Sie sind auf Seite 1von 23

Design Displacements for Base

Isolators Considering Bidirectional


Seismic Effects
Arturo Tena-Colunga,a M.EERI, and Miguel ngel Prez-Osorniob

Currently, there is not available in the literature a specific recommendation


to combine peak displacements in orthogonal directions for the design of base
isolators. The 100% + 30% combination rule prescribed by many seismic
codes worldwide to account for orthogonal effects in the design of conventional structures is sometimes used for base-isolated structures. This
combination rule was formerly proposed to account for lateral forces, but it
should be studied whether or not this rule still applies for the design
displacements of isolated systems. This paper presents a parametric study
where it is found that the statistical responses to combine peak displacements
in orthogonal directions are not independent from the effective period for both
linear and bilinear isolators, so a constant combination rule is somewhat
incorrect. DOI: 10.1193/1.2216737
INTRODUCTION
In the seismic design of three-dimensional structures, it is quite common to take into
account the simultaneous action of the horizontal ground motions through what it is
known as the 100% + 30% combination rule. For instance, in a seismic analysis through
the static method or response spectrum method, the effects of both horizontal components of the ground motion are combined taking, for each direction in which the structure is analyzed, 100 percent of the ground component parallel to the main direction of
analysis and 30 percent of the ground component perpendicular to the main direction of
analysis, using all sign combinations that would lead to the safest design envelope for
the resisting element. The 100% + 30% combination rule is widely used and recommended by most seismic design codes worldwide, and it is based on studies conducted
almost 30 years ago Rosenblueth and Contreras 1977, where the 30% was suggested in
order to minimize the errors introduced by the proposed linear approximation when considering the three orthogonal ground components modeled in terms of elastic spectral
accelerations Menun and Der Kiureghian 1998. This 100% + 30% combination rule is
independent of the fundamental period for the structure and its dynamic characteristics
of response elastic or nonlinear, soil conditions rock, firm soil, soft soil, etc., correlation between ground components, the type of fault mechanism that generates the
ground motions, etc. This fact makes the 100% + 30% combination rule very simple and
general, therefore, very attractive to many practicing engineers and even researchers.
a

Departamento de Materiales, Universidad Autnoma Metropolitana, Av. San Pablo # 180, 02200 Mxico, DF,
Mexico
b
Areva, Lago Victoria 74, Col. Granada, 11520, Mxico, DF, Mexico

803

Earthquake Spectra, Volume 22, No. 3, pages 803825, August 2006; 2006, Earthquake Engineering Research Institute

804

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

Recently, more general procedures to account for orthogonal effects 3-D have been
proposed and compared with the 100% + 30% rule Menun and Der Kiureghian 1998,
Lpez et al. 2001. Menun and Der Kiureghian 1998 and Lpez et al. 2001 based
their methods on the pioneering study of Penzien and Watabe 1975 in which, proposing an orthogonal set of principal axes, and using orthogonal transformations identical in
form to those used in the transformation of stress, they found that the major principal
axis is horizontal and directed to the epicenter, the intermediate axis is in the transverse
horizontal direction, and the minor principal axis is vertical. Penzien and Watabe 1975
used 18 acceleration records from six different earthquakes, three in California 1933
Long Beach, 1940 El Centro, and 1952 Taft, and three in Japan 1968 Tokachi-Oki,
1970 Hiddaka-Sankei, and 1974 Izu-Hanto-Oki. Although the procedures presented by
Menun and Der Kiureghian 1998 and Lpez et al. 2001 are very interesting, they are
still based on combinations of elastic response spectrum quantities, mainly spectral accelerations; that is, ground motions are still represented in terms of elastic pseudoacceleration spectra. In fact, Menun and Der Kiureghian 1998 and Lpez et al. 2001
illustrate their methods using a single pseudo-acceleration design spectrum.
Surprisingly, most researchers are not utilizing simultaneous orthogonal ground motions recorded in recent earthquakes to study orthogonal effects. It is worth noting that
nowadays there are larger reliable databases available worldwide than the one used by
Penzien and Watabe 1975. One exception is the study presented by Burton et al.
2002, where they perform a comparison of response values obtained from modal
analyses of bridges using the 100% + 30% and the 100% + 40% combination rules
three-directional and from time-history analyses using frequency-scaled records. They
found that results obtained for the 100% + 30% and the 100% + 40% combinations rules
were conservative with respect to time-history analyses, although they used few records
5 for their study.
Therefore, the validity of the 100% + 30% rule before the state of current knowledge
has not been fully justified using a reasonable database of simultaneous orthogonal
ground motions recorded in recent earthquakes, primarily for more recent technologies
as systems for the control of the seismic response of structures, among them, base isolation.
The seismic design of isolation systems is accomplished mainly in terms of the
maximum displacement that the isolators can sustain for dynamic stability. These peak
design displacements for isolators are more directly related to the ground displacements
rather than to the ground accelerations. Furthermore, it is common to design isolation
systems to behave inelastically through hysteresis of the material rubber, lead, steel,
etc., by friction, damping, combinations, etc.
With respect to structural design with seismic isolation, a specific proposal to combine the maximum displacement for isolation systems that occur in orthogonal directions does not exist. Therefore, the 100% + 30% combination rule is commonly used
with little in-depth consideration. To the authors knowledge, few people have evaluated
whether or not this rule, obtained from the maximum response combinations from
pseudo- acceleration response spectra of a relatively reduced sample, is valid to estimate

