Investigation into the Effects of

Foundation Uplift on Simplified

Seismic Design Procedures
Chad Harden,a… Tara Hutchinson,b… M.EERI, and Mark Moore,c… M.EERI

Uplifting of and yielding below shallow foundations supporting rigid

lateral force–resisting elements can provide additional nonlinearity into a
system’s overall force-deformation behavior. While this nonlinearity may be
advantageous, potentially reducing seismic demands, displacement
compatibility may result in overstress of lateral and/or gravity-resisting
elements. Incorporating this balance of benefit versus consequence in
structural design is one goal of performance-based earthquake engineering
共PBEE兲. There are a variety of approaches in design codes for estimating
seismic demands and incorporating “performance” as a design goal. Such
methods generally account for the displacement of an equivalent SDOF system
by reducing the design strength, however, not explicitly for the case of
foundation uplift. To address this shortcoming, this paper investigates the
relationship between the strength ratio R and the displacement ratio C1 using
the beam on nonlinear Winkler foundation 共BNWF兲 concept. Numerical
models were constructed considering a range of soil-structure natural periods
and a range of design R values. Nineteen ground motions with a broad range of
characteristics are used to conduct nonlinear time-history analyses. Results
from these simulations indicate that current suggestions for C1-R relations are
highly unconservative when uplifting foundations are anticipated. Revised
C1-R relations for uplifting foundations are presented and an example
numerical comparison provided. 关DOI: 10.1193/1.2217757兴

INTRODUCTION
It is well established that soil yielding beneath foundations can be an effective en-
ergy dissipation mechanism; however, this benefit may come with the expected costs of
excessive transient and permanent deformations. To realistically account for the sys-
tems’ performance, these consequences of permanent settlement and rotation must be
reasonably estimated and accounted for. A practical example, where the propagation of
the effects of a rocking foundation-superstructure becomes prominent is shown in Figure
1. The shear wall is typically stiffer than the frame and hence tends to attract load if the
subgrade is stiff, while the more flexible frame may be damaged by large displacements

a兲
RBF Consulting, 14725 Alton Parkway, Irvine, CA 92618-2027; E-mail: [email protected]
b兲
Department of Civil and Environmental Engineering, University of California, Irvine, CA 92697-2175;
E-mail: [email protected]
c兲
Rutherford and Chekene, 55 Second Street, Suite 600, San Francisco, CA 94105; E-mail:
[email protected]

663
Earthquake Spectra, Volume 22, No. 3, pages 663–692, August 2006; © 2006, Earthquake Engineering Research Institute
664 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

Figure 1. Example components of a combined structural system, where compliance at the

foundation level is allowed, for 共a兲 a stiff and strong foundation and 共b兲 a soft and flexible foun-
dation, below the rocking superstructure. 共Adapted from ATC-40 关ATC 1996兴.兲

due to a more compliant subgrade. Thus selection of the foundation strength and stiff-
ness is important to determine which structural component is the “weakest link” in the
load path. Incorporating a bed of Winkler foundation springs below each of these foot-
ings and modeling the system, as a whole, would provide a reasonable account for the
demands into both the frame and the shear wall. In recognition of this, present design
guidelines such as ATC-40 共ATC 1996兲 and FEMA-356 共ASCE 2000兲 recommend using
such a Winkler-based model to represent the stiffness of the soil-structure interface, and
provide methods to estimate stiffness of the individual vertical springs. Moreover, a non-
linear Winkler-based model can be used to more accurately conduct a performance-
based design 共PBD兲, as both the benefits and consequences of allowing the structure to
rock can be reasonably represented. However, in design practice, more simplified pro-
cedures are often desired, without fully modeling some aspects of the system 共such as
the foundation兲.
To evaluate the accuracy of simplified design procedures, in this paper, simulation
results considering foundation rotation and possible uplift are compared with current de-
sign methods used to account for the increase in displacement of an equivalent single-
degree-of-freedom 共SDOF兲 system when reduced design strength is provided. The basis
for the approach is the estimation of a displacement amplification factor C1, for a system
with a prescribed design strength ratio R 共C1-R estimations兲.

