Você está na página 1de 12

Influence of the ground motion records

variability on the biased nonlinear structural


responses
L. Martins
1
, M. Marques
1
& R. Delgado
1

1
Department of Civil Engineering, Faculty of Engineering of the
University of Porto, Portugal
Abstract
It is a well known fact that one of the major sources of uncertainty in the seismic
design and assessment of structures is the inherent variability of the real seismic
action. The common practice to address this issue is to criteriously select a set of
ground motion records large enough to account for loading variability on the
overall structural behaviour. Design codes and technical literature provide some
guidance on the criterion to select ground motion records and on the minimum
number of accelerograms needed to ensure valuable data, however there is little
published research on the effects of seismic loading variability on structural
response. This work focuses on the scatter of the set of real ground motion
records for analysis, calling on the distribution of the consequent structural
effects, so as to point out a relationship between both action and response
variability. This topic is addressed by analysing the response of a set of
numerical models based on real RC moment frame structures subjected to a
seismic load of increasing intensity, using both nonlinear dynamic and nonlinear
static procedures. The seismic load is introduced by fifty five natural
accelerograms selected to match a target magnitude/distance relationship and
fault mechanisms. From the distribution of the performance points, conclusions
are drawn on the dependency between structural effects and seismic action, and
their evolution with the intensity level, using statistical quantities, such as
average and standard deviation. This study also addresses the influence of the
intensity measure used for scaling the records and the sensitivity of the structural
response to the selection criterion of the ground motion records.
Keywords: Seismic behaviour, Nonlinear analysis, RC moment frames
1 Introduction
The increase on computer power seen in the past few decades enabled
engineering researchers and practitioners to resort to nonlinear analysis as a
fundamental method to assess a structure's performance under seismic loading.
Using nonlinear static or nonlinear dynamic analysis unquestioning one of the
most important sources of uncertainty in the analysis is the inherent variability of
the seismic action [1].
For design and assessment proposes the common practice on the selection of
ground motion records is based on code provisions, that usually recommend a
minimum of three to seven natural or artificial records for dynamic analysis,
matching in a range of periods the design response spectrum [2]
In order to address the load variability in nonlinear analysis it is accepted to use
the average or peak responses of the set of selected ground motion records.
Alternatively, and mostly on research studies, it is also common to use a group
of stochastic ground motion records large enough to predict the overall structural
behaviour and somehow include the variability of its characteristics, such as
duration, peak ground quantities, soil type, magnitude and distance to rupture.
This paper addresses the effects of the seismic action variability in the structural
response of two real RC moment frames subject to continuously increasing
intensity values. Throughout nonlinear incremental dynamic analysis and
pushover methods the relationships of statistical measures to the intensity level
are evaluated and conclusions are drawn from the results.
The effects of the chosen scaling technique, such as peak ground accelerations,
PGA, and spectral accelerations at the fundamental period of the buildings,
Sa(T
1
,), in the evolution of the response variability are also assessed in the
present study.

2 Ground motion records
2.1 Selection criteria
One of the major sources of uncertainties in any structural analysis using ground
motion records from real earthquakes is the inherent variability of the seismic
action.
The common practice to address this issue is to select a group of records large
enough to allow the prediction of the overall structural behaviour and the most
likely failure mechanism. It has been proved that the quality of the structural
performance assessment though nonlinear dynamic analyses is closely related to
the record selection.
Most authors recommend that selecting ground motion records for seismic
design or assessment purposes should, especially, take into consideration the
range of magnitudes and distances to rupture expected for the region of interest.
Bommer and Acevedo [3] argue that, due to the strong bond between the
magnitude of an earthquake and its duration, this criterion should always be
considered. At this point one must state that earthquake duration plays a major
role in a structure's performance due to strength degradation effects caused by
cyclic loading [4, 5]. Bommer and Acevedo suggest that the magnitude of the
selected ground motion records should be constrained to the interval defined by
the expected magnitude 0.20M
W
. Despite being one of the most common
selection methods, the selection based on the earthquake magnitude has been
questioned by some authors like Shome et al [6] or Iervolino and Cornell [7].
The first one claims that the best results using nonlinear dynamic analysis are
achieved, not by selecting records to match a target range of magnitudes but
instead determining the expected spectral acceleration at the site from attenuation
laws and probabilistic seismic hazard assessments, and then scaling the records
to that intensity level. On the second paper the authors state they haven't found
significant changes in the response dispersion computed using a set of records
chosen based on the magnitude criterion and another group randomly selected.
One may further refine the technique of records selection by minimizing the
deviation of the response spectra to a target spectrum. Equation (1) provides a
way to analyse the deviation between spectra, being N the number of periods of
vibration in which the comparison is to be made [8].