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

805

the maximum design displacement that could be experienced in the isolation system in
any given direction Tena-Colunga and Gmez-Sobern 2002. In this regard, some design recommendations, such as FEMA-273 ATC 1997 in section 9.2.4.5.C, basically
endorses the 100% + 30% combination rule to compute the maximum design displacement for the isolation system. Recently, the new AASHTO draft bridge provisions incorporate a 100% + 40% combination rule MCEER/ATC-49, 2003. The 1997 Uniform
Building Code UBC ICBO 1997 establishes that in the design of isolation systems
through dynamic analysis, the combination of the two orthogonal components of the
ground motions in its horizontal direction must exceed 1.3 times the design spectrum.
Almost every base isolation project to date as per UBC code has had to use the time
history method of analysis and the combined response spectra of the two components of
the time histories should be within 10% of 1.3 times the site specific spectra Reviewer
2005. As demonstrated by Naeim and Kelly 1999, the vectorial sum of the 1997 UBC
proposal of scaling 1.3 times the design spectrum and the sum of the 100% + 30% combination rule lead to different results.
It is the opinion of the authors that, with respect to the horizontal ground motions,
the following questions must be answered regarding orthogonal effects for the design of
isolation systems: 1 is the 100% + 30% combination rule still valid and reasonable to
compute design displacements? and 2 can a rational rule for the combination of orthogonal effects be independent of the structural period and the characteristics of the
ground motion?
This paper presents a parametric study where the statistics of the ratio between peak
displacements of base isolators when subjected to bidirectional seismic input 2D with
those obtained for unidirectional seismic input 1D is computed for base-isolated
structures within an effective period range between 1.5 and 3.0 seconds 1.5 s TI
3.0 s. Both linear and bilinear k2 / k1 = 0.10 isolators were considered. In total, 154
pairs of accelerograms recorded in rock or firm soil sites of the Mexican Pacific Coast
during 13 seismic events of Ms 6.4 in the last 20 years were used for the parametric
study. To the authors knowledge, this might be the first study worldwide where a comprehensive database of acceleration records is used to study orthogonal effects, and the
first to directly consider the nonlinear behavior of the structural system.
From the simulations done in this study, it is found that the statistical responses
mean, mean plus a standard deviation of the 2D / 1D ratio is not independent from the
effective period for both linear and bilinear isolators. Therefore, a constant combination
rule for example, 100% + 30%, although practical, is somewhat incorrect. In addition,
the 2D / 1D ratio also depends on the characteristics of the isolation system, as the obtained results varied from linear and nonlinear systems. The obtained amplification factors are also compared with the worldwide accepted 100% + 30% combination rule, discussing the advantages and disadvantages of this rule for the seismic design of
elastomeric bearings.

806

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

Figure 1. Benchmark model: a plan, b elevation, c 3-D layout, and d isolation system.

BENCHMARK MODEL CONSIDERED


The subject three-story building for the parametric studies benchmark model is depicted in Figure 1. The building is regular in elevation and symmetric with respect to two
main orthogonal axes both in mass and stiffness. The building has four frames in each
direction with a typical bay width of 7 m and a story height of 3 m. Typical RC columns
are of square cross section 50 50 cm and typical rectangular RC beams are 35
75 cm. The building is supported by 16 cylindrical elastomeric bearings one for each
column line below a rigid floor slab, as depicted in Figure 1. Braces of A-36 steel are
located as shown in Figure 1 and are of square box section 25 25 cm with a plate
thickness of 0.8 cm. The total weight of the structure above the isolation system is W
= 991 Ton.
The first six fixed-based periods of vibration for the building model computed with
ETABS are summarized in Table 1. These structural periods plus their corresponding
mode shapes were used in the nonlinear dynamic analysis according to this modeling
option allowed by the 3D-Basis program Nagarajaiah et al. 1991a, b. The mode shapes
obtained in ETABS are purely translational and rotational, as the structural model was
completely symmetric in stiffness and mass. The fundamental fixed-based periods in the
main orthogonal directions are Tx = Ty = 0.187 s. The ratio between the torsional and lateral frequencies for the superstructure is s = 1.20. The plan aspect ratio is L / b = 1.

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

807

Table 1. Periods of vibration for the first six fixed-base mode shapes of
benchmark structure
Mode
1 Translation X
2 Translation Y
3 Rotational

Period s

Mode

Period s

0.187
0.187
0.156

4 Translation X
5 Translation Y
6 Rotational

0.063
0.063
0.054

This model has also been used as symmetric benchmark models in previous studies
dealing with torsional effects for example, Tena-Colunga and Gmez-Sobern 2002,
Tena-Colunga and Zambrana-Rojas 2004.
PROPERTIES OF THE ISOLATION SYSTEMS

In this study, both linear elastic and bilinear cylindrical isolators with a post- to preyielding stiffness of 10% k2 / k1 = 0.10 were designed for the following effective period
range of base-isolated structures TI : 1.5 s TI 3 s. The isolators were designed following some available recommendations of the New Zealand practice Skinner et al.
1993 and the 1997 Uniform Building Code UBC-97, as presented in greater detail
elsewhere Prez-Osornio 2004. Isolators were modeled in the 3D-Basis software accounting for biaxial interaction between the two main orthogonal directions.
The mechanical properties for the bilinear isolation systems were set to comply with
a recommendation of the UBC-97 that establishes that the effective stiffness of the isolation system at the design displacement keff must be greater than one-third of the effective stiffness at 20% of the maximum design displacement keff2, that is, keff
1 / 3keff2, as depicted in Figure 2. For bilinear systems with k2 / k1 = 0.10, it can be dem-

Figure 2. Design envelope curve for bilinear isolators that follow the restrictions of the UBC.