SEISMIC DISPLACEMENT DEMAND ESTIMATION METHODS

There are a variety of approaches suggested in design codes for estimating seismic
displacement demands. Design documents such as FEMA-356 共ASCE 2000兲, ATC-40
共ATC 1996兲, and UBC-97 共ICBO 1997兲 suggest procedures ranging from simple elastic
static analyses to nonlinear dynamic analyses. Four analysis methods suggested in cur-
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 665

rent codes for the design or rehabilitation of a building include the linear static proce-
dure 共LSP兲, the capacity spectrum approach 共CSA兲, the nonlinear static procedure
共NSP兲, and the nonlinear dynamic procedure 共NDP兲 共after FEMA-356 关ASCE 2000兴兲.
Linear refers to a linear elastic system, and nonlinear refers to systems, which behave
nonlinearly or exhibit inelastic response. A static procedure considers response 共forces
and displacements兲 from an applied set of monotonic loads or a target floor displace-
ment. A dynamic procedure may be either linear or nonlinear and includes a full time-
history analysis of a representative analytical model considering earthquake motion in-
put.
To apply the nonlinear static procedure, a target displacement at each floor, ␦t, may
be calculated:

T2e
␦t = C0C1C2C3SA g 共1兲
4␲2
where:
C0 = modification factor to relate spectral displacement of an equivalent SDOF
system to the roof displacement of the building MDOF system
C1 = modification factor to relate expected maximum inelastic displacements to
displacements calculated for linear elastic response
= 1.0 for Te ⱖ TS
= 关1.0+ 共R − 1兲TS / Te兴 / R for Te ⬍ TS
= 1.5 maximum
Te = effective fundamental period of the building
TS = characteristic period of the response spectrum
R = ratio of elastic strength demand to calculated yield strength
C2 = modification factor to represent the effects of pinched hysteresis shape, stiff-
ness degradation, and strength deterioration on maximum displacement re-
sponse 共equal to 1 for linear procedure兲
C3 = modification factor to represent increased displacements due to dynamic P -
delta effects
SA = spectral acceleration at the fundamental period and damping ratio of the
building
The target displacement at each floor is then used to calculate the resulting forces in
the component members, and acceptance or rejection of the design or retrofit results
based on the acceptance criteria of the component as a function of the designated target
building performance level.

PREVIOUS INVESTIGATIONS INTO CODE PRESCRIPTIONS

There have been a number of investigations evaluating the applicability of simplified
spectral-based design procedures, specifically force reduction–displacement ductility
共R-µ⌬兲 relation methods. Such works have involved considering the response of systems
to pulse-type excitations 共Veletsos and Newmark 1960; Veletsos and Newmark 1964;
and Cuesta and Aschheim 2000, 2001a, b, c兲, elasto-plastic systems 共Newmark and Hall
666 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