( )
( )
2
1
target
target 1

=
|
|
.
|

\
|
=
N
i
i
record
i record
rms
PGA
T Sa
PGA
T Sa
N
D
(1)

Iervolino et al [7] proposed and alternative equation (equation (2)) to avoid
resorting to peak acceleration as a normalization measure.

( ) ( )
( )
2
1
target
target 1

=
|
|
.
|

\
|
=
N
i
i
i i record
i
T Sa
T Sa T Sa
N
o
(2)

Last but not least Beyer and Bommer [9] also addressed this subject and
proposed their own equation (equation (3)) that adds to equation (1) a scaling
term , minimizing the deviation of the scaled record to the target spectrum. In
this equation, the match to the target spectrum is evaluated at a range of periods
defined between T
k
and T
j
.
( ) ( ) ( )
2
target
1
1

=

+
=
k
j i
i i record rms
T Sa T Sa
j k
D o
(3)

2.2 Minimum number of records for nonlinear analyses
In order to correctly assess the performance of a structure through nonlinear
dynamic analyses using accelerograms from real earthquakes is necessary to
ensure a minimum number of records. The European standard for seismic design
of structures (Eurocode 8 [10]) allows the engineer to use a lower bound of three
or seven records, whether considering the maxima or average structural response
quantities, respectively. Shome et al [6] suggest the use of a minimum of seven
records in nonlinear dynamic analyses, on the other hand Boomer and Acevedo
[3] recommend at least ten different records to ensure valuable data from the
analysis.
Studies from Hancock et al [11] have shown that the standard deviation of the
structural response decreases when records are scaled based on the spectral
acceleration, in comparison with responses obtained from real unscaled ground
motion records.

3 Methodology
3.1 Numerical models
This study was based on two real reinforced concrete moment frames selected to
represent two extremes in the seismic structural behaviour, both structures are
described in detail in Marques [12]. The first one is a three-storey building built
in Europe during the 70's and designed mainly to withstand permanent loads, the
second one is a twelve-storey building with high ductility. The first structure is
an asymmetrical frame with three spans of 4.05m, 2.00m and 3.50m length. This
frame has rectangular columns with cross-section variable in height ranging from
0.80x0.20 m
2
to 0.40x0.20m
2
and rectangular beams with 0.20x0.20m
2
. The
second structure has six alignments of columns evenly spaced in spans of 4.0m
each. The frame is constructed using 0.70x0.70m
2
columns and T-shaped beams
with 0.65m in height with a web and flange thickness of 0.35m and 0.14m
respectively. Figure 1 presents a representation of both moment frames
Both finite element models were constructed using beam finite elements with
distributed plasticity and five integration points each.

Figure 1: Analysed structures
[12].

3.2 Selected ground motion records
The ground motion records on the basis of this study were selected to match the
expected range of magnitudes, distances to rupture and fault mechanisms,
compatible with the most recent Hazard models proposed to the Portuguese
mainland [13, 14]. The resultant set of records contain fifty five ground motion
records retrieved from the PEER Strong Motion Database [15] with PGA
ranging from 0.009g to 0.822g with an average of 0.303g and standard deviation
of 0.223g. Figure 2 presents in the top plot the fifty five individual response
spectra obtained from the acceleration records and on the bottom the average and
median spectra are shown.

Figure 2: Response spectra =5%.

3.3 Methods of analysis
In this study both nonlinear dynamic and nonlinear static analyses were applied
to predict the overall structural behaviour of the RC moment frames under
seismic loading of increasing intensity. The intensity measure levels were fixed
based on the spectral acceleration to the structure's fundamental period.
Therefore, each individual spectrum was scaled to match a reference spectral
acceleration at T
1
. All the analyses were performed considering a minimum and
maximum intensity levels of 0.1g and 2.0g, respectively, for both structures.
Newmarks numerical integration method was chosen to solve the system of
equations of motion in the nonlinear time-history analysis, considering the
parameters =1/2 and =1/4 lead to unconditional stability and second order
precision. To improve the numerical stability and better reproduce the energy
dissipation in the elastic range, a 2% tangent stiffness proportional damping was
considered in the analysis. Likewise, to minimize the possibility of finding
unwanted damping forces the analysis algorithm was developed in such a way
that the proportionality term has to be updated at every time step [16].
In what nonlinear analysis is concerned, first of all one must refer that only
conventional pushover algorithms were used. Thus, capacity curves were
determined based on three different load patterns; the first one, even along the
building's height; the second, proportional to the first mode of vibration; and the
third, with a triangular shape. For this study the N2 method was the Nonlinear
0 0.5 1 1.5 2 2.5 3 3.5 4
0
0.5
1
1.5
2
2.5
3
T [s]
S
a