808

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

onstrated that the maximum ductility demand that can be accepted in order to satisfy this
mechanic restriction is 9.0 e.g., Tena-Colunga 1997, 2002; Tena-Colunga and
Gmez-Sobern 2002.
It is worth noting that this mechanical restriction was added late in the development
of the UBC in an attempt to prevent the use of very long period isolation systems and is
very rarely used in design in the United States Reviewer 2005. The real limitations
with elastomeric isolators made by quality manufacturers are the buckling and ultimate
shear strain limits; some isolators are able to achieve 400% shear strain nowadays Reviewer 2005.
According to what is outlined in detail in Prez-Osornio 2004 and briefly discussed
in the following sections regarding the selected ground motions, five yield strength ratios for the isolation system were considered in this study: Vy / W = 0.03, 0.05, 0.08,
0.10, and 0.12, where W = 991 Ton is the total weight for the structure above the isolation system. This was done for the effective period range of interest 1.5 s TI 3 s
taking an effective period increment of 0.1 second; that is, there were 16 different designs for the isolation system in the considered effective period range. Details on the
designs of the isolation system are presented in Prez-Osornio 2004.
The yield levels of 10% and 12% are of academic interest since they satisfy the ductility limits related to the UBC-97 mechanic restriction described above for seismic
zones C and D of the Manual of Civil Structures MOC-93 code Inst. de Invest. Elctricas 1993, as it will be evident in following sections. It is worth noting that there are
no projects to date that use a yield level beyond 7% Reviewer 2005.
SELECTED ACCELERATION DATA BASE
A comprehensive set of accelerograms typical of subduction earthquakes recorded in
firm soil sites or rock during the past two decades in the Mexican Pacific Coast were
used in the present study. These records were obtained from the Strong Motion Mexican
Data Base Sociedad Mexicana 2000, and were filtered and corrected for baseline errors
with the procedure that is described and illustrated in detail in Prez-Osornio 2004. In
addition, the seismic zonation of Mexico proposed by the MOC-93 Inst. de Invest. Elctricas 1993, where the Mexican Republic is divided in four seismic zones Figure 3,
was considered in this study. Some details on the seismic provisions of MOC-93 code
are presented in Tena-Colunga 1999.
GENERAL SELECTION CRITERIA FOR PAIRS OF ACCELERATION RECORDS

Recording stations located in seismic zones B, C, and D of the Mexican Pacific


Coast for firm soils and rock were selected in this study. The selected stations are identified in detail in Prez-Osornio 2004 and the location of some of them is depicted in
Figure 4. The selection of recorded accelerograms was made utilizing the following criteria:
Only acceleration records of Mexican earthquakes with magnitude M 6.4 were
considered. The following 13 Mexican earthquakes were then selected: a 19
September 1985 M = 8.1, b 21 September 1985 M = 7.6, c 30 April 1986

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

809

Figure 3. MOC-93 seismic zone map of Mexico. Courtesy of Servicio Sismolgico Nacional

M = 7.0, d 25 April 1989 M = 6.9, e 24 October 1993 M = 6.6, f 14


March 1994 M = 6.8, g 14 September 1995 M = 6.4, h 9 October 1995
M = 8.0, 21 October 1995 M = 6.5, j 15 July 1996 M = 6.5, k 11 January
1997 M = 6.9, l 15 July 1999 M = 6.5, and m 30 September 1999 M
= 7.5.
Selected records must have peak ground accelerations amax 10 cm/ s2 in both
orthogonal horizontal components.

Figure 4. Some of the recording stations in firm soil and rock selected for this study.

810

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

The integral curve obtained from calculation from the Arias intensity must show
an adequate form for an acceleration record, that is, the curve of such integral
should not be similar to the one which is obtained for a pulse or a noise sign. For
this purpose, each acceleration record was independently evaluated using the
Degtra program Ordaz 2001. These criteria must be simultaneously fulfilled by
the two orthogonal horizontal ground components as well.
As a result, 154 pairs of records 308 individual records in total for the 13 Mexican
earthquakes complied with all the selection criteria described above, as presented in detail in Prez-Osornio 2004.
ADDITIONAL SELECTION AND SCALING CRITERIA OF PAIRS OF
ACCELERATION RECORDS FOR BILINEAR ISOLATORS

It is not important to scale acceleration records to obtain the ratio between peak displacements of linear elastic base isolators when subjected to bidirectional seismic input
2D with those obtained for unidirectional seismic input 1D. However, it is very important to scale acceleration records for inelastic systems, among other considerations,
because the accelerograms should have sufficient energy to cause the isolators to respond in the nonlinear range. Therefore, it was necessary to define scaling criteria for the
acceleration records that would cause the isolation systems in the study to have a reasonable inelastic response, but at the same time would maintain the original intensity
relationship between the two orthogonal horizontal ground components. That is, the
scaling factor to use should be the same for the two components E-W and N-S, for
each pair of accelerograms under study.
Thus the original set of 154 pairs of acceleration records previously selected were
individually scaled in order that the so-called dominant ground component would match
the spectral acceleration for T = 2.24 s of the design spectrum of interest of MOC-93
code zones D-I, C-I, and B-I, as illustrated in Figure 5 and presented and discussed in
greater detail in Prez-Osornio 2004. The design spectra specified by MOC-93 can be
found in Tena-Colunga 1999. The effective period of the isolated structure T = 2.24 s
was arbitrarily selected as is an intermediate period in the period range 1.5 s TI
3.0 s where it is recognized in the literature that base isolation is most appropriate, as
has been done in previous studies Tena-Colunga 2002.
The final selection criterion for each pair of scaled acceleration records was that both
records should simultaneously agree that a the scaled peak ground accelerations would
not surpass amax = 1.2 g, and b the scaling amplification factor would not surpass 100
fsc 100. Both orthogonal records should fulfill this criterion in order to be used.
Minimum, peak, and average scaling amplification factors used in the selected records
were, respectively, 0.74, 66.5, and 20.3 for the zone D-I; 0.54, 96.2, and 18.7 for zone
C-I; and 0.21, 45.3, and 9.9 for zone B-I. This criterion, while arbitrary, was taken in
order to avoid amplifying peak ground accelerations of the Mexican earthquakes to unrealistic values, taken as reference peak ground accelerations recorded in firm soils and