1973, 1982; Veletsos et al. 1965; Veletsos 1969; Veletsos and Vann 1971; and Lai and
Biggs 1980兲; degrading systems 共Riddell and Newmark 1979, Al-Sulaimani and Roes-
sett 1984, Riddell et al. 1989, and Vidic et al. 1994兲; and softening systems 共Hidalgo and
Arias 1990兲, and other important parameters deemed to potentially affect the accuracy of
these estimations. These studies have improved upon such factors, providing semi-
empirically derived values to support increased accuracy 共Elghadamsi and Mohraz 1987,
Peng et al. 1988, Takada et al. 1988, Nassar and Krawinkler 1991, Miranda 1993, Rid-
dell 1995, Ordaz and Pérez-Rocha 1998, Chopra and Goel 1999兲. To refine the accuracy
of the analysis, the force reduction may be presented as a function of ductility, damping,
stiffness characteristics, natural and characteristic period, and/or site soil-profile type.
Miranda and Bertero 共1994兲 provide a review of the evolution of the strength reduction
factor R from Newmark and Hall 共1973兲 to the publication date. More recent studies
include work by Cuesta et al. 共2003兲, who investigate the relationship between R as a
function of µ⌬ and the ratio T / TS, where µ⌬ is the displacement ductility demand of the
system, T is the natural period of the system and TS is the characteristic ground motion
period. Cuesta et al. 共2003兲 found that both the FEMA and ATC methods were appro-
priate for models with significant stiffness degradation. Fajfar 共2000兲 describes the N2-
method, where expected inelastic displacements of an equivalent SDOF system are cal-
culated using inelastic spectra as a function of ductility, similar to current code
recommendations. The method presented in this study is fundamentally different than
the N2 capacity spectrum method as described in Fajfar 共2000兲, in that elastic spectra are
used to estimate inelastic displacements as a function of both R and the relative periods
of the SDOF and the demand spectrum, and attempt to include the effect of soil-
structure interaction.
Relative to rocking-spectrum analyses, Makris and Konstantinidis 共2002兲 find that
rocking structures should not be replaced with SDOF simplifications, because of the
nonlinear response of the soil-structure interaction and its sensitivity to the input de-
mand. However, the study presented in this paper attempts to incorporate the sensitivity
of the Winkler model to the characteristics of the input ground motion through 共1兲 the
relative magnitude of the soil-structure systems’ natural period to the ground motion
characteristic period, and 共2兲 the relative magnitude of the spectral acceleration of the
elastic rocking system to the elastic rocking system with uplift allowed. Therefore, the
lack of demand and response coupling in current simplified design approaches, as ob-
served by Makris and Konstantinidis 共2002兲, is accounted for in this paper.

SCOPE OF THIS STUDY

The approach in current design codes allows one to compromise between strength
and ductility while still attempting to control performance targets, through a
displacement-based design procedure. An accurate assessment of these strength and dis-
placement values when evaluating a building for rehabilitation becomes increasingly im-
portant due to cost-performance trade-offs. Often the advantage of allowing period elon-
gation and energy dissipation through the foundation compliance is of great interest in
these situations. However, if a simplified analysis procedure is to be performed, an
evaluation of the accuracy of these procedures is warranted. Therefore, in this paper a
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 667

suite of nonlinear dynamic analyses is performed using a Winkler model with a simple
lumped mass attached, considering two foundation conditions 共cases兲: 共1兲 elastic springs
共rocking system兲 and 共2兲 elastic springs allowed to uplift 共rocking-uplifting system兲.
Nonlinearity in the analysis is introduced through possible uplift for case 2. Results from
these analyses are compared with design code relations for the strength 共R兲 factors and
displacement 共C1兲 factors to be used in a simplified analysis method. All analyses are
performed in the OpenSees 共Open System for Earthquake Engineering Simulation兲 plat-
form, developed by the Pacific Earthquake Engineering Research Center 共PEER兲 at the
University of California, Berkeley.

CONSISTENT DEFINITIONS C1-R

Consistent definitions for the quantities C1 and R must be adopted in the context of
the rocking structural system to allow proper comparison with design approaches. In its
most basic definition, the R factor is the ratio of elastic demand to the specified yield
force in the nonlinear system:

