormal

random variable with expected (mean) value

E[log Sa ] described by the attenuation equation; i.e. [17]

E[log Sa(M,R,S) ] = A(M,R,S) (6)

Note that the above equation describes the randomness in attenuation alone when M, R, and S

are known. M, R, and S are also random variables, which would influence the demand on the

structural system.

16
3.4.3 Bayesian Logic Tree method for modeling uncertainty evaluation:

The formulation of a logic tree is one of the used modeling tools for accounting the epistemic

uncertainty. The logic tree includes assigning weights to each of the multiple alternative

hypotheses in a single model. Thus each probable analytical model suitable for the situation is

assigned a probability of success value and thus the uncertainty arising through knowledge

uncertainty can be included. A logic tree consists of a series of nodes at which models are

specified, and a number of branches that represent the different models under consideration.

Each alternative model (or node) is assigned a weighting factor representing the likelihood of

that model being correct. The sum of probabilities from all branches connected to a node must be

equal to 1.0. A sample flowchart depicting this procedure can be found in Figure 2. A seismic

hazard analysis (deterministic or probabilistic) is carried out for each combination of models

associated with a terminal branch. The result of each analysis is weighted by the relative

likelihood of its particular combination of branches. The final result is the sum of the weighted

individual results.

Fig 2: Sample flowchart depicting the logic tree method [18]

17
3.4.4 Capacity Uncertainties:

1) MATERIAL PROPERTIES: The test data up to 1980 found in the report by Ellingwood et
al. (1980) shows the material property variability that is persistent. For example COV of
compressive and tensile strength of concrete is found to be around 18% and that of the yield
strength of steel reinforcement is around 10%. [19]

2) Member Capacity under Monotonic Load: Besides the material property randomness there
is an additional randomness in the accounting of the member dimensions and hence resulting
in the member capacity uncertainty..

3) System Capacity against Damage: Often there is no standard rule to define the system
capacity and it is often described in terms of the drift capacity for different performance
levels, such as the median drift capacity and its cov. The commonly accepted distribution that
best fits the capacity variable is lognormal distribution.

4. Methodology

In performance evaluation and reliability analysis of structures the system performance is often
described in terms of demand and capacity. The demand can be the force (shear, bending
moment etc.) or the response (displacement, acceleration, drift) caused by the ground excitation.
And the capacity of the system is the maximum forces or response that the system can withstand
without failure. Uncertainty modelling includes modelling of two different types: (1) record-to-
record variability and (2) Analytical uncertainty. Results shows that, a simple nonlinear static
analyses increases this uncertainty even further in the estimation of structural demand. Thus the
current design codes are based on the use of nonlinear response history analyses for the
performance-based design procedures like the ATC-58 PEER.

Also a various sources of uncertanities arises in the input parameters to the OPENSEES model
like in:

18
 Mass

 Live loading

 Material properties (Yield Strength, Elastic Modulus)

 Damping value

4.1 Demand uncertainties:


Randomness:
The uncertainty that the user can face in estimation of Demand comprises of various factors.
First is the accounting of the randomness in Demand that is inherent in its prediction . This arises
because of change in demand value because of change in ground motion that the structure is
subjected to (Record to record variability). So this can be stated as the randomness arising due to
variation in the actual ground motion accelerogram (Represented as ) This uncertainty is
accounted in the form of factor in SAC project procedure.

Uncertainty:
Another major source of uncertainty that creeps into the procedure is the analysis uncertainty that
arises because of the uncertainty associated with specific analytical modeling tools and
techniques used to estimate the demand levels. This is represented by . The is nominally
composed of four parts as follows: associated with uncertainties related to the extent that
the nonlinear time history analysis procedure will represent the actual behavior;
associated with uncertainty in estimating the damping value of the structure; associated
with uncertainty in live load; associated with uncertainty in material properties [9].
The density function (PDF) properties of these quantities can be found in Table 1.

The is assumed to be 0.15, 0.20, and 0.25 for 3-, 9-, and 20-story buildings, respectively,
based on judgment and on understanding [9].

Final that represents the final uncertainty value in demand estimation can be obtained by:

√ where,

√ (7)

19
4.2 Capacity Uncertainties:
Structural capacity is the maximum IDR value that a member or a system can withstand without
failure without reaching a prescribed limit state. Capacity uncertainty also comprises of two
types namely: ‖Randomness‖ and ―Uncertainty‖ inherent in prediction of structural capacity
( . This uncertainty is accounted in the form of factor in SAC project procedure.