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

811

Figure 5. Scaling procedure for some selected accelerograms to match the spectral acceleration
for the design spectrum of zone D-I of MOC-93 code for T = 2.24 s.

rock worldwide during strong earthquakes. After this selection process, 73 pairs of acceleration records fulfilled the criteria when scaling to the design spectrum of zone D-I,
90 pairs for zone C-I, and 135 pairs for zone B-I.
The mean response spectrum obtained from the SRSS combination of the response
spectrum related to the two orthogonal components of the 73 scaled pairs of time histories that fulfilled the criteria for zone D-I is compared with 1.3 times the design spectrum of zone D-I of MOC-93 code in Figure 6. It can be observed from this figure that
the average SRSS spectrum compares reasonably well with the scaled MOC-93 D-I design spectrum in the period range of interest of this study 1.5 s TI 3 s. For shorter
periods T 1 s the resulting SRSS spectrum considerably surpasses the scaled design
spectrum, whereas for longer periods the SRSS spectrum T 3 s is notably below the
scaled design spectrum.
AMPLIFICATION FACTORS FOR THE DISPLACEMENTS OF BASE
ISOLATORS DUE TO ORTHOGONAL EFFECTS
Dynamic analyses linear and nonlinear were conducted for the different baseisolated models using 3D-BASIS program Nagarajaiah et al. 1991a,b, where the superstructure is modeled to behave elastically. For each pair of acceleration records there
are 16 different models, as an effective period increment of 0.1 seconds was taken for
the base-isolated structures in the effective period range of interest 1.5 s TI 3 s.

812

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

Figure 6. Comparison of the mean average spectrum obtained from the SRSS combination of
the 73 scaled pairs of time histories for zone D-I and 1.3 times the design spectrum of zone D-I
of MOC-93 code.

Also, for each one of the 16 models, three different loading cases were considered. The
models were subjected to the action of 1 the E-W component alone, 2 the N-S component alone, and 3 both horizontal orthogonal components N-S and E-W, E-W and
N-S, Figure 7. Please note that as a consequence of having a completely symmetric system using cylindrical isolators of circular cross section, peak displacements for the isolators for the bidirectional loading conditions depicted in Figures 7c and 7d are exactly
the same, although the specific angle where they occur changes. However, as this study
is only interested in the 2D / 1D ratio and not in the angles where they occur, this observation allowed us to reduce the number of required simulations from 64 to 48, as only
the bidirectional loading depicted in Figure 7c or 7d is needed to compute this ratio.
All told, 48 different analyses were computed for each pair of accelerograms for the
effective period range of interest, giving a total of 7,392 simulations for the linear-elastic
isolators. For the bilinear isolators, for each considered yield strength, 6,480 nonlinear
dynamic simulations were conducted for zone B-I Vy / W = 0.03 and Vy / W = 0.05,
4,320 simulations for zone C-I Vy / W = 0.08 and Vy / W = 0.10, and 3,504 simulations
for zone D-I Vy / W = 0.08, Vy / W = 0.10, and Vy / W = 0.12. Thus a total of 39,504
simulations were conducted for this parametric study.
The results of the analyses were organized by classifying the data files by event.
From the time history results, peak dynamic displacements for the isolators were selected. Due to the symmetry of the benchmark model and to the absence of a rotational

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

813

Figure 7. Benchmark symmetrical system: a unidirectional E-W input, b unidirectional N-S


input, and c and d bidirectional E-W and N-S input.

component of the ground, all 16 isolators Figure 1d experienced the same displacement.
An interesting issue was to define what it is called the dominant component for the
ground motion. Some researchers define the dominant component in terms of the peak
ground acceleration, while others use spectral ordinates, usually in terms of pseudoacceleration response spectra or the Fourier Amplitude Spectra. In this work, it was important to define the dominant component for the effective period range of interest
1.5 s TI 3 s, using a criteria that it should not be based on a single peak response
or a crude average. As discussed in Prez-Osornio 2004, one can assess different options. For example, a possible criterion is to select the smallest amplification ratio
2D / 1D regardless of the direction of the ground component E-W or N-S, for each
given effective period, that is, the dominant ground component can switch direction
E-W to N-S or vice versa among the effective period range of interest. This envelope
criterion, though very easy to implement, would yield to the smallest 2D / 1D amplification ratios. However, it is not conceptually strong, since it does not really define a

814

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

Table 2. Peak displacements for linear-elastic isolators computed with 3DBASIS program for the ground records of CHIL station 15 June 1999
earthquake

dominant ground component. This procedure was not used because the 100% + 30%
combination rule is based on studies where spectral values have been combined in terms
of defining a dominant component.
The criterion used in this study can be illustrated with the help of Table 2. The numbers distinguished in bold italics in columns 5 and 6 of Table 2 depict the smallest
2D / 1D amplification ratios for a given effective period TI in the effective period range
of interest 1.5 s TI 3 s. In order to define the dominant ground component column
7, it was decided to take an indirect criteria where the dominant ground component in
the effective period range of interest was defined in terms of the smallest total sum of all
the 2D / 1D ratios for the E-W component column 5 or N-S component column 6 in
the effective period range of interest. Then, for the example shown in Table 2, the smallest 2D / 1D total sum 20.603 is connected to the N-S component column 6, and it
was therefore decided with this criteria that the N-S component of CHIL station is the
dominant ground component from a displacement perspective in the effective period
range of interest. It is obvious that others could consider different procedures.