SAmg
R= 共2兲
Hy
where SA = spectral acceleration at the fundamental period and damping ratio of the
building, m = seismic mass, g = acceleration due to gravity, and Hy = lateral yield
strength.
In this work, the yield force for the analysis of structures supported on rocking-uplift
foundations 共case 2兲 is defined as the force applied at the top of the superstructure,
which just causes separation of the soil-foundation interface, or uplift. One conventional
definition of Hy is the intersection between the first significant pre-yield and post-yield
stiffness of the pushover analysis, which may be calculated by an equivalent-area bilin-
ear response. This definition can be applied in the current analysis, where the pre-yield
and post-yield stiffnesses are defined at the onset of uplift.
For this study, a range of R factors is prescribed and, considering the elastic state of
stress, a corresponding vertical factor of safety FSq against bearing capacity mobiliza-
tion can be calculated 共e.g., Allotey and Naggar 2003兲. This will define the limits of
validity for the current study, such that the soil structure interaction will behave elasti-
cally even while uplift occurs. Consider the vertical factor of safety FSV, under a purely
vertical load. As a footing rotates, the bearing capacity at an extreme end will increase.
A variable FSq can be defined as the ultimate bearing capacity divided by the maximum
bearing pressure beneath the footing 共at an extreme end兲. Note that under a purely ver-
tical load, FSq is equal to FSV, but FSq decreases as lateral load is delivered to a system
and the footing rotates. This relationship is described graphically in Figure 2, and de-
fines the limits of validity of the current study. For a lateral force “R” times the yield
force Hy 共force required to cause incipient uplift兲, the subgrade will remain elastic for
footings with a sufficient FSV such that the curve does not fall below FSq = 1.0. Note that
typical “R” values fall in the range of FSV found in conventional design practice, which
is approximately FSV = 3 – 4.
668 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

Figure 2. Parameter range for an assumption of a foundation remaining in the elastic range.

The second quantity under investigation is C1, the ratio of displacement demand on
an elastic system to the displacement demand on the nonlinear system. C1 is commonly
defined as a function of the natural period of the structure. The analysis presented is an
attempt to verify the current code C1-R functions for a range of periods.
From a numerical standpoint, a robust approach to the problem is to define structures
that have an exact R over a range of periods, which are subject to a large collection of
earthquakes. The structures defined for this task will be simple in nature; a lumped mass
supported by an elastic column and connected to an elastic foundation on an elastic Win-
kler subgrade. The definition of R over a range of periods and for this simple system
must be consistent.
Settlement of an elastic foundation of length L, superstructure weight W, and global
soil vertical stiffness KZ may be determined as:

W
s= 共3兲
KZ
Since in this study the yield point of the system is defined as the point of uplift, this
state must be well defined. Psycharis 共1981, 1983兲 gives the critical rotation at incipient
uplift, shown graphically in Figure 3:

2s
tan共␪cr兲 = 共4兲
L
This assumes the structure above is stable with respect to overturning. Thus the soil
springs are stiff enough to prevent instability due to overturning.
The critical moment applied to the base that would just cause uplift, Mcr, is found by
summing the moments about the center of the footing. The 共uniform兲 vertical displace-
ment at any point is s, and the additional contribution to this value due to rotation is
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 669

Figure 3. Notation to define critical rotation for an elastic subgrade modulus.

labeled s␪共x兲. Since force is a function of displacement via the vertical stiffness of the
system, the critical moment is found by integrating the displacements as a function of
distance from the center of the footing, multiplied by stiffness per unit length of footing.

Mcr = 冕 L/2

−L/2
共s + s␪共x兲兲
KZ
L
xdx 共5兲

The critical moment can then be simplified to the “kern” assumption:

