Vous êtes sur la page 1sur 13

ARTICLE IN PRESS

Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering


journal homepage: www.elsevier.com/locate/soildyn

Seismic performance of three-dimensional frame structures with


underground stories
H. El Ganainy, M.H. El Naggar 
Department of Civil and Environmental Engineering, University of Western Ontario, London, Ontario, Canada N6A 5B9

a r t i c l e in fo abstract

Article history: This paper investigates the seismic performance of moment-resisting frame steel buildings with
Received 27 September 2008 multiple underground stories resting on shallow foundations. A parametric study that involved
Accepted 18 February 2009 evaluating the nonlinear seismic response of five, ten and fifteen story moment-resisting frame steel
buildings resting on flexible ground surface, and buildings having one, three and five underground
Keywords: stories was performed. The buildings were assumed to be founded on shallow foundations. Two site
Soil–structure interaction (SSI) conditions were considered: soil class C and soil class E, corresponding to firm and soft soil deposits,
Performance-based design (PBD) respectively. Vancouver seismic hazard has been considered for this study. Synthetic earthquake records
Seismic design of buildings compatible with Vancouver uniform hazard spectrum (UHS), as specified by the National Building Code
Underground stories
of Canada (NBCC) 2005, have been used as input motion. It was found that soil–structure interaction
Uniform hazard spectrum (UHS)
(SSI) can greatly affect the seismic performance of buildings in terms of the seismic storey shear and
Ground response analysis
moment demand, and the deformations of their structural components. Although most building codes
postulate that SSI effects generally decrease the force demand on buildings, but increase the
deformation demand, it was found that, for some of the cases considered, SSI effects increased both
the force and deformation demand on the buildings. The SSI effects generally depend on the stiffness of
the foundation and the number of underground stories. SSI effects are significant for soft soil conditions
and negligible for stiff soil conditions. It was also found that SSI effects are significant for buildings
resting on flexible ground surface with no underground stories, and gradually decrease with the
increase of the number of underground stories.
& 2009 Elsevier Ltd. All rights reserved.

1. Introduction approach is common in practice, it might differ in details


depending on the procedure used in the seismic analysis of the
The current state-of-practice for seismic design of buildings building.
with multiple underground stories involves approximate ap- It is important to incorporate the underground stories, base-
proaches that primarily differ according to the designer’s judg- ment walls, foundation soil and side soil explicitly in the
ment and experience. This is a consequence of lack of relevant mathematical model of the structure to be able to assess the
recommendations in building codes. Most building codes treat effect of the underground part of the building adequately on its
low and medium rise regular buildings with multi-level under- seismic performance. This is also essential since the current trend
ground stories with the same recommendations used for build- of using performance-based design approaches in lieu of tradi-
ings with surface foundations. tional force-based design approaches in the seismic design of
In general, buildings with multiple underground stories are buildings dictate that soil–structure interaction (SSI) analysis
designed by cropping the superstructure and analyzing it as a becomes an integral part of methods used in the seismic
fixed base structure founded on the ground surface. On the other evaluation of buildings. Perhaps the most popular approach in
hand, the substructure is designed for the seismic base shear and modeling the nonlinear response of the foundation soil and side
moment demand resulting from the superstructure in addition to soil is the Beam-on-a-Nonlinear Winkler Foundation (BNWF)
the seismic earth pressure acting on the basement walls, due to approach due to its merit of simplicity in defining the parameters
the oscillating mass of side soil. Even though this two step involved in the model.
The main objective of this paper is to better understand the
seismic performance of three-dimensional (3D) frame structures
 Corresponding author. Tel.: +1 519 661 4219; fax: +1 519 6613942. with multiple underground stories. To achieve this objective,
E-mail addresses: helganai@uwo.ca (H. El Ganainy), helnaggar@eng.uwo.ca nonlinear direct integration time–history analyses for 3D moment-
(M.H. El Naggar). resisting frame steel structures with above-ground stories ranging

0267-7261/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2009.02.003
ARTICLE IN PRESS

1250 H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261

from five to fifteen stories, and underground stories ranging from 2. Description of model buildings
zero (i.e. no basement) to five underground stories were
performed. The nonlinear structural analysis program 2.1. Description of model geometry and structural system
Perform-3D [1] was chosen for this research since it is dedicated
mainly for the performance assessment of 3D structures in the The models adopted herein are 4  5 bays moment-resisting
context of performance-based design (PBD). Its material library frame steel buildings, having a constant bay width of 7.2 m and
contains a wide variety of structural components formulated constant story height of 3.6 m. Fig. 1 shows the plan of the
to account for both geometric and material nonlinearity in repetitive story of the buildings. The lateral resisting system of the
structures. building constituted four perimeter frames along the periphery of
the building where the girders were rigidly connected to the
columns, except where the girders were connected to the weak
side of the columns. Fig. 2 shows the layout of the lateral resisting
system of a typical model building. On the other hand, the inner
frames work mainly as the gravity load carrying system where
girders were pin-connected to the columns.
The parametric study involves evaluating the seismic perfor-
mance of five, ten and fifteen story buildings with three under-
ground stories. The buildings were assumed to be resting on
shallow foundations. To further explore the effect of the number
of underground stories on the seismic performance of buildings,
the ten story building was analyzed for zero (i.e. no basement),
one and five underground stories.
The thickness of the reinforced concrete basement walls was
assumed 0.25 m considering that they will resist the lateral earth
pressure only. Their reinforcement ratio was 0.25%, in accordance
with the specifications of FEMA 310 [2] document. Although they
were not designed to be part of the lateral resisting system of the
building, they were included in its structural model since they
should affect its seismic response due to their large mass and
in-plane bending stiffness.
On the other hand, the thickness of the slabs was taken as
Fig. 1. Plan of the repetitive story of the buildings. 0.25 m to be consistent with approximately 1/30 of the slab span

Fig. 2. Lateral resisting system for a typical model building.