Variations due to future factors that cannot be predicted are termed randomness. The principle
portion of randomness in global capacity is due to the variation in the earthquake accelerograms.
So comes from randomness estimation of IDA results. The capacity is dependent on the
material properties, under consideration, and methods and models used in describing the
capacity. Thus in this work the capacity uncertainty of materials is also considered [9].

The member and system capacity depend directly on the material strength which is inherently
random. This randomness can be modelled by taking median data for different parameters under
consideration. The structure is modelled with the median parameter values and the uncertainty in
the response is taken into account. This variation is termed uncertainty.

Final that represents the final uncertainty value in capacity estimation can be obtained by:

√ Where,

√ (8)

The global drift capacity of a building is determined using the incremental dynamic analysis
(IDA) procedure. This is based on the use of nonlinear time history (NTH) analysis.

20
Table 1: Statistical characteristics of the input random variables for model building [20].
Parameter Median values used COV Distribution

Damping 5% 0.4 lognormal

Mass 250kg 0.1 lognormal

Yield strength 248Mpa 0.05 lognormal

Elastic modulus 200Gpa 0.1 lognormal

Live load 2.5kN 0.2 lognormal

4.3 Mathematical Formulation:


The use of square root sum of squares (SRSS) method for finding the equivalent dispersion value
by combining the individual dispersion values arising from various random and uncertain
sources as mentioned earlier is only valid when the underlying values are independent. Hence
one needs to establish the independence before proceeding with the determination of final
dispersion value.

Evaluation method:

For a given lognormal random variable ‗X‘ is defined as the standard deviation of logarithmic
of X obtained from the given data points.

( ( (9)

This value is obtained from data sampling by using the following equation:

21
Final dispersion value evaluation:

Mathematical formulation behind the method employed for evaluating the final dispersion value
is explained here. In this work the interstorey drift ratio value is considered as the engineering
parameter of study whose values are recorded to evaluate the demand and capacity parameters.
And in calculating the demand dispersion value two factors are considered where represents
the record to record randomness and represents the uncertainty arising from modeling
approximations and other factors as mentioned in sections 3.4.
(10)

Where,

= IDR value recorded from each run subjected to different material and earthquake properties;
( (

= Median of all the IDR values;

= Record to record variability factor; ( (

= Analysis variability factor; ( ) (

4.4 Incremental Dynamic Analysis (IDA)


The IDA is a new procedure suggested by Cornell in 2002 which can be seen as an of the simple
nonlinear response history analysis in corollary with stating the nonlinear static pushover
analysis as an extension for static analysis. IDA is done by conducting a series of nonlinear lime-
history analyses using the same earthquake record by increasing the intensity levels..

4.5 Computing the drift capacity


The drift capacity of a structure depends on the earthquake record if dynamic behaviour is taken
into account. Non-linear dynamic analysis is run on all the earthquake records and the
corresponding drift capacity value is computed as given in Yun et al. 2002 [9]

22
Fig 3: Computation of drift capacity of structure [21]

In Figure 3, Line 1 represents the elastic slope (Se). Lines 2 and 3 represent the two extreme
cases that often arise during an IDA run. It is often observed that many members start yielding
when the records are scaled beyond a particular level. Line 2 represents this case. The structure is
assumed to have failed as soon as the slope drops below a limiting value (in the range 0.1Se-
0.2Se)—in this paper we have taken the limit conservatively as 0.1Se [11]. Line 3 represents
when structural members do not fail in succession.

5. Model Building and Analysis

5.1 Multi-Storey Frame model:

In order to account for more complex representation of the real structures a typical three bay four

storied frame (Figure 4) is studied in this work. On account of increasing complexity and the

larger time taken by the FEM model to obtain the desired results the multi storey frame analysis

is done by scripting in OPENSEES. It is an object-oriented software and allows one to

create finite element applications for seismic analysis. The dynamic analysis done using the

scripting language in OPENSEES saves a huge amount of time in the order of 1000 times faster

than other FEM applications. Both the geometric as well as material nonlinearity is accounted in

this code as well. The ground type for design is assumed as Type C

23
To accurately model the connection between the column and beam, and in order to be able to

realistically represent its behavior, a combination of rigid links and displacement-based fiber

elements were used. The rigid links were used to model and take into account the geometry of

the members and the connection while only a part of the gusset plate, equivalent to a length of

two times the thickness of the gusset plate, was modeled. It was shown by Astaneh-Asl et al.