KZ 2 WL
Mcr = L tan共␪cr兲 = 共6兲
12 6
For these analyses, the beam-column stiffness is taken as sufficiently large, as envi-
sioned for a stiff shear wall. It is therefore reasonable to neglect any contribution of the
shear wall to the horizontal displacement of the system, since the variable of interest is
foundation rocking and uplift. The assumption of a rigid beam 共foundation兲 and column
共shear wall兲 in the following derivations is accurate to within approximately 0.1% error
for even the highest bound of soil stiffnesses given a concrete shear wall as the dominant
superstructure element. However, to avoid any contribution of superstructure flexibility
to the flexibility of the entire system in general applications, the “beam” element should
be sufficiently stiff such that Eb* Ib / 共Es* Is兲 ⱖ 50, and the “column” element should be
sufficiently stiff such that Ec* Ic / 共Es* Is兲 ⱖ 100, where E and I are elastic modulus and
moment of inertia for the beam 共subscript “b”兲, column 共subscript “c”兲, and subgrade
共subscript “s”兲.
Continuing with the rigid beam-column assumption, knowing Mcr and ␪cr at uplift,
the two final factors needed to solve for a set of lumped mass models with identical
subgrade reactions, which will give a constant line of R for a range of periods, is the
magnitude of the lumped mass and the distance from the foundation to the lumped mass
共i.e., the lever arm兲. Solving for Hy from Equation 2 and substituting Mcr gives the mo-
670 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

ment arm or effective height h to the center of mass at the top of the structure:

RL
h= 共7兲
6SA
Given the range of periods through which the line of constant R will sweep, and the
range of h’s previously defined, a range of lumped masses can be calculated. The elastic
period of the structure can be defined using the horizontal stiffness of the system:

Hy Mcr
KL = = 共8兲
⌬top h⌬top
Substituting the lateral stiffness and the displacement of the top of the structure as a
function of rotation into the definition of the natural period a structure and solving for
mass gives

m=
KZ LT
48 h␲
冉 冊 2
共9兲

Since the natural period is a function of mass and stiffness, and the R-value is a func-
tion of SA and the effective height h; then for a given R-value with a fixed subgrade
modulus and foundation aspect ratio, but over a range of natural periods and further a
group of earthquakes, each point on a graph of C1 versus T will have a different mass m
and effective height h.

GROUND MOTIONS SELECTED

A suite of ground motions was chosen to represent a wide range of earthquake char-
acteristics for use as input to the models. In this work, the simulations use the longitu-
dinal components of the 19 unscaled time histories described by Somerville and Collins
共2002兲. Table 1 lists the characteristics of interest for the longitudinal directions of the
ground motions. These ground motions represent a broad range of peak ground accel-
erations 共PGA兲, peak ground velocities 共PGV兲, and peak ground displacements 共PGD兲
levels. PGA levels range from 0.13 to 0.75 g, PGV ranges from 9.1 to 84.8 cm/ s, and
PGD ranges from 1.2 to 18.7 cm.
The characteristic period TS in this work is defined as the intersection of the constant
acceleration and constant velocity region of the time history 共Newmark and Hall 1982兲.
To calculate TS for each motion, a mean fit through the values of pseudo-acceleration,
pseudo-velocity, and pseudo-displacement was performed. The range of characteristic
periods for these motions is TS = 0.26 to 0.88 seconds.
Figure 4a and 4b show the mean 共µ兲 and mean ± standard deviation 共␴兲 for all 19
ground motions 共longitudinal component兲 of the spectral acceleration and spectral dis-
placement versus period, respectively. As evident in the µ ± ␴ plots, the suite of motions
selected have broad spectral characteristics. Figure 4c and 4d shows the mean 共µ兲 and
mean ± standard deviation, normalized by the mean, for all 19 ground motions 共longi-
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 671