ARTICLE IN PRESS

H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261 1251

Table 1 Table 2
Unit weights and distributed loads used in defining the gravity loads acting on the Soil properties assigned for soil class C and soil class E.
buildings.
Soil class C Soil class E
Unit weights of materials (kN/m3)
Unit weight of steel 77 Shear wave velocity, Vs (m/s) 560 150
Unit weight of concrete 25 Dry unit weight, gdry (kN/m3) 21.00 18.00
Angle of internal friction, f1 40 30
Wall–soil friction angle, d1 25 20
Equivalent uniformly distributed load (kPa) Material damping ratio, e 0.05 0.05
Nonstructural components 1.1 Poisson ratio, n 0.35 0.35
Live load 2.4

as specified by FEMA 310 [2] document. The slabs were Table 3


represented in the structural model of the building using its Parameter used in calculating the seismic loads on the buildings using the NBCC
2005 equivalent static force procedure considering the Vancouver seismic zone.
weight in the gravity load case and as concentrated masses at the
center of gravity of each floor for the seismic analysis. In addition, Elastic design response spectrum parameters (g)
all the nodes lying in the plane of each floor were assigned a rigid Peak ground acceleration (PGA) 0.5
diaphragm constraint. However, the slabs were not modeled Sa (0.2) 1.00
Sa (0.5) 0.68
explicitly and consequently their bending stiffness was neglected.
Sa (1.0) 0.34
This is consistent with the assumption that the moment-resisting Sa (2.0) 0.18
frames form the lateral resisting system of the building.
Equivalent static force procedure parameters
Importance factor, IE 1.0
2.2. Gravity loads
Higher mode factor, Mv 1.0
Ductility-related force modification factor, Rd 5.0
The gravity loads assigned to the buildings were the own Overstrength-related force modification factor, Ro 1.5
weight of structural components, including the steel girders and
columns and the reinforced concrete slabs and basement walls.
It also included the weight of the nonstructural components
(e.g. cladding, partitions, floor finishing, etc.) in addition to the 4. Preliminary analysis of the buildings
live load assigned to the slabs.
Since the slabs were not modeled explicitly, their weight and The five, ten and fifteen story buildings were designed using
the live load they carry were included in the structural model by the structural analysis program ETABS [6] assuming fixed base
distributing its reaction on the supporting girders. Table 1 lists the condition at the ground surface. This step provided preliminary
unit weights and distributed loads used in defining the gravity sections for the structural members of the buildings, which would
loads acting on the buildings. be augmented by the underground stories, foundation soil and
side soil for further seismic analysis. Although the column
sections should increase below the ground level in consideration
3. Materials strengths and moduli of the added gravity loads from the underground stories, they
were not changed since the seismic performance of the buildings
3.1. Steel and concrete rather than its seismic design is the objective of this study.
The steel design feature included in ETABS was utilized to
The steel members of the building and the reinforcing steel of perform the seismic design of the buildings. The seismic loads
the basement walls were assumed to be of the same grade. The were calculated using the equivalent static force procedure as
steel yield strength was taken as 482,633 kPa, with an elastic specified by the NBCC 2005 for a building in Vancouver. Table 3
modulus of 199,948 MPa. The steel hardens to 689,476 kPa at a lists the parameters used in calculating the seismic loads acting
strain of 0.1, corresponding to a post-yield strain hardening ratio on the buildings. The structural members of the buildings were
of 1.1%. The steel Poisson’s ratio was taken as 0.3. The concrete of designed according to the loading cases and guidelines specified
0
the basement walls had f c ¼ 82,737 kPa, elastic modulus ¼ 37,232 by the NBCC 2005, the Canadian Standards Association (CSA) and
MPa and Poisson’s ratio ¼ 0.25. the Canadian Institute of Steel Construction (CISC). Table 4 lists
the preliminary sections for the girders and columns of the
buildings as obtained from the ETABS analysis and design stage.
3.2. Foundation and side soil

The buildings site was assumed to have a 30-m-thick deposit of 5. Intended behavior and performance levels
homogeneous soil underlain by the bedrock. Therefore, the
average properties in the top 30 m were used for calculating the The seismic performance of the model buildings was examined
foundation and side soil mechanical properties in accordance with with an emphasis on the effect of underground stories, foundation
the National Building Code of Canada (NBCC) 2005 specifications. soil and side soil on the building performance. To achieve this goal
Two scenarios were assumed for the soil deposit used in the the nonlinear structural analysis program Perform-3D [1] is used.
current study, namely: soil class C corresponding to ‘‘very dense
soil and soft rock’’; and soil class E corresponding to ‘‘soft soil’’ in 5.1. Intended behavior of structural components
accordance with the site classification of the NBCC 2005. Table 2
lists the properties assigned for these two soil classes in the The perimeter frames were considered the primary structural
current study from the ranges specified by NBCC 2005, Das [3,4] component and comprise the lateral resisting system of the
and FEMA 356 [5] document. buildings. Therefore, the perimeter frames are intended to
ARTICLE IN PRESS