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

815

Figure 8. Histogram that summarizes the statistics of dominant periods E-W vs. N-S in the
effective period range of interest for the 154 pairs of records selected for the linear-elastic isolators under study.

As also illustrated with the help of Table 2, it is not always simple to define the
dominant ground component. If one compares the smallest 2D / 1D amplification ratios
of columns 5 and 6 for each effective period in the period range of interest, it can be
observed that for some periods the dominant component is the E-W component seven,
column 5 and for some others the dominant component is the N-S component nine,
column 6, therefore, it is not so clear for this station to define the dominant ground
component as it is for other ground motions. In fact, both columns 5 and 6 could reasonably be identified as the dominant ground component, so it is difficult to take a
simple decision in this regard. It is precisely in cases like this one where the proposed
criterion shows its added value. If in the decision process the smallest sum of the
2D / 1D ratios of columns 5 and 6 is taken into account, we have an additional parameter to make a reasonable decision, as can be observed in Table 2, since the smallest sum
corresponds to column 6 20.603. Then, it was decided that the N-S component column 6 represents the dominant ground component for displacements in the period
range of interest and, therefore, it is identified as such in column 7.

816

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

Figure 9. Histogram that summarizes the statistics of dominant periods E-W vs. N-S in the
effective period range of interest for the 73 pairs of records selected for the bilinear isolators
under study for zone D-I and Vy / W = 0.10.

Figure 8 is a histogram where the statistics of 2D / E-W vs. 2D / N-S ratios for
linear-elastic isolators in the effective period range of interest for the ground motion
records used in this study are grouped in terms of similar statistics, that is, the number
of events where a component E-W or N-S dominates in the period range of interest.
For example, in Table 2, the E-W component dominated in 7 periods and the N-S dominated in 9 periods, so defining the stats in terms of the E-W component vs. the N-S
component E-W vs. N-S, the score is 7-9. These histograms were used to help appreciate how many cases of a clear dominant component were detected for example, scores
of 16-0, 0-16, 15-1, 1-15, 14-2, 2-14, 13-3, and 3-13 with respect to those where this
was difficult to assess 8-8, 9-7, 7-9, 10-6, and 6-10, rather than finding if there were
more dominant E-W components than N-S components. From the histogram depicted in
Figure 8 one can observe that the cases where a component is clearly dominant are practically equal to the number of cases where a dominant component is difficult to assess.
Therefore, the smallest sum of the 2D / 1D ratios criterion used in this study was very
important in the definition of the dominant ground component and, therefore, in the results presented in this paper.
The same procedure was used for defining the dominant ground component for bilinear isolators and, as illustrated in the histograms of Figures 9 and 10, the criterion of
the smallest sum is very useful, since the definition of the dominant ground component
for nonlinear systems depends also on the yield force of the isolation system, as discussed and shown in greater detail in Prez-Osornio 2004.

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

817

Figure 10. Histogram that summarizes the statistics of dominant periods E-W vs. N-S in the
effective period range of interest for the 73 pairs of records selected for the bilinear isolators
under study for zone D-I and Vy / W = 0.08.
STATISTICAL CRITERION USED

Mean values and the mean plus one standard deviation + Sdev for the
2D / 1D amplification ratios of all simulations were considered in this study to evaluate
bidirectional orthogonal effects for the peak displacements of base isolators. This statistical criterion is considered reasonable for many engineering applications.
2D / 1D AMPLIFICATION FACTORS FOR LINEAR-ELASTIC ISOLATORS

As mentioned in previous sections, 154 pairs of acceleration records corresponding


to 13 different seismic events and a total of 7,392 simulations were considered for this
parametric study. For the parametric study, the 2D / 1D ratios for Mexican earthquakes
of M 6.4 were first classified by event and magnitude, so the statistical response was
first computed event by event in order to discern first if there exists a strong correlation
between the amplification factors and the characteristics of each event for example,
magnitude. Based on this extensive study that is shown in Prez-Osornio 2004, it was
found that the mean values of 2D / 1D ratios depend on many factors. Therefore, a
more complex study with a considerably larger database is needed in order to discern if
there is a relationship between 2D / 1D ratios and the magnitude for the earthquake,
something that is beyond the scope of the present study. Also, a larger database is needed
if a dependency on the earthquake source mechanism is to be evaluated subduction,
fault normal, fault parallel, strike-slip fault, etc., and to study the impact of near-fault
acceleration records with velocity pulses recorded in large magnitude earthquakes such
as Northridge 1994, Kobe 1995, Chi-Chi 1999, and Izmit 1999.
Mean values and the mean plus one standard deviation + Sdev for the
2D / 1D amplification factor for linear-elastic isolators were computed for the 154 pairs

818

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

Figure 11. Amplification factors due to orthogonal effects 2D / 1D for linear-elastic isolation systems.