[
g
]


0 0.5 1 1.5 2 2.5 3 3.5 4
0
0.5
1
1.5
T [s]
S
a

[
g
]


Average
Average+Standard Deviation
Average-Standard Deviation
Median
Response Spectra
Static Procedure (NSP) chosen to compute the performance points of the
structures. Originally developed by Fajfar and Gaperi [17] the N2 method has
been implemented in EC8 as the reference NSP for performance assessment,
when using pushover analysis.

4 Results and general discussion
4.1 Scaling method
The intent to analyse the evolution of the structures response to a seismic input
of increase level of intensity lead to the need of scaling the records. In this study,
was chosen a scaling technique to match the same spectral acceleration of the
response spectra at the fundamental period of the structure.
Even though all spectral accelerations are the same it is still possible to get
variability in the structural response due to the nonlinearity effects. As the
structure is loaded the changes on its natural period of vibration make that the
spectral acceleration used to calculate the performance point in pushover
analyses vary from the one used to define the intensity level. This effect is
particularly noticed when using response spectra form natural records due to
their irregular shape. It is possible to illustrate the aforementioned issue when
focusing on the equivalent SDOF system for the computation of the performance
point in pushover analyses. It is known that the equivalent system derives from
the linearization of the capacity curve of the real structure, this fact itself makes
the natural period of vibration different from the one of the real structures and it
can be even more different as the structure is loaded beyond yielding. As
illustrated in Figure 3 this can lead to significant changes in the spectral
acceleration taken into the calculations.

Figure 3: Scaling method.
4.2 Structural response
The structural performance of both numerical models was characterised through
their global response, drawn in terms of capacity curves, presented in Figure 4,
using conventional pushover algorithms and the three load profiles previously
mentioned. In this figure the top plot refers to the three-storey frame while the
bottom one was drawn using the twelve-storey structure. In figure 4 is visible
that using a uniform load pattern lead to a maximum base shear 15% higher than
the one calculated using the other two load profiles. The modal and triangular
load patterns produced almost the same lateral force vs displacement
relationship.

Figure 4: Capacity curves.

From the capacity curves and applying the N2 method one can determine the top
displacement at the performance point.
Taking the average and the standard deviation from the top displacement
calculated using both nonlinear dynamic and static analysis Figure 5 was plotted.
The top two charts in this figure are related to the three-storey building while the
remaining charts are from the second structure.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
0
50
100
150
200
250
300
Top displacement [m]
B
a
s
e

s
h
e
a
r

[
k
N
]


0 0.2 0.4 0.6 0.8 1 1.2
0
2000
4000
6000
8000
Top displacement [m]
B
a
s
e

s
h
e
a
r

[
k
N
]


Uniform load pattern
Modal load pattern
Triangular load pattern
Uniform load pattern
Modal load pattern
Triangular load pattern


Figure 5: Average top
displacement and
standard deviation
results

From Figure 5 one can state that the average top displacement calculated using
pushover analysis is in agreement with the one defined through nonlinear
dynamic analysis. This fact is testimony of the ability of pushover methods to
predict structural behaviour, as an alternative to more time consuming methods
like nonlinear dynamic analysis. In this figure it can be seen that for the lower
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Sa [g]
A
v
e
r
a
g
e

t
o
p

d
i
s
p
l
a
c
e
m
e
n
t

[
m
]


Dynamic analysis
Pushover analysis - uniform load pattern
Pushover analysis - modal load pattern
Pushover analysis - triangular load pattern
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
0
0.01
0.02
0.03
0.04
0.05
0.06
Sa [g]
S
t
a
n
d
a
r
d

d
e
v
i
a
t
i
o
n



Dynamic analysis
Pushover analysis - uniform load pattern
Pushover analysis - modal load pattern
Pushover analysis - triangular load pattern
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
0
0.1
0.2
0.3
0.4
0.5
Sa [g]
A
v
e
r
a
g
e

t
o
p

d
i
s
p
l
a
c
e
m
e
n
t

[
g
]