1252 H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261

experience inelastic behavior in flexure but to remain essentially On the other hand, the basement walls are modeled using
elastic in axial and shear deformations. The connection panel inelastic fiber wall elements that could experience nonlinear
zones between the girders and columns of the perimeter frames behavior in in-plane bending, including: concrete fibers cracking
are also considered primary structural components, as they affect and crushing; and steel fibers yielding. However, they were
the performance level of the building, and are intended to assigned an elastic shear material to behave essentially elastic in
experience inelastic behavior in shear. However, the interior shear.
girders and columns that comprise the gravity load carrying
system are intended to behave elastically in flexure, axial and
shear deformations, since they are considered secondary structur- 5.2. Definition of performance levels
al components. These designations are in accordance with the
guidelines given by ASCE 41 [7] in classifying the structural To assess the performance of buildings, ASCE 41 [7] defines the
components of buildings in the context of the PBD principles. acceptance criteria of the structural components of the building in
The basement walls of the building contribute to its lateral terms of strength demand capacity ratios or deformation demand
resistance because of their orientation within the structural capacity ratios, depending on the force–deformation actions of the
system. Therefore, they also can be considered as primary structural components whether they are force-controlled or
structural components and hence are intended to experience deformation-controlled, respectively.
nonlinear behavior in in-plane bending. However, they should Perform-3D automatically calculates the strength and defor-
remain essentially elastic in shear, since shear failure in reinforced mation demand on the structural components of the building
concrete is a brittle mode of failure. This renders inelastic shear throughout the analysis steps. However, ASCE 41 [7] gives
behavior in structural members an undesired target performance. deformation capacities for the inelastic components correspond-
Finally, the slabs are intended to behave elastically and, as stated ing to the three target performance levels for structural compo-
before, were not included in the structural model. nents, namely: immediate occupancy (IO), life safety (LS)
To achieve these intended behaviors, the perimeter girders and and collapse prevention (CP). It defines the deformation
columns and the connection panel zones are modeled using capacities as multiple of the yield deformations of the compo-
inelastic frame and connection panel zone elements, respectively. nents. Table 5 gives the deformation capacities of the inelastic
They are assigned deformation-controlled force–deformation structural components encounter in the model buildings corre-
actions in bending and shear, respectively, in accordance with sponding to the IO, LS and CP performance levels and in
ASCE 41 [7] guidelines for structural steel components. The accordance with ASCE 41 [7] specifications. Deformation capa-
interior girders and columns are assigned force-controlled cities for perimeter girders and columns are expressed as multi-
force–deformation actions in flexure, axial and shear deforma- ples of the chord rotation (yy) at yield. The deformation capacities
tions (i.e. the components’ strengths are assigned to the elastic for the connection panel zones (assuming an improved
structural members without defining the associated plastic WUF-bolted web connection for the moment connections
deformations) as well as axial and shear modes of deformation between girders and columns) are expressed as functions of the
in perimeter girders and columns. The components’ strengths can girders’ depth (d). It should be noted that these deformation
be calculated in accordance with the established principles of capacities are plastic rotations and angular shear deformations,
mechanics (e.g. Load and Resistance Factor Design (LRFD) which dictates adding the yield deformations to them in order to
specifications for structural steel design considering a strength get the total deformation capacities.
reduction factor equals to unity). Perform-3D calculates the ASCE 41 [7] specifies that the strength capacity of structural
components’ strength from the geometric properties of members’ components should be assigned different values corresponding to
cross section (e.g. section modulus) and the associated mechan- the considered performance level. In the current study, compo-
ical properties of the cross section’s material (e.g. yield strength of nents that have force-controlled force–deformation actions are
steel). required to remain elastic. Therefore, for the performance levels
IO, LS and CP, the strength capacities were taken equal to the
nominal capacities of the components.
Table 4 To reduce the volume of analysis output results, Perform-3D
Preliminary sections for girders and columns of the model buildings as obtained groups demand capacity ratios of similar components together to
from ETABS. distill the results down to few ‘‘Limit States’’ that can be easily
used in assessing the performance of buildings. Each limit state
5 story building 10 story building 15 story building
groups similar demand capacity ratios (e.g. end rotation of
Perimeter girders W 27  146 W 27  94 W 27  146 perimeter girders) at a certain performance level (e.g. LS
E–W interior girders W 18  106 W 18  97 W 18  106 performance level). It then calculates the maximum demand
N–S interior girders W 24  104 W 24  104 W 24  104 capacity ratio, within each time step, for all the components in the
Perimeter columns W 14  90 W 14  145 W 14  233
Interior columns W 14  90 W 14  132 W 14  193
limit state. Perform-3D defines this maximum demand capacity
ratio as the ‘‘Usage Ratio’’ of the limit state at this time step. The

Table 5
Deformation capacities for inelastic structural components of buildings corresponding to IO, LS and CP performance levels.

Immediate occupancy Life safety Collapse prevention

Plastic rotation angle (radians)


Perimeter girders 1yy 6yy 8yy
Perimeter columns 1yy 6yy 8yy

Angular plastic shear deformation (radians)


Connection panel zones 0.01–0.00015d 0.0139–0.0002d 0.021–0.0003d
ARTICLE IN PRESS