(1985) that modeling only a length equivalent to two times the thickness of the gusset plate

allowed accurate numerical results when compared to experimental data. Figure 4 below shows

typical joints in the models where the black squares represent fully restrained connections and

empty circles represent hinge connections. For each line element representing a beam, column,

brace element or gusset plate, a displacement-based fiber cross-section and uniaxial material was

assigned. The rigid links were modeled using an elastic beam-column element with high area and

modulus of elasticity.

(a) (b)

Fig 4: (a) Modeling technique employed to represent the link between the beam and column (b)

Four Storey 3 Bay Frame structure used for analysis

24
5.2 SAP 2000 Model:

In order to validate the OPENSEES model and its results a corresponding model of the same

frame structure is made in SAP 2000 which can be seen in Figure. Eigenvalue analysis and static

pushover analysis is done using this model and the results are then compared with that of

OPENSEES model.

Fig 5: SAP 2000 model of the 4 storey 3 bay frame structure


5.3. Real time earthquake records:
According to latest guidelines of FEMA 695 [23] a set of 22 real time earthquake records (44

individual components) is proposed to be used for nonlinear dynamic analysis of buildings and

evaluation of the probability of collapse for Maximum Considered Earthquake (MCE) ground

motions. These records meet a number of conflicting objectives like Code (ASCE/SEI 7-05)

Consistent, Very Strong Ground Motions, Structure Type Independent, Site Hazard Independent

An idealized SDOF system with natural frequency corresponding to that of the given frame

structure is used for scaling purpose and Newark‘s algorithm is employed for the evaluation of

spectral acceleration. The desired spectral acceleration values are obtained from the so called

uniform hazard spectrum graphs developed for Mumbai city given in Figure 6.

25
(a) (b)
Fig 6: Response spectra for different site conditions at Mumbai with: (a) 2% probability of
exceedance in 50 years (b) 10% probability of exceedance in 50 years. [10]

Justification for scaling:

Scaling of the same earthquake records is used in this work to cover a large range of response in

the inelastic region. But although some authors argue that the inherent frequency content and the

total duration of the earthquakes are still being the same it is found that when spectral

acceleration is used as a measuring parameter the errors from this approximation is small

(Shome et al 1998).

Justification for using other earthquakes:

Also in this work the earthquake records used were not site specific records and they haven‘t

occurred anywhere near the site so one can state that the ground motions may not represent any

possible future events. But since the earthquake records are tried to match to hazard level

through scaling using the uniform hazard spectra obtained for the specific site one can assume

that these ground motions would produce structural response that actually represents those that

of actual site.

26
6. Results

6.1 Eigenvalue Analysis:

A preliminary analysis was first conducted to validate the OPENSEES model. For this the

structural models that were developed are subjected to eigenvalue analysis and nonlinear static

pushover analysis. Table 2 shows the comparison of the first three modal periods (T1 to T3) for

the model from both the OPENSEES and SAP 2000. The value shows that the model accurately

represents the modal behavior of the structure.

Table 2: Time period comparison from eigenvalue analysis

Configuration
T1(s) T2(s) T3(s)
`

OPENSEES Model 1.20 0.61 0.50

SAP Model 1.27 0.65 0.53

(a) (b)

Fig 7: Four Storey 3 Bay Frame structure (a) First Mode of vibration (b) Second mode

27
6.2 Pushover Analysis:

The pushover analysis of the structural modal is also performed so as to verify the failure

behavior and the ductile nature of the structure. This also yielded similar responses well

accepting with the suggested design base shear values as per IS code. Figure 8, 9 shows the

pushover curves (base shear versus roof displacement) obtained for the model. A summary of

the design base shear and the base shears corresponding to the first significant yield point and

ultimate capacity derived from the pushover curves is contained in Table 3, 4. Overstrength

factors, obtained as the ratio of the base shear corresponding to first significant yield to the

design base shear, are also reported in the table.

Table 3: Comparison of Design base shear and the base shear strength from pushover analysis

Design Yield Strength Ultimate Capacity OverStrength

Strength (kN) (kN) (kN) Ratio

OPENSEES MODEL 2050 2560 3000 1.20

The pushover results are then compared with that obtained from SAP 2000 which can be seen in

table 4. The results show a close agreement of the pushover curves obtained from SAP 2000 and

OPENSEES. From the hinge failure model of SAP it can be observed that the failure occurs at

the first storey of the frame agreeing with the eigenvalue result which shows that the first mode

of vibration is the dominant mode.