Dynamic analysis
Pushover analysis - uniform load pattern
Pushover analysis - modal load pattern
Pushover analysis - triangular load pattern
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
0
0.05
0.1
0.15
0.2
Sa [g]
S
t
a
n
d
a
r
d

d
e
v
i
a
t
i
o
n


Dynamic analysis
Pushover analysis - uniform load pattern
Pushover analysis - modal load pattern
Pushover analysis - triangular load pattern
intensity levels the displacements computed using nonlinear dynamic procedures
are closer to the ones determined using the modal and triangular load patterns,
however as the intensity of the load is increased the average top displacement
from the dynamic methods approximate to the one predicted by the uniform load
pattern, particularly for the three-storey frame. This behaviour is related to the
development of a typical soft-storey mechanism at the base of the building as it
is increasingly loaded, imposing a change on the distribution of the inertial
forces along the buildings height.
With regard to the evolution of standard deviation with the load intensity level it
is shown in Figure 5 that as the intensity level is increased there is a similar
growth on the standard deviation of the structural response, until a certain level
in which the standard deviation practically stabilizes. One may advance that this
type of behaviour is due to the elongation of the structure's natural period of
vibration. As the load is incremented the structure's progressive loss of stiffness
implies that the relative displacement of the structure approximates to the ground
displacement leading to a loss in the dispersion of the structural response.

4.3 Sensitivity of structural response to the intensity measure
Despite the fact that spectral acceleration (Sa) is frequently used as an intensity
measure for the seismic performance assessment of structures, a new set of
pushover analyses was performed using the peak ground acceleration (PGA) as
the intensity measure.
Once computed the performance points using the N2 nonlinear static procedure,
the average top displacement and standard deviation were determined for each
intensity level.
With the intention to establish a comparison between the results from both
scaling methods it is needed to harmonise the scales of the intensity measures,
converting each intensity level measured in PGA to its correspondent level
measured in Sa.
The approach followed in this study was to take the median spectral acceleration
in the structure's natural period of vibration for all the considered intensity levels.
From these studies it was identified a PGA value of 1.0g for the three-storey
frame and 1.5g for the twelve-storey frame, that corresponds to a maximum Sa
of about 2.0g.In Figure 6 are plotted together the results taken from scaling to
PGA and Sa. The initial two plots refer to the three-storey structure while the
other two to the twelve-storey structure.


Figure 6: Comparison between
scaling methods.

From analysing Figure 6 it is noticeable that no significant differences are found
in the computed average top displacement when the record scaling is based on
peak ground accelerations or spectral accelerations; however this figure also
shows that the standard deviation is more sensitive to the scaling method. It is
clearly visible in the figure that for the twelve-storey structure the scaling based
on PGA has produced higher standard deviation for all the considered intensity
levels, while the three-storey frame only exhibit this behaviour on the upper
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Sa [g]
A
v
e
r
a
g
e

t
o
p

d
i
s
p
l
a
c
e
m
e
n
t

[
m
]


Uniform load pattern-Sa
Modal load pattern-Sa
Triangular load pattern-Sa
Uniform load pattern-PGA
Modal load pattern-PGA
Triangular load pattern-PGA
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Sa [g]
S
t
a
n
d
a
r
d

d
e
v
i
a
t
i
o
n



Uniform load pattern-Sa
Modal load pattern-Sa
Triangular load pattern-Sa
Uniform load pattern-PGA
Modal load pattern-PGA
Triangular load pattern-PGA
0 0.5 1 1.5 2 2.5
0
0.1
0.2
0.3
0.4
0.5
Sa [g]
A
v
e
r
a
g
e

t
o
p

d
i
s
p
l
a
c
e
m
e
n
t

[
m
]


Uniform load pattern-Sa
Modal load pattern-Sa
Triangular load pattern-Sa
Uniform load pattern-PGA
Modal load pattern-PGA
Triangular load pattern-PGA
0 0.5 1 1.5 2 2.5
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Sa [g]
S
t
a
n
d
a
r
d

d
e
v
i
a
t
i
o
n



Uniform load pattern-Sa
Modal load pattern-Sa
Triangular load pattern-Sa
Uniform load pattern-PGA
Modal load pattern-PGA
Triangular load pattern-PGA
intensity levels. These findings are similar to the ones obtained by Marques [12],
in which the author also noticed that using PGA as the intensity measure seemed
to lead into a higher variability in the structural response.