H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261 1253

performance of the building can be thus assessed by ensuring that Table 6


the usage ratios of the target performance level of the components Shallow foundations’ properties used with the model buildings.
have not exceeded unity throughout the seismic event.
Soil class C Soil class E
Perimeter footings
Plan dimensions (m) 1.7  1.7 3.1  3.1
6. Foundation system Depth of foundation (m) 1.5 2.0
Ultimate bearing capacity, qu (kPa) 4206.63 1235.54
The foundation system of the buildings comprises shallow Soil passive resistance along footing front face (kPa) 20.00 15.00
Foundation vertical stiffness, kv (kN/m3) 18,48,932.57 57,725.55
spread footings. For the ease of modeling, it was assumed that
Foundation horizontal stiffness, KH (kN/m) 7,255,630.5 6,99,894.48
they are square in plan, and concentric with the supported
columns. Two footing models were considered for each building:
Interior footings
one for the interior columns and one for the perimeter columns,
Plan dimensions (m) 2.1  2.1 4.0  4.0
due to the substantial difference between the vertical load acting Depth of foundation (m) 1.5 2.0
on them. The strip footings beneath the basement walls and the Ultimate bearing capacity, qu (kPa) 4594.07 1359.5
semelles and straps connecting the footings were neglected in this Soil passive resistance along footing front face (kPa) 20.00 15.00
study. Foundation vertical stiffness, kv (kN/m3) 13,99,578.64 41,870.41
Foundation horizontal stiffness, KH (kN/m) 8,015,932.36 801,808.65
The side soil is assumed to be homogeneous throughout the
embedment depth of the building. It is assumed that it possesses
the same mechanical and physical properties of the foundation
soil.
The foundation and side soil were assumed to experience shear modulus, G ¼ 0.8Go, was used in calculating the
nonlinear behavior under seismic shaking. Therefore, they were vertical and horizontal stiffness of the footings. Table 6 lists
modeled using the Beam-on-a-Nonlinear Winkler Foundation the shallow foundations’ properties used with the model
approach that is capable of simulating the important aspects of buildings.
the nonlinear behavior of the foundation and side soil. El Ganainy [9] has shown that based on the BNWF approach,
the cyclic rocking, vertical and horizontal responses of shallow
6.1. Shallow foundations foundations can be modeled effectively using an assemblage of a
curvature hinge (or a moment–rotation hinge), shear hinge
The model buildings encountered in this study involved a wide connected in series with an elastic frame member attached to
range of footing vertical dead loads corresponding to different the bottom end of ground story columns. El Ganainy [9] has
scenarios considered. In addition, two soil classes were considered derived three bounding surfaces to couple the responses of these
in this study: soil class C corresponding to ‘‘very dense soil and hinges to be able to model the complete 3D response of shallow
soft rock’’ and soil class E corresponding to ‘‘soft soil’’. This wide foundations. To account for the ‘‘Soil Squeeze Out’’ phenomenon
range of variable parameters makes the task of sizing the footings [9] observed in the cyclic rocking response of shallow foundation,
cumbersome. Using a constant value for the vertical bearing El Ganainy [9] has shown that assigning an appropriate energy
capacity safety factor would result in a wide range of footing plan degradation factor to the curvature hinge would result in
dimensions, and consequently different foundations’ bearing adjusting the material damping from the cyclic moment–rotation
capacities and stiffness. This would complicate the comparison response of the footings and yield hysteretic moment–rotation
of the seismic performance of consistent buildings (e.g. ten story loops consistent with the S-shape loops observed from experi-
buildings with zero, one, three and five underground stories). mental results.
Therefore, a variable vertical bearing capacity safety factor was This modeling approach was adopted herein in modeling the
used when sizing the foundations, since it is neither the seismic shallow foundations of the model buildings encounter in this
design of the footings that is being investigated in this research study. Bilinear approximations for the moment–rotation relation
nor the evaluation of the code recommendations for the seismic and the horizontal force–shear displacement relation were
design of shallow foundations is to be done. assigned to the curvature and shear hinges, respectively [9].
For each soil class, two model footings were used: perimeter The geometric and mechanical properties of the curvature hinge,
columns footings and interior columns footings. These footings shear hinge and the elastic frame member were calculated,
were sized so that the vertical bearing capacity safety factor utilizing the mechanical properties of the specified soil classes,
ranged from 7.0 for five story buildings to 2.0 for fifteen story using the procedure outlined in El Ganainy [9].
buildings. The bearing capacities of the foundations are calculated To account for the soil squeeze out phenomenon, energy
using Terzaghi’s standard bearing capacity formula for square degradation factors were assigned to the curvature hinges at
footings. The soil passive resistance along the front face of the 0.55 and 0.8 for soil classes C and E, respectively. These values
footing is taken according to the presumptive values recom- provided good fit with the experimental results obtained from
mended by FEMA 356 [5] document for different soil types. TRISEE experiment for the high density (HD) and low density (LD)
However, the side friction along the footing side–soil interface is tests, respectively [9], noting that soil classes C and E are
neglected. approximately consistent with the relative densities of 85% and
The vertical and horizontal elastic stiffness of the foundations 45% of the HD and LD tests [9]. Hence, the corresponding energy
is calculated using the frequency-independent formulas given degradation factors values were used herein.
by FEMA 356 [5] document. To account for the cyclic nature A P–MB–ML bounding surface was assigned for the curvature
of the seismic load on the footings, the unload–reload stiffness of hinge to account for the interaction between the vertical and
the footing was used in lieu of the initial elastic stiffness. rocking responses of the footing [9]. El-Tawil and Deierlein’s
Allotey and El Naggar [8] recommend using an effective shear bounding surface [10,11], which is built-in Perform-3D, was used
modulus of 0.8 of the elastic shear modulus of the soil in in this regard. The fitting exponent m that controls the shape of
calculating the vertical stiffness of the foundation. Therefore, the P–M bounding surface was assigned a value of 2 and the
the elastic shear modulus, Go, of the soil was calculated from exponent n that controls the shape of the bounding surface in the
its shear wave velocity and mass density, then an effective MB–ML plane was assigned a value of 1.8 [9].
ARTICLE IN PRESS