Table 4: Comparison of base shear strength from pushover analysis of OPENSEES and SAP

Yield Strength Ultimate Capacity OverStrength


(kN) (kN) Ratio
OPENSEES MODEL 2560 3000 1.20
SAP Model 2250 2800 1.25

28
Fig 8: Pushover Curve obtained from SAP 2000 model

3500000

3000000

2500000

2000000
Series1
1500000

1000000

500000

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Fig 9: Pushover curve obtained from OPENSEES

29
Fig 10: Hinge Failure Sequence during pushover analysis in SAP model

6.3 Nonlinear Time history Analysis:

The frame model is now subjected to the scaled earthquake records corresponding to the hazard

level under consideration and the maximum storey drift values are evaluated in each earthquake

record. This involves only the randomness due to record to record variability, so the dispersion

(β) values from other sources of uncertainty is to be invoked in the final lognormal

approximation. In the time history analysis, three main engineering demand parameters, namely

base shear, peak roof drift ratio and peak inter-story drift ratio, are examined. The peak inter-

story drift ratio is defined as the ratio of the maximum inter-story drift over the duration of the

ground motion to the height of the story, and the peak roof drift ratio is obtained as the ratio of

the maximum top displacement to the overall height of the structure. Figure 11 shows typical

time history plots for inter-story drift ratio (IDR) at 1st story level under the effect of San

Fernando Earthquake.

30
First Floor IDR
1
0.8
0.6
0.4
0.2
IDR (%)

0
-0.2 0 2 4 6 8 10 12 14
-0.4
-0.6
-0.8
-1
-1.2
Time (s)

Fixed

Fig 11: Time history plot of first floor IDR when subjected to San Fernando Earthquake

Interstorey Drift Ratio variation:

Figure 12 show typical variation of peak inter-storey drift ratios (IDR) along the height of the

frame for three ground motion records respectively. For all the ground motions the maximum

peak inter-story drift ratios over the height of the structure were less than the 2% design limits.

The maximum peak IDR clearly occurred in the first story level further strengthening the

assumption that the first mode contributes the highest in the structural response
5

4
Story number

3
Series2
2
Series3
1 Series1

0
0 0.5 1 1.5
IDR(%)

Fig 12: Typical variation of peak IDR (%) over the height of frame for 3 different earthquakes
31
6.4 Evaluation of Demand distribution:

To evaluate the demand distribution one needs to model the hazard level owing to the inherent

randomness. The 4-storey 3-bay frame model is then subjected to the scaled earthquake records

accordingly in each of these methods and the final log-normal demand distributions so obtained

are compared.

6.4.1 Record to record variability estimation: The step by step procedure adopted in this

method for predicting the demand distribution is as follows:

1. Generation of the desired spectral acceleration (Sa) corresponding to a hazard level

randomly. This is accomplished with the help of the inverse transformation technique on the

hazard curve for Mumbai city. A uniform random number ranging between 0 to 1 is generated

and the corresponding Sa from the hazard curve is used as the scaling factor.

2. Selecting an earthquake record from the 44 selected earthquake records and finding IDRmax.

This above procedure is repeated for 132 iterations and the final distribution is evaluated.

Fig 13: Interactive figure depicting the inverse transformation technique used to randomly
generate the Spectral Acceleration values (Sa)
The approximate lognormal distribution obtained from this method is found to be with mean

0.0023 and variance of 2.4e-06 resulting in a record to record dispersion value of 0.63.

32
6.4.2 Analytical variability estimation:

The contribution of yield strength, elastic modulus, mass and damping values variability in the

final dispersion value needs to be evaluated and the step by step method adopted for this is as

follows:

1) Consider any one of the earthquake record from the set of 44 of FEMA. (R = R1)

2) Then generate yield samples randomly from the prescribed yield strength distribution

3) Estimate the median value of y

4) Evaluate y (yield factor) value accordingly corresponding to each randomly generated

yield strength value.

5) Estimate the βy from the obtained data set of yield factor values using Equation

6) Repeat the same procedure for different earthquakes and for estimating the β values

resulting from other sources of uncertainties under consideration.

7) Plot the final distribution as shown in Figure 14.

The final evaluated dispersion values of β are tabulated in Table 5 and the final demand

distribution is obtained by combining the final analytical βa with record to record βRD as

shown in Figure 15.