of accelerations records and the results plotted in Figure 11, where they are compared
with the 100% + 30% combination rule. It can be observed from Figure 11 that the mean
values for the 2D / 1D ratios vary in the effective period range of interest, that is, they
are not constant. Therefore, it is clear that this amplification factor depends on the period. Furthermore, it can also be observed that in the period interval between 2.2 and 2.7
seconds, where maximum values are obtained, the amplification curve takes a sinusoidal
form with two peaks, as well as a higher standard deviation. In general, the mean plus
one standard deviation + Sdev surpasses the 100% + 30% combination rule for the entire period range, so if + Sdev is used as the design criteria for orthogonal effects, then
the 100% + 30% rule is unconservative for linear-elastic isolators subjected to typical
ground motions of firm soils and rock of the Mexican Pacific Coast. However, the
100% + 30% combination rule is quite conservative for the mean results. This relation
needs to be evaluated for near-fault strong motions elsewhere in the world, something
that is beyond the scope of the present study.
2D / 1D AMPLIFICATION FACTORS FOR BILINEAR ISOLATORS

The parametric study for each zone of MOC-93 code was organized in the following
way. For zone B-I, nonlinear dynamic analyses were conducted for bilinear isolators
with k2 / k1 = 0.10 and yield forces Vy / W = 0.03 and 0.05; therefore, 12,960 simulations
were needed for this zone. For zone C-I, nonlinear dynamic analyses were conducted for
bilinear isolators with k2 / k1 = 0.10 and yield forces Vy / W = 0.08 and 0.10; therefore,
8,640 simulations were needed for this zone. For zone D-I, nonlinear dynamic analyses
were conducted for bilinear isolators with k2 / k1 = 0.10 and yield forces Vy / W = 0.08,
0.10 and 0.12; therefore, 10,512 simulations were needed for this zone. Therefore, a total
of 32,112 nonlinear dynamic analyses were conducted to study the statistics of 2D / 1D
ratio for bilinear isolators.
Among the processed results for all the considered zones are curves for the mean

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

819

Figure 12. Amplification factors due to orthogonal effects 2D / 1D for MOC-93 seismic
zone B-I for bilinear isolation systems with k2 / k1 = 0.10, Vy / W = 0.03 and 0.05.

values and the mean plus one standard deviation + Sdev for a the amplification
factors due to orthogonal effects 2D / 1D, and b their corresponding displacement
ductility demands = 2D / y, as shown in detail in Prez-Osornio 2004. The results
for the amplification factors due to orthogonal effects 2D / 1D are presented in following sections for zones B-I, C-I, and D-I. However, because of space constraints, the
results obtained for displacement ductility demands = 2D / y are only discussed for
zone D-I.
Results for Zone B-I
Mean values and the mean plus one standard deviation + Sdev for the
2D / 1D amplification factor for bilinear isolators with Vy / W = 0.03 and 0.05 computed
with the 135 selected pairs of accelerations records selected for zone B-I are given in
Figure 12, where the statistical quantities are also compared with the 100% + 30% combination rule. The following observations can be made from Figure 12 regarding the
2D / 1D ratios:
2D / 1D ratios are reasonably constant in the effective period range of interest.
Results for the mean values are below the line that represents the 100% + 30%
combination rule. Curves for the mean plus one standard deviation are also below the 100% + 30% line, except for Vy / W = 0.03 for the period range 1.5 s
TI 1.8 s.
The isolation system has a higher displacement response as the yielding force
Vy / W is reduced; therefore, 2D / 1D amplification factors increase as the isolation system has a higher displacement response. This can be confirmed observing that 2D / 1D ratios for Vy / W = 0.03 are higher than those computed for
Vy / W = 0.05.

820

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

Figure 13. Amplification factors due to orthogonal effects 2D / 1D for MOC-93 seismic
zone C-I for bilinear isolation systems with k2 / k1 = 0.10, Vy / W = 0.08 and 0.10.

Results for Zone C-I


Mean values and the mean plus one standard deviation + Sdev for the
2D / 1D amplification factor for bilinear isolators with Vy / W = 0.08 and 0.10 computed
with the 90 selected pairs of accelerations records selected for zone C-I are given in Figure 13, where these statistical values are also compared with the 100% + 30% combination rule. The results are similar to those obtained for zone B-I regarding 2D / 1D ratios, but it can be seen that smaller values are obtained for zone C-I Figure 13 with
respect to zone B-I Figure 12, particularly for the period range between 1.5 and 1.9
seconds. In fact, the curves for the mean values and the mean plus one standard
deviation + Sdev are both below the line that represents the 100% + 30% combination
rule for the entire period range of interest.
Results for Zone D-I
Mean values and the mean plus one standard deviation + Sdev for the
2D / 1D amplification factor for bilinear isolators with Vy / W = 0.08, 0.10, and 0.12
computed with the 73 selected pairs of accelerations records selected for zone D-I are
given in Figure 14, where these statistics are also compared with the 100% + 30% combination rule. The results are similar to those obtained for zones B-I and C-I regarding
2D / 1D ratios, but the following additional observations can be made:
1. 2D / 1D amplification factors for the considered yield strength ratios
Vy / W = 0.08, 0.10, and 0.12 approach a similar value when TI 2.7 s.
2. The computed 2D / 1D amplification factors for zone D-I are greater than those
obtained for zones B-I and C-I for the period range under consideration.
3. 2D / 1D curves for the mean plus one standard deviation are below the
100% + 30% line in the entire period range of interest only for Vy / W = 0.12,
and above the 100% + 30% line in the entire period range of interest only for
Vy / W = 0.08, and for Vy / W = 0.10, the curve is below this 100% + 30% line for
the period range 1.5 s TI 2.1 s and above this line for the period range
2.2 s TI 3.0 s.