1254 H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261

A VB–VL bounding surface was assigned for the shear hinge to structures. For flexible structures, such as moment-resisting
account for the interaction between the horizontal responses frame structures, the effect of the radiation damping would be
of the footing along its width and length. The elliptical equation of minimal and can be neglected.
this bounding surface is built-in Perform-3D and was used in this
regard [9]. Finally, the radiation damping through the foundation In the current study, the elastic stiffness and nonlinear behavior
soil was neglected. This is considered an acceptable approxima- of side soil were modeled. However, the oscillating mass of the
tion, since the hysteretic damping is more important in the case of side soil and the radiation damping effects were neglected.
seismic loading. Currently, there are no nonlinear bar elements in Perform-3D
capable of modeling the nonlinear backbone curve of the side soil
6.2. Side soil as shown in Fig. 3. Therefore, approximating assumptions were
introduced in order to make use of the available nonlinear bar
The effect of side soil on the seismic performance of buildings elements in Perform-3D to model the nonlinear response of the
with underground stories can be grouped into three aspects: side soil adequately.
The approximation done herein can be better understood if the
(1) Side soil serves as a flexible support to the building in lateral lateral pressure–lateral deflection relation of the side soil is
deformation. In the static case, it acts on the basement walls represented into two distinct parts as follows (with reference to
with a lateral pressure corresponding to the active earth Fig. 4):
pressure. Under seismic shaking, as the building oscillates
back and forth towards and away from the side soil, it (1) Under static loading condition, the side soil acts on the
responds like horizontal elastic springs. As the intensity of the basement walls with a static pressure corresponding to the
seismic shaking increases, the side soil could experience a active earth pressure.
nonlinear behavior, in which it cannot provide a lateral (2) As the building oscillates, the side soil acts like horizontal
pressure on the basement walls more than its passive nonlinear springs, where their ultimate compression capa-
resistance, Pp, while the building is swaying towards the cities are PpPa. However, they possess no tension capacity.
backfill. Also, it cannot provide a lateral pressure less than its
active resistance, Pa, while the building is swaying away from It should be noted that the soil considered in this study is
the backfill. In some types of soils, especially cohesive soil, cohesionless and should not experience gapping (i.e. the mini-
gapping could occur between the basement walls and the mum earth pressure cannot drop below the active value). Thus,
backfill as a result of the building oscillation. In this case, the this approach is considered to be adequate in representing the
lateral pressure of the backfill on the basement walls drops to backbone curve for the lateral pressure–lateral deflection relation
zero. This nonlinear behavior can result in hysteretic for- of the side soil.
ce–deformation actions in the side soil where the resulting Briaud and Kim [12] have recommended a set of static Py
hysteretic damping provides an additional source for dissipat- curves for sand and clay to be used within a beam–column
ing the earthquake energy. Fig. 3 shows the backbone curve method for the design of tieback walls. They validated the curves
for the hysteretic lateral pressure–lateral deflection relation of by comparing their predictions with the measured behavior of
the side soil. four-full scale tieback walls in sand and in clay. The Py curves
(2) Under severe seismic shaking, where the backfill experiences recommended by Briaud and Kim [12] for sand are used herein in
nonlinear response, the wedge of the soil behind the base- modeling the backbone curve of the hysteretic lateral pressur-
ment walls could fail and begin oscillating with the building, e–lateral deflection relation of the side soil.
either in-phase or out-of-phase. This oscillating mass of soil The active, Pa, and passive, Pp, earth pressures and the
could affect the seismic response of the building by altering its corresponding wall deflections ya and ya, respectively, are
effective oscillating mass. However, this oscillating soil mass calculated for sand as follows:
could be neglected in comparison to the mass of the structural
components and basement walls of the building, without Pa ¼ K a gZ cosðdÞ (1)
affecting the seismic response of the building significantly.
(3) Side soil dissipates the earthquake energy through radiation Pp ¼ K p gZ cosðdÞ (2)
damping. This additional damping can affect the seismic
response of the building, since it increases its effective ya ¼ 1:3 mm (3)
damping ratio. However, this effect would be most significant
for stiff structures, such as shear wall and braced frame yp ¼ 13 mm (4)

Fig. 3. Backbone curve of the hysteretic lateral pressure–lateral deflection relation for the side soil.
ARTICLE IN PRESS

H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261 1255

Fig. 4. Approximate representation of the lateral pressure–lateral deflection relation for the side soil.

where design response spectrum of the Vancouver area, specified by the


2
NBCC 2005, were used.
cos ðfÞ
Ka ¼ h pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i2
cosðdÞ 1 þ sinðf þ dÞ sinðfÞ= cosðdÞ

cos2 ðfÞ 7.1. Uniform hazard spectra (UHS) and compatible


Kp ¼ h pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i2
cosðdÞ 1  sinðf þ dÞ sinðfÞ= cosðdÞ earthquake records

g is the unit weight of the soil, Z is the depth at which the lateral The NBCC 2005 has introduced uniform hazard spectra that
earth pressure is calculated, f is the angle of internal friction of have a constant probability of exceedance of 2% in 50 years as a
the soil, and d is the wall–soil friction angle. function of spectral period. These spectra are based on a
Using Eqs. (1)–(4) and the soil properties for site class C and probabilistic seismic hazard assessment for different zones across
site class E listed in Table 2, the backbone curves for the hysteretic Canada [13]. The UHS eliminate the need to use standard spectral
lateral pressure–lateral deflection relation of the side soil were shapes scaled to the peak ground acceleration, thus providing a
calculated at selected depths. more site-specific description of the earthquake spectrum and
The backbone curve for the side soil can be adequately ensuring a uniform hazard level to be achieved for all spectral
modeled using the horizontal nonlinear springs and static active periods [13].
earth pressure distributed over the basement walls’ area. Thus, The UHS can be considered as a composite of all earthquake
nonlinear inelastic horizontal bar elements distributed horizon- events that contribute most strongly to the hazard at the specified
tally and vertically over the surface area of the basement walls probability level. In general, the dominant contributor to the
were used to model the nonlinear behavior of the side soil. The short-period ground-motion hazard comes from small-to-moder-
bar elements were equally spaced vertically at 1.2 m and ate earthquakes at close distances, whereas larger earthquakes at
horizontally at 7.2 m (i.e. the bar elements were distributed along greater distance contribute most strongly to the long-period
the underground perimeter columns, so that one bar is located at ground-motion hazard [14].
each story level and two intermediate bars are equally spaced The artificial ground-motion time histories compatible
within each story). The backbone curve for each bar element was with the 2% in 50 year UHS of the NBCC 2005 for the
calculated using its corresponding depth. The minimum active Vancouver area suggested by Atkinson and Beresnev [14] were
earth pressure acting on the basement walls was represented by used as input motion. They proposed an event of M6.5 at a
static concentrated loads acting at the bar elements’ locations. distance of 30 km to represent the short-period hazard and an
These loads were calculated as the value of the active earth event of M7.2 at a distance of 70 km to represent the long-period
pressure at the bar level multiplied by the horizontal and vertical hazard. In addition, an earthquake of M8.5 for the Cascadia event
spacing of the bar elements. scaled by a factor of 2.2 is used to simulate a great earthquake on
the Cascadia subduction zone. Therefore, three earthquake
records are required to cover the entire hazard represented by
7. Earthquake loads the NBCC 2005 UHS for the Vancouver area. Figs. 5–7 show the
three artificial acceleration records used in the dynamic analysis
The model buildings are assumed to be located in the of the buildings. Each building was analyzed for each of these
Vancouver area. Thus, earthquake records compatible with the three records.
ARTICLE IN PRESS

1256 H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261

Fig. 5. Acceleration record for M6.5 at a distance of 30 km event.