33
Fig 14: Distribution of  factor values corresponding to each uncertainty.

Fig 15: Final demand distribution plots considering only record variability and total uncertainty

34
Table 5: Dispersion values obtained for different quantities under consideration
Uncertain Mean value of  Dispersion value
PDF Distribution
Quantity ratio (β)

Mass 0.97 0.081 Lognormal

Young Modulus 1.02 0.046 Lognormal

Yield Strength 1.04 0.0672 Lognormal

Vertical Load 1.01 0.0217 Lognormal

Damping 1.01 0.0426 Lognormal

6.4.3 Limit State Capacity and Reliability Calculation for deterministic capacity:

Thus from the FEMA 273 guidelines the probability of failure is calculated which is tabulated in
the following table.

Table 6: Probability of failure for different performance levels


Performance Probability of
Reliability Limit IDR value
Level Failure

IO 0.0323 0.9677 0.70%

LS 3.13E-05 0.999969 2.50%

CP 2.20E-07 0.99999978 5%

6.5 Capacity Distribution:

As opposed to the usage of determinstic capacity limit values based on codal guidelines we can

also evaluate the capacity distribution for a given structure and then use monte carlo simulations

to obtain the final probability of failure. For the estimation of the so called capacity distribution

both the uncertainity and randomness due to record to record variability should be taken into

account. Incremental Dynamic Analysis procedure is employed for the evaluation of drift

capacity of a given structure. Sample IDA curves obtained from the above mentioned procedure

35
is shown in Figure 16. The capacity values are obtained from these set of IDA curves based on

the selected criteria as mentioned earlier. The so obtained dispersion value from this method

represents the record to record variability (βRC) and the dispersion caused due to the material

properties uncertainty (βUC) is added to evaluate the final βC

3
"first-mode" spectral acceleration Sa

2.5

2
(T1, 5%) (g)

1.5

0.5

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Maximum Inter Storey Drift Ratio

Fig 16: Sample IDA Curves of the multi storey frame when subjected to different earthquake
excitations depicting mostly strain hardening cases.

Fig 17: Final Capacity distribution of Multi Storey frame with and without other uncertainties

36
6.6 Correlation coefficient evaluation:

Before proceeding with the monte carlo simulation to evaluate the final probability of failure
with the help of developed demand and capacity distributions we need to evaluate the correlation
coefficient between these parameters since both of these are dependent on similar input
properties and there is greater chance that they are not independent. The method employed to
evaluate this coefficient is as follows:

1. Fix an earthquake which is to used from the suite of 44 earthquakes.

2. Evaluate the demand value (IDRmax) by following Method 1 as described before.

3. Evaluate the capacity value by Incremental Dynamic Analysis.

4. Repeat the above procedure for all the 44 earthquakes.

The correlation coefficient can be evaluated by the following equation and is found to be equal to
0.28.
i n

 ( D  D)*(C  C )
i i
rxy  i 1

(n  1) sx s y
(11)

7. Final Reliability evaluation:

7.1 Analytical Method: By using the evaluated distribution properties of both demand and
capacity variables we can obtain the probability of failure value analytically. The value is found
to be equal to 10*10-5.

( ( where,

( ( ; ( (

And the probability of failure is given by:

( ( (

Here Z is again a lognormal variable whose mean and standard deviation are given by:

[ ( ] ; ( √

37
7.2 Simulation Method: Monte Carlo simulation technique is employed for the reliability
evaluation. The demand and capacity distributions developed along with the correlation
coefficient value calculated were used to generate 10,000,000(n0) sample points. And it is
observed that about 890(n) data points resulted in failure. Thus the probability of failure is
evaluated to be 8.9E-5.

n
P ( D  C  0) 
n0 (12)

8. Conclusion

In this work the application of performance based design implied in FEMA/SAC project codes is
studied as an improvement to the current earthquake resistant design code. Different advantages
of Performance based design over the Indian code are identified and then studied. The major
problem faced in the performance based design ie uncertainty of various input variables for
design like input ground motion, hazard level, material properties and modeling techniques etc is
incorporated and the reliability level is estimated accordingly. It is found that the dispersion
caused by the various material dependent properties is insignificant when compared to the record
to record variability arising from the uncertainty in the earthquake ground motion levels. Thus
we can conclude that the uncertainties in these quantities can be neglected in future studies. The
same observation is made in the capacity distribution evaluation as well. Based on the obtained
distribution properties of demand and capacity parameters the reliability index value is found to
be around 4. A reliability index in this range falls within the life safety level and couldn‘t prevent
adverse event like collapse in case of major earthquake occurrence. So the design is to be
readjusted accordingly depending on the structural importance and usage. Thus this work
established the methodology for evaluating the structural reliability levels of steel frame
structures in general by covering various aspects of uncertainties.