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

821

Figure 14. Amplification factors due to orthogonal effects 2D / 1D for MOC-93 seismic
zone D-I for bilinear isolation systems with k2 / k1 = 0.10, Vy / W = 0.08, 0.10, and 0.12.

Their corresponding displacement ductility demands for bidirectional input are


depicted in Figure 15. It can be observed for the curves associated with the mean response that peak displacement ductility demands for the isolators for both Vy / W
= 0.10 and Vy / W = 0.12 do not exceed the limiting value 9.0 associated with the primary curve defined by the 1997 UBC for the entire effective period range of interest.
However, for Vy / W = 0.08, the curve is above = 9 in the period range 1.5 s TI
1.7 s. For the plots corresponding to the mean plus one standard deviation + Sdev,
peak displacement ductility demands for the isolators do not exceed the limiting value
9.0 for Vy / W = 0.12 for the entire period range of interest, but for Vy / W = 0.10, is
exceeded in the period range 1.5 s TI 1.8 s and for Vy / W = 0.08, is exceeded in
the period range 1.5 s TI 2.3 s.

Figure 15. Displacement ductility demands = 2D / y for bilinear isolation systems with
k2 / k1 = 0.10 and Vy / W = 0.08, 0.10, and 0.12, subjected to bidirectional seismic input, for
MOC-93 seismic zone D-I.

822

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

From the viewpoint of an efficient design of bilinear isolators with k2 / k1 = 0.10 in


terms of an acceptable displacement ductility demand for a stable behavior 9.0, it
can be deduced from Figure 15 that the optimal design curve for seismic zone D-I of
MOC-93 should be closer to a yielding force for the isolation system Vy / W = 0.10 rather
than Vy / W = 0.08 excessive demands for the period range 1.5 s TI 1.7 s, if mean
values are considered.
COMPARISON WITH THE 100% + 30% COMBINATION RULE

As already shown and discussed for linear-elastic and bilinear isolators, the mean
values obtained for the 2D / 1D ratios are not constant, and they depend on the period
and the mechanical characteristics for the isolators.
In general, displacement amplification factors due to orthogonal effects 2D / 1D
ratios corresponding to mean values are always below the 100% + 30% combination rule for the entire period range for both linear-elastic and bilinear isolators. The results for the mean plus one standard deviation + Sdev of the 2D / 1D ratios exceed
the 100% + 30% combination rule for the entire period range for linear-elastic isolators,
and depending on the yield strength Vy / W and the seismic zone under consideration,
they can also surpass the 100% + 30% combination rule for nonlinear isolators.
Although the observed variations on the 2D / 1D ratios can be taken into account
with a simple straight line, the question is if this line should be a constant, as the one
given by the 100% + 30% combination rule, or if it is more appropriate to provide a linear equation for this rule that will depend on the period. In the opinion of the authors, to
provide a linear equation depending on the period will make the design process more
transparent to the practicing engineers, at minimum additional effort.
CONCLUDING REMARKS
A parametric study devoted to study the statistical response of displacement amplifications factors due to orthogonal effects 2D / 1D ratios for linear-elastic and bilinear
isolation systems was presented. The study considered acceleration records for 13 subduction earthquakes M 6.4 of the Mexican Pacific Coast recorded in rock or firm
soils during the past 20 years. An effective period range for base-isolated structures between 1.5 and 3.0 seconds 1.5 s TI 3.0 s was considered.
The parametric study was rigorous with the selection criteria, filtering, and scaling of
the acceleration data base, as discussed in detail in the paper. For linear-elastic isolators,
a database composed of 154 pairs of acceleration records was used. For bilinear isolators, 73 pairs of acceleration records fulfilled the criteria when scaling to the design
spectra of zone D-I of MOC-93 code, 90 pairs for zone C-I, and 135 pairs for zone B-I.
It was shown from the results of the parametric study that the mean values for the
2D / 1D ratios for linear-elastic and bilinear isolators are not constant, and they depend
on the period and the mechanical characteristics for the isolators. Also, for bilinear isolators it was found that the 2D / 1D amplification factors increase as the isolation system has a higher displacement response lower yield strength, Vy / W.

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

823

In general, displacement amplification factors due to orthogonal effects 2D / 1D


ratios corresponding to mean values are always below the 100% + 30% combination rule for the entire period range for both linear-elastic and bilinear isolators. However, the curves for the mean plus one standard deviation + Sdev of the 2D / 1D ratios surpass the 100% + 30% combination rule for the entire period range for linearelastic isolators, and depending on the yield strength Vy / W and the seismic zone
under consideration, they can also surpass the 100% + 30% combination rule for bilinear isolators.
The authors stated in the introduction of this paper that the questions to answer regarding orthogonal effects for the design of isolation systems are 1 is the 100%
+ 30% combination rule still valid and reasonable to compute design displacements? and
2 can a rational rule for the combination of orthogonal effects be independent of the
structural period and the characteristics of the ground motion?
Assessing results of the parametric study briefly described in this paper, the authors
have the following answers to these questions:
1. The 100% + 30% combination rule is a reasonable and conservative method to
compute design displacements, as the 2D / 1D ratios obtained in this study for
the mean curves are well covered, and this rule is not exceeded by a large margin even when using the results of the mean plus one standard deviation
+ Sdev. Nevertheless, this rule should be upgraded.
2. Although the observed variations on the 2D / 1D ratios can be taken into account with a simple straight line, the question is if this line should be a constant,
as the one given by the 100% + 30% combination rule, or if it is more appropriate to provide a linear equation in terms of the effective base-isolated period
for the structure. The use of constants in engineering practice has already lead to
several judgment errors in earthquake-resistant design. In the opinion of the authors, to provide a linear equation depending on the period would make the design process more transparent to practicing engineers at a minimum additional
effort, and would reduce the risk of judgment errors. However, it must also be
recognized that adopting such a recommendation would add complexity to our
already complex design codes. Building officials and code developers should
discuss this issue and take the best decision for engineering practice.
Finally, based on the extensive study presented in Prez-Osornio 2004, it was
found that the mean values of 2D / 1D ratios may depend on other factors such as the
magnitude for the earthquake explored or the earthquake source mechanism not explored. Therefore, more extensive studies with considerably larger databases are needed
in order to discern if there is a relationship between 2D / 1D ratios and the magnitude
for the earthquake or the earthquake source mechanism, and to study the impact of nearfault strong motions of large-magnitude earthquakes. Future research efforts should be
conducted in this direction.