Fig. 6. Acceleration record for M7.2 at a distance of 70 km event.

Fig. 7. Acceleration record for M8.5 Cascadia event scaled by a factor of 2.2.

7.2. Ground response analysis motions at the considered foundation levels. In general, the
results show attenuation for the three bedrock motions, which is
The UHS given by the NBCC 2005 are defined with reference to most pronounced for M6.5 at a distance of 30 km event.
site class C that is defined as ‘‘very dense soil and soft rock’’. Thus,
the compatible records can be considered bedrock motions.
In general, the characteristics of the bedrock motion can be 8. Nonlinear dynamic analysis
amplified or attenuated while propagating from the bedrock
towards the ground surface. This alteration depends mainly on the Nonlinear dynamic analyses were performed to assess the
frequency content of the bedrock motion and the properties of the seismic performance of the model buildings. A series of nonlinear
soil deposit. Firm soil deposits, such as site class C, will probably direct integration time–history analyses were conducted using the
not alter the characteristics of the bedrock motion since they can nonlinear structural analysis software Perform-3D [1].
be considered part of the bedrock. Therefore, all buildings founded The seismic responses of five, ten and fifteen story buildings
on site class C can be analyzed for the bedrock motions shown in with underground stories ranging from zero (i.e. founded on the
Figs. 5–7, whether they are surface building or have multiple ground surface) to five underground stories were investigated.
underground stories. The response of the buildings was evaluated in terms of: (1) the
On the other hand, a soft soil deposit as site class E would magnitudes and distribution of the envelope of the story shear
probably alter the characteristics of the bedrock motion, by and moment demand on the buildings throughout each seismic
amplification or attenuation. Thus, the ground motions shown in event; (2) the maximum usage ratio of the limit states defining
Figs. 5–7 were propagated within a 30-m-thick deposit of soil site the performance level of the primary structural components of
class E (Table 2), performing nonlinear free-field site response the building. The performance of the following structural
analyses using the one-dimensional (1D) site response analysis components has been investigated: (1) perimeter columns’ end
program DEEPSOIL [15]. The G/Gmax modulus reduction curve and rotation; (2) perimeter girders’ end rotation; (3) connection panel
the equivalent damping ratio versus shear strain relationship for zones’ shear deformation, in terms of three performance levels,
sand given by Seed and Idriss [16] were assigned to the soil namely: immediate occupancy, life safety and collapse prevention,
deposit [17]. as specified by the ASCE 41 [7] standard.
Three ground response analyses were conducted, one for each
of the three ground motions as input bedrock motion and the
response of the soil deposit was evaluated. The ground motion 8.1. Nonlinear direct integration time–history analysis
was calculated at four foundation levels corresponding to
buildings having five, three and one underground stories and Perform-3D utilizes step-by-step integration through time
at the ground surface for buildings with surface foundations. using the constant average acceleration (CAA) method (also
Figs. 8–10 show the results of the ground response analyses in known as the trapezoidal rule or the Newmark b ¼ 14 method) to
terms of the acceleration time histories of the three bedrock calculate the seismic response of buildings.
ARTICLE IN PRESS

H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261 1257

Fig. 8. Ground response analysis results for M6.5 at a distance of 30 km event.

Fig. 9. Ground response analysis results for M7.2 at a distance of 70 km event.


ARTICLE IN PRESS

1258 H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261

Fig. 10. Ground response analysis results for M8.5 Cascadia event scaled by a factor of 2.2.

The input ground motions resulting from the ground response Rayleigh damping could overestimate the viscous damping in
analyses and the corresponding bedrock motions were used in nonlinear structures [18]. Using the modal damping could
the analysis of the buildings. The earthquake direction was set alleviate this defect.
to the W–E direction for all buildings. The ground motions were In the modal damping approach, the damping matrix is
given in 0.01 s time steps. Therefore, the integration time step for calculated for linear analysis using the elastic mode shapes and
the analysis was taken 0.01 s in order to accurately capture the periods of the structure utilizing the specified damping ratio. This
input ground motions. Also, it is sufficiently small to capture damping matrix is kept constant throughout the analysis steps.
the structure response, since it is considerably smaller than For nonlinear analysis, the deformed shape for the nonlinear
1
the recommended practical value of 12 of the structure period [18] structure generally contains contributions from the elastic mode
that ranges around 3.0 s for all model buildings encountered in shapes. However, the effective periods of vibration for these
this study. shapes are not the linear periods. Consequently, the mode shapes
There are two sources of damping in nonlinear structures: (1) are still damped, but since the effective period may have changed
for a structure that is essentially elastic, the earthquake energy is after yielding of structural components (probably increased)
dissipated through viscous damping; (2) after the structure yields, while the damping matrix is unchanged, the amount of damping,
hysteretic damping resulting from the inelastic behavior of the expressed as a proportion of critical damping, generally changes
structural components would add to the total dissipated energy. [18]. A shortcoming of using modal damping in nonlinear analysis
Modeling the structural elements of the building using inelastic is that only the calculated elastic mode shapes are damped.
components inherently accounts for this source of damping [18]. However, the higher modes are undamped.
To simulate viscous damping in buildings, either the modal To provide reasonable damping values, avoiding the pitfalls of
damping or Rayleigh damping can be used. both methods, a combination of modal damping and a small value
Rayleigh damping calculates the damping matrix of the of the Beta-K Rayleigh damping (with no Alpha-M damping) could
structure using a combination of the mass matrix and the initial be used. This is to insure that the Beta-K part will serve in
elastic stiffness matrix of the structure, multiplied by scaling damping the higher modes of vibration, and the modal damping
factors, a and b, for the mass and stiffness matrices, respectively. serves in damping the lower modes (i.e. elastic modes). For the
The Rayleigh damping is widely used in linear structural analysis. current study, the damping ratio was assigned to the model
However, it can lead to unrealistic large damping values in buildings as a combination of 3% modal damping in addition to
nonlinear analysis. The Rayleigh damping matrix is calculated 0.1% Beta-K Rayleigh damping, and six modes of vibration were
once at the beginning of the analysis using the initial elastic calculated for the buildings.
stiffness matrix of the structure and is used throughout the The analyses involved a gravity load case followed by a series
analysis. However, as the intensity of the seismic shaking of independent dynamic analyses, each having the gravity load
increases and the structure experiences nonlinear behavior in case as the preceding case. The self weight of the structure and the
the form of plastic hinging, the structure would soften and its active earth pressure acting on the basement walls were applied
stiffness would decrease and become much less than the elastic in the gravity load case. The p–d effects were considered for all the
value used initially in calculating the damping matrix. Hence, the vertical components of the building. The elastic frame members
ARTICLE IN PRESS