38
9. References

1. Structural Engineering Association of California (SEAOC), ―Vision 2000, Performance


based Seismic Engineering of Buildings,‖ April 1995.

2. Applied Technology Council (ATC) [1996] Seismic Evaluation and Retrofit of Concrete
Buildings, Report No. ATC-40, Volume 1–2, Redwood City, California.

3. Federal Emergency Management Agency: ―NEHRP Commentary on the Guidelines for the
Seismic Rehabilitation of Buildings‖, FEMA 273 (1997)

4. Federal Emergency Management Agency: ―Recommended seismic design criteria for new
steel moment frame buildings‖, FEMA 350 (2000)

5. Wen, Y. K., et al. "Uncertainty modeling in earthquake engineering." Mid-America


earthquake center project FD-2 report (2003).

6. Bureau of Indian Standards ―Earthquake resistant design and construction: Code of Practice‖,
IS 4326 (1993)

7. Somerville, P.G., Smith, N, Punyamurthula, S, and Sun, J., ―Development of Ground Motion
Time Histories for Phase 2 of the FEMA/SAC Steel Project,‖ Report No. SAC/BD-97/04,
SAC Joint Venture, Sacramento, California, 1997.

8. Cornell, C.A. et al.: "Probabilistic Basis for 2000 SAC Federal Emergency Management
Agency Steel Moment Frame Guidelines", Journal of Structural Engineering 128-4 (2002)

9. Yun, S.Y. et al.: "Seismic Performance Evaluation for Steel Moment Frames", Journal of
Structural Engineering 128-4 (2002)

10. Kanth, Raghu STG, and R. N. Iyengar. "Seismic hazard estimation for Mumbai city."
Current Science 91.11 (2006): 1486-1494.

11. Vamvatsikos, Dimitrios, and C. Allin Cornell. "Incremental dynamic analysis." Earthquake
Engineering & Structural Dynamics 31.3 (2002): 491-514.

12. Wen, Y. K., et al. "Uncertainty modeling in earthquake engineering." Mid-America


earthquake center project FD-2 report (2003).

39
13. Toro, Gabriel R., Norman A. Abrahamson, and John F. Schneider. "Model of strong ground
motions from earthquakes in central and eastern North America: best estimates and
uncertainties." Seismological Research Letters 68.1 (1997): 41-57.

14. Cornell, C. Allin. "Engineering seismic risk analysis." Bulletin of the Seismological Society
of America 58.5 (1968): 1583-1606.

15. Baker, Jack W. "An introduction to probabilistic seismic hazard analysis (PSHA)." White
paper, version 1 (2008): 72.

16. Bolt, B. A. (1970). Causes of earthquakes, in Earthquake Engineering, R. L. Wiegel, Editor,


Prentice Hall, Englewood Cliffs, NJ, 21-45.

17. Boore, D. M. and Joyner, W. B. ― Prediction of Ground Motion in North America‖, ATC-35-
1,1994, pp.6-1to 6-41.

18. Abrahamson, Norman A., and Julian J. Bommer. "Probability and uncertainty in seismic
hazard analysis." Earthquake spectra 21.2 (2005): 603-607.

19. Ellingwood, B. R., Galambos T.V., MacGregor J. G., and Cornell C. A., 1982.
―Development of a probability based load criteria for American National Insitute A58‖, NBS
Special Publication 577, June 1980.

20. Dolsek, Matjaz. "Incremental dynamic analysis with consideration of modeling


uncertainties." Earthquake Engineering & Structural Dynamics 38.6 (2009): 805-825.

21. Patra, P. and Bhattacharya, B.: "An assessment of IS codal provisions for the design of low
rise steel moment frames through incremental dynamic analysis", Bulletin of Earthquake
Engineering 9-2 (2010)

22. Astaneh-Asl, Abolhassan, Subhash C. Goel, and Robert D. Hanson. "Cyclic out-of-plane
buckling of double-angle bracing." Journal of structural Engineering 111.5 (1985)

23. Venture, NEHRP Consultants Joint. "Evaluation of the FEMA P-695 Methodology for
Quantification of Building Seismic Performance Factors." NIST, USA (2010).

24. Shome, N, Cornell, A. C., Bazzurro, P., and Carballo, J. E. ― Earthquakes, records, and
nonlinear responses‖, Earthquake Spectra, Vol. 14, No. 3, August 1998, pp.469-500.

40

You might also like