824

A. TENA-COLUNGA AND M. . PREZ-OSORNIO

ACKNOWLEDGMENTS
Financial support of the National Science and Technology Council of Mexico Conacyt and Universidad Autnoma Metropolitana Azcapotzalco are gratefully acknowledged. Additional comments of anonymous reviewers of the manuscript were helpful in
improving this paper and are gratefully acknowledged.
REFERENCES
Applied Technology Council ATC, 1997. NEHRP Guidelines for the Seismic Rehabilitation of
Buildings, prepared for the Building Seismic Safety Council, published by the Federal Emergency Management Agency, FEMA-273, Washington, D.C.
Burton, M. R., Cronin, C. J., and Mayes, R. L., 2002. Effect of vertical motions on seismic
response of highway bridges, J. Struct. Eng. 128 12, 15511564.
Instituto de Investigaciones Elctricas, 1993. Manual de Diseo de Obras Civiles: Diseo por
Sismo, MOC-93, Comisin Federal de Electricidad in Spanish.
International Conference of Building Officials ICBO, 1997. Uniform Building Code, Whittier,
CA.
Lpez, O. A., Chopra, A. K., and Hernndez, J. J., 2001. Evaluation of combination rules for
maximum response calculation in multicomponent seismic analysis, Earthquake Eng. Struct.
Dyn. 30, 13791398.
Menun, C., and Der Kiureghian, A., 1998. A replacement for the 30%, 40% and SRSS rules for
multicomponent seismic analysis, Earthquake Spectra 14 1, 153156.
Multidisciplinary Center for Earthquake Engineering Research MCEER, 2003. Recommended
LRFD Guidelines for the Seismic Design of Highway Bridges, Part I: Specifications, Part II:
Commentary and Appendices, MCEER/ATC-49 Joint Venture Report MCEER-03-SP03.
Naeim, F., and Kelly, J. M., 1999. Design of Seismic Isolated Structures, First edition, John
Wiley & Sons, New York.
Nagarajaiah, S., Reinhorn, A. M., and Constantinou, M. C., 1991a. 3D-Basis: Nonlinear Dynamic Analysis of Three-dimensional Base Isolated Structures: Part II, Technical Report
NCEER-91-0005, National Center for Earthquake Engineering, State University of New
York at Buffalo.
, 1991b. Nonlinear dynamic analysis of 3D-base isolated structures, J. Struct. Eng. 117
7, 20352054.
Ordaz, M., 2001. Programa Degtra, manual en lnea, Instituto de Ingeniera, UNAM in Spanish.
Penzien, J., and Watabe, M., 1975. Characteristics of 3-dimensional earthquake ground motion,
Earthquake Eng. Struct. Dyn. 3, 365374.
Prez-Osornio, M. A., 2004. Amplificacin de los Desplazamientos de Aisladores de Base por
Efectos Bidireccionales, M.Sc. thesis, Divisin de Estudios de Posgrado de la Facultad de
Ingeniera, Universidad Nacional Autnoma de Mxico, October in Spanish.
Reviewer, 2005. Personal communication during the Earthquake Spectra peer review process.
Rosenblueth, E., and Contreras, H., 1977. Approximate design for multicomponent earthquakes, J. Eng. Mech. 103, 895911.

DESIGN DISPLACEMENTS FOR BASE ISOLATORS CONSIDERING BIDIRECTIONAL SEISMIC EFFECTS

825

Skinner, R. I., Robinson, W. H., and McVerry, G. H., 1993. An Introduction to Seismic Isolation, John Wiley & Sons, London.
Sociedad Mexicana de Ingeniera Ssmica, 2000. Base Mexicana de Datos de Sismos Fuertes,
volumen 2, CD-ROM, December.
Tena-Colunga, A., 1997. Evaluacin de un mtodo de diseo esttico para el aislamiento ssmico de estructuras de la costa Mexicana del Pacfico, Revista de Ingeniera Ssmica, SMIS,
57, 134 in Spanish;
, 1999. International seismic zone tabulation proposed by the 1997 UBC code: Observations for Mexico, Earthquake Spectra 15 2, 331360;
, 2002. Seismic design of base-isolated structures using capacity spectra, J. Earthquake
Eng. 6 4, 553586.
Tena-Colunga, A., and Gmez-Sobern, L. A., 2002. Torsional response of base-isolated structures due to asymmetries in the superstructure, Eng. Struct. 24 12, 15871599.
Tena-Colunga, A., and Zambrana-Rojas, C., 2004. Torsional response of base-isolated structures due to stiffness asymmetries of the isolation system, Proceedings, 13th World Conference on Earthquake Engineering, Vancouver, Canada, Paper No. 2022, CD-ROM.

Received 20 October 2004; accepted 29 November 2005

Das könnte Ihnen auch gefallen