H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261 1259

used within the assemblage utilized in modeling the shallow


foundations’ response were not assigned p–d effects to eliminate
any additional artificial moments on the footings that could result
from the self weight of the structure.

9. Results and discussion

Because of the extensive amount of results obtained from the


parametric study, only some representative results are presented
here. The complete set of results can be found in El Ganainy [9].
The results are presented in the form of graphs comparing
response quantities for each of the five, ten and fifteen story
buildings with these conditions: fixed base, flexible foundation
(i.e. zero underground stories) and having one, three or five
underground stories. The response quantities presented include:
(1) the envelope of the story shear and moment demand on the
buildings throughout the earthquake events; (2) the maximum
usage ratio of the limit states defining the performance level of
the structural components of the building.

9.1. Story shear and moment demand

Figs. 11 and 12 show the envelope of story shear demand on


Fig. 12. Story shear demand on five story buildings—M8.5 Cascadia event scaled
five story buildings for the M8.5 Cascadia event scaled by a factor by a factor of 2.2 (soil class E).
of 2.2 for soil classes C and E, respectively, while Figs. 13 and 14
show the envelope of moment demand for the same conditions.
The figures show that the SSI decreased the base shear and
moment demands on buildings founded on stiff soil, but increased
the base story shear and moment demand on buildings founded
on soft soil conditions. For example, it increased by about 10–25%
of the fixed base buildings values for buildings founded on soil
class E. This shows that the common assumption that SSI has a
favorable effect by decreasing the seismic forces postulated by
almost all the design code does not always hold. The figures also
show that the behavior of buildings with underground stories is
closer to that of fixed base, i.e., as the number of underground
stories increased, the SSI effects decreased. This could be
attributed to the rigidity of the basement walls together with
the rigid diaphragm action of the underground stories’ slabs

Fig. 13. Story moment demand on five story buildings—M8.5 Cascadia event
scaled by a factor of 2.2 (soil class C).

rendering the embedded part of the building essentially a rigid


box, hence fixing the structure. These fixing effects would
probably increase with the number of the underground stories.
The results for the 10 and 15 story buildings (found in El Ganainy
[9]) show that the SSI effects are less pronounced for buildings
with longer period.
Inspecting the whole set of graphs for the envelope of story
shear and moment demand on the buildings found in El Ganainy
[9], the following observations are made:

Fig. 11. Story shear demand on five story buildings—M8.5 Cascadia event scaled (1) For buildings with underground stories or flexible foundations
by a factor of 2.2 (soil class C). founded on soil class C, the envelope of the story shear and
ARTICLE IN PRESS

1260 H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261

Fig. 15. Usage ratio of limit states for five story buildings—M8.5 Cascadia event
scaled by a factor of 2.2 (soil class C).
Fig. 14. Story moment demand on five story buildings—M8.5 Cascadia event
scaled by a factor of 2.2 (soil class E).

moment demand almost has not changed, either in magnitude


or in shape of distribution, compared to the fixed base
buildings case. The minimal SSI effect in this case is attributed
to the high soil stiffness and hence minimal change in the
dynamic characteristics of the soil–structure system.
(2) The effects of the SSI on the seismic loads are pronounced for
buildings founded on soil class E for all seismic events
considered. The envelopes of story shear and moment demand
for buildings with underground stories or flexible foundations
have changed in magnitude compared to the case of buildings
with fixed base condition. The magnitudes of base shear and
moment have mostly increased, especially at the base where
the increase generally ranged from about 10% to 25% of the
fixed base values.