5 Final remarks
The present paper addressed the effects of the variability of seismic load in the
response of two reinforced concrete buildings. The subject was introduced by
referencing past research dedicated to this issue and presenting the most common
ground motion record selection criteria to perform nonlinear analyses.
Taken the presented selection criteria and the seismologic characteristics of
Portuguese mainland as reference, a set of fifty five accelerograms were chosen
from PEER-Strong motion database.
Using the selected ground motion records the structural performance of two
reinforced concrete moment frames was assessed by the means of both nonlinear
dynamic and static analyses.
The studies made clear that the variation of the standard deviation of the
structural response has two different stages. For the lower intensity levels the
standard deviation increases with the intensity, however after a certain point the
structure's loss of stiffness implies that its maxima relative displacement gets
closer to the ground displacements, leading to a decrease in the variation of
standard deviation with the increase of the intensity of the seismic load.
When comparing the computed average top displacement and the standard
deviation using different Intensity Measures it has been shown that only the
standard deviation seems to be affected by the scaling method. For the twelve-
storey frame differences of over 100% in the standard deviation were recorded
being the highest values computed using the peak ground acceleration as
intensity measure. In the three-storey frame this behaviour only appears for the
upper intensity levels and the maximum differences were smaller than the ones
registered on the taller building.

References
[1] Kwon, O.-S. & Elnashai, A., The effect of material and ground motion
uncertainty on the seismic vulnerability curves of RC structure. Engineering
Structures, 2006. 28(2): p. 289-303.
[2] Hachem, M.M., et al. An international comparison of ground motion
selection criteria for seismic design. in Codes in structural engineering -
Developments and needs for international practice. 2010. Dubrovnik,
Croatia.
[3] Bommer, J.J. & Acevedo, A.B., The use of real eathquake accelerograms as
input to dynamic analysis. Journal of Earthquake Engineering, 2004.
8(sup001): p. 43-91.
[4] Kramer, S.L., Geotechnical earthquake engineering. Prentice-Hall
international series in civil engineering and engineering mechanics. 1996,
Upper Saddle River, N.J.: Prentice Hall. xviii, 653 p.
[5] Chung, Y.S., Meyer, C. & Shinozuka, M., Seismic Damage Assessment of
Reinforced Concrete, Technical report NCEER-87-0022, Nacional Center
for Earthquake Engineering Research, State University of New York, 1987
[6] Shome, N., et al., Earthquakes, Records, and Nonlinear Responses.
Earthquake Spectra, 1998. 14(3): p. 469-500.
[7] Iervolino, I. & Cornell, C.A., Record selection for nonlinear seismic analysis
of structures. Earthquake Spectra, 2005. 21(3): p. 685-713.
[8] Katsanos, E.I., Sextos, A.G. & Manolis, G.D., Selection of earthquake
ground motion records: A state-of-the-art review from a structural
engineering perspective. Soil Dynamics and Earthquake Engineering, 2010.
30(4): p. 157-169.
[9] Beyer, K. & Bommer, J.J., Selection and scaling of real accelerograms for
bi-directional loading: A review of current practice and code provisions.
Journal of Earthquake Engineering, 2007. 11(SUPPL. 1): p. 13-45.
[10] CEN, Eurocode 8 - Design of structures for earthquake resistance Part1:
General Rules, seismic actions and rules for buildings,EN1998-1,2010,
CEN: Brussels. p. 230
[11] Hancock, J., Bommer, J.J. & Stafford, P.J., Numbers of scaled and matched
accelerograms required for inelastic dynamic analyses. Earthquake
Engineering and Structural Dynamics, 2008. 37(14): p. 1585-1607.
[12] Marques, M., Avaliao probabilistica da segurana ssmica de edifcios,
Tese de Doutoramento, Universidade do Porto, 2011 (in Portuguese)
[13] Borges, J.F., et al., Seismotectonics of Portugal and its adjacent Atlantic
area. Tectonophysics, 2001. 331(4): p. 373-387.
[14] Vilanova, S.P. & Fonseca, J.F.B.D., Probabilistic seismic-hazard assessment
for Portugal. Bulletin of the Seismological Society of America, 2007. 97(5):
p. 1702.
[15] PEER, PEER Strong Motion Database, 2000,
http://peer.berkeley.edu/smcat/index.html
[16] Charney, F., Unintended Consequences of Modeling Damping in Structures.
Journal of Structural Engineering, 2008. 134(4): p. 581-592.
[17] Fajfar, P. & Gaperi, P., The N2 method for the seismic damage analysis
of RC buildings. Earthquake Engineering and Structural Dynamics, 1996.
25(1): p. 31-46.

Você também pode gostar