9.2. Usage ratio of the limit states

Figs. 15 and 16 show the maximum usage ratio of the limit


states for five story buildings for the M8.5 Cascadia event scaled
by a factor of 2.2 for soil classes C and E, respectively. Fig. 15 shows
Fig. 16. Usage ratio of limit states for five story buildings—M8.5 Cascadia event
that for soil class C, the deformations of the structural compo- scaled by a factor of 2.2 (soil class E).
nents of buildings with flexible foundations or underground
stories are slightly different from those of buildings with fixed
base conditions. On the other hand, Fig. 16 shows that for the soil especially for the connection panel zones’ shear deformations
class E, the deformations of the structural components of performance levels. This observation clearly demonstrates that SSI
buildings with flexible foundations or underground stories are effects on the seismic performance of buildings increase as the
substantially different (larger) from those of buildings with fixed soil stiffness decreases.
base conditions. However, as the number of underground stories
increased, the deformation of the structural components gradu-
ally decreased approaching the fixed buildings values. Similar 10. Summary and conclusions
observations can be made from the rest of figures found in El
Ganainy [9], for ten and fifteen story buildings and different The seismic performance of buildings with multiple under-
seismic events. In general, the SSI effects are less pronounced as ground stories was investigated. Five, ten and fifteen story 3D
the period of the building increased. Comparing the deformations moment-resisting frame steel buildings with underground stories
of the structural components of buildings founded on soil class E ranging from zero to five underground stories have been
to that of the corresponding buildings founded on soil class C, it is examined. The buildings were assumed to be founded on shallow
noted that the deformation level generally increases for the soil foundations. Two site conditions were considered: soil class C and
class E case. The increase ranges from about 50% to about 300% soil class E, corresponding to firm and soft soil deposits,
ARTICLE IN PRESS

H. El Ganainy, M.H. El Naggar / Soil Dynamics and Earthquake Engineering 29 (2009) 1249–1261 1261

respectively. Vancouver seismic area has been considered for this case for buildings founded on soil class E. This would in turn
study. Synthetic earthquake records compatible with the Vancou- increase the lateral deflection of the whole building. Thus, SSI can
ver UHS, as specified by the NBCC 2005, have been used as input have a detrimental effect on the performance of buildings.
motion. For buildings founded on site class C, the bedrock motions
have been utilized in the seismic analyses of the buildings.
However, for buildings founded on site class E, ground response References
analyses have been performed to evaluate the characteristics of
the ground motion within the soil profile at each of the desired [1] Computers and Structures Inc. (CSI). Perform-3D—A computer program for
nonlinear analysis and performance assessment of 3D structures. 2007: 4.0.3.
foundation levels for the considered buildings. [2] American Society of Civil Engineers (ASCE). FEMA 310-handbook for the
The building foundations were modeled using an assemblage seismic evaluation of buildings: a prestandard. Washington, DC: Federal
of a curvature hinge, shear hinge connected in series with an Emergency Management Agency; 1998.
[3] Das BM. Principles of foundation engineering. Toronto, Ontario, Canada:
elastic frame member attached to the bottom end of the ground Thomson; 2007.
story columns. El Ganainy [9] has shown that this approach can be [4] Das BM. Fundamentals of geotechnical engineering. Pacific Grove, CA: Brooks/
used in modeling the 3D cyclic rocking, vertical and horizontal Cole; 1999.
[5] American Society of Civil Engineers (ASCE). FEMA 356—Prestandard and
responses of shallow foundations effectively. The behavior of the commentary for the seismic rehabilitation of buildings. Washington, DC:
side soil has been modeled using an approximate method: Federal Emergency Management Agency; 2000.
nonlinear bar elements were used to model its flexibility and [6] Computers and Structures Inc. (CSI). ETABS—an integrated building design
software; 2008 9.2.0.
ultimate horizontal capacity and static earth pressure was applied
[7] American Society of Civil Engineers. ASCE 41—seismic rehabilitation of
to the basement walls in order to simulate the effect of the initial existing buildings. American Society of Civil Engineers; 2006; ASCE/SEI 41-06.
active earth pressure prior to the seismic event. [8] Allotey N, El Naggar MH. Analytical moment–rotation curves for rigid
foundations based on a Winkler model. Soil Dyn Earthquake Eng
The dynamic analysis was conducted using the nonlinear
2003;23:367–81.
structural analysis software Perform-3D. The results indicate that [9] El Ganainy HM. Seismic performance of buildings with multiple underground
the SSI can considerably affect the seismic response of surface stories. M.E.Sc. thesis, University of Western Ontario, London, Ontario, Canada
buildings as well as buildings with underground stories founded 2008.
[10] El-Tawil S, Deierlein GG. Nonlinear analysis of mixed steel–concrete frames. I:
on soft soil conditions. In general, the results showed that SSI element formulation. J Struct Eng 2001;127:647–55.
effects are important for buildings founded on soft ground [11] El-Tawil S, Deierlein GG. Nonlinear analysis of mixed steel–concrete frames.
conditions. However, for firm ground conditions its effects can II: implementation and verification. J Struct Eng 2001;127:656–65.
[12] Briaud J, Kim N. Beam–column method for tieback walls. J Geotech
be neglected. Geoenviron Eng 1998;124:67–79.
The results showed that the SSI effects increased the base shear [13] Adams J, Atkinson G. Development of seismic hazard maps for the proposed
and moment demand on buildings founded on soft soil. This is 2005 edition of the national building code of Canada. Can J Civ Eng 2003;30:
255–71.
contrary to the common assumption that SSI effect on seismic [14] Atkinson GM, Beresnev IA. Compatible ground-motion time histories for new
forces is always favorable, postulated by almost all the design national seismic hazard maps. Can J Civ Eng 1998;25:305–18.
codes. It is clearly demonstrated that the base shear and moment [15] Hashash Y, Park D, Tsai CC. DEEPSOIL—a computer program for one-
dimensional site response analysis. University of Illinois at Urbana-Cham-
demand on short-period buildings could increase by up to 25% of
paign; 2005: 2.6.
the fixed base buildings values. [16] Seed HB, Idriss IM. Soil moduli and damping factors for dynamic response
The deformations of the structural components of the build- analysis. Berkeley, California: Earthquake Engineering Research Center,
ings have also been affected by the SSI. The deformations of University of California; 1970: UCB/EERC-70/10.
[17] Hashash Y, Park D, Tsai CC. DEEPSOIL—tutorial and user manual. University of
buildings with flexible bases have shown a considerable increase Illinois at Urbana-Champaign; 2005; 2.6.
that ranged from 50% to about 300% compared to the fixed base [18] Computers and Structures Inc. (CSI). Perform-3D—user manuals. 2007: 4.0.3.