Table 1. Ground motions used in this study 共longitudinal component兲

EQ Dist* PGA PGV PGD TS

Note Station Name 共km兲 共g兲 共cm/s兲 共cm兲 共sec兲

1 Palm Springs Airport NPS_plma 9.6 0.204 12.1 2.4 0.26

2 Canoga Park, Topanga Canyon Blvd NR_cnpk 17.7 0.331 31.6 12.6 0.35
2 Encino, Ventura Blvd #1 NR_env1 17.7 0.503 56.6 16.0 0.45
2 Encino, Ventura Blvd #9 NR_env9 17.9 0.247 31.3 8.1 0.48
2 North Hollywood, Lankershim Blvd #1 NR_nhl2 18.4 0.185 27.6 5.5 0.49
2 Arleta, Nordhoff Fire Station NR_nord 9.4 0.344 40.6 14.6 0.64
2 Northridge, Roscoe #1 NR_nrr1 13.7 0.390 37.4 15.8 0.56
2 Sun Valley, 13248 Roscoe Blvd NR_rosc 10.8 0.444 38.2 10.0 0.88
2 Sepulveda VA Hospital NR_spva 9.2 0.753 84.8 18.7 0.58
2 Van Nuys, Sherman Way #1 NR_vns1 12.8 0.375 37.7 9.2 0.68
2 Van Nuys, Sherman Circle #1 NR_vnsc 12.8 0.474 31.4 7.6 0.52
2 Van Nuys, 7-story hotel NR_vnuy 11.3 0.469 54.4 11.3 0.42
2 Woodland Hills, Oxnard Street #4 NR_whox 20 0.318 41.8 12.3 0.39
3 Los Angeles, 14724 Ventura Blvd SF_253 16.3 0.203 21.5 12.3 0.34
3 Los Angeles, 15910 Ventura Blvd SF_461 16.2 0.151 17.2 8.4 0.76
3 Los Angeles, 15250 Ventura Blvd SF_466 16.4 0.164 24.1 11.5 0.48
3 Glendale, Muni Bldg, 633 E Broadway SF_glen 18.8 0.227 32.0 10.0 0.83
3 Van Nuys, 7-story hotel SF_vnuy 9.5 0.134 23.3 14.9 0.87
4 Caltech, Braun Athletic Building WH_athl 16.6 0.149 9.1 1.2 0.41

Notes:
共1兲 Earthquake location and date: North Palm Springs, 8 July 1986
共2兲 Earthquake location and date: Northridge, 17 January 1994
共3兲 Earthquake location and date: San Fernando, 9 February 1971
共4兲 Earthquake location and date: Whittier Narrows, 1 October 1987
*
Distance is defined as the closest surface distance from the fault rupture to the ground motion recording station.

tudinal component兲 of the spectral acceleration and spectral displacement versus period,
respectively. These plots further substantiate the broad spectral characteristics of the mo-
tions selected.

ANALYSIS RESULTS AND DISCUSSION

Numerical models were constructed considering a range of natural periods 共T = 0.1 to
2.0 seconds, at increments of ⌬T = 0.2 s兲 and a range of design R-values 共R = 1.5, 2.0,
4.0, and 8.0兲. Each of these models was subjected to the ground motions listed in Table
1. Simulation results for these cases are presented in Figure 5, for the condition of R
= 4.0. Figure 5 shows C1 values versus normalized period T / TS, where Ts is the charac-
teristic period noted in Table 1. The displacement ratio C1 in this study is defined as
672 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

Figure 4. Mean 共µ兲± standard deviation 共␴兲 of 共a兲 spectral acceleration 共SA兲 versus period and
共b兲 spectral displacement 共SD兲 versus period for the longitudinal component of the ground mo-
tions considered in this study. Also shown are the mean 共µ兲± standard deviation 共␴兲 normalized
by the mean for 共c兲 spectral acceleration 共SA兲 versus period and 共d兲 spectral displacement 共SD兲
versus period for the longitudinal component of the ground motions considered in this study

Figure 5. C1 versus normalized period for R equal to 4.0 共Six data points above C1 equal to
20兲.
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 673

⌬nonlinear
C1 = 共10兲
⌬elastic
where ⌬nonlinear is that calculated for the system allowed to uplift, while ⌬elastic results
from the system where uplift is not allowed. For comparison, the FEMA-356 recom-
mended value for C1, as well as a mean fit to the data is shown. ATC-40 suggests a limit
of C1 = 2.0 at T ⬍ 0.1 seconds, whereas FEMA-356 suggests a limit of C1 = 1.5 with the
intersection of the curve defined in Equation 1. The nonlinear regression to the data is
taken as the mean 共C1-ave兲 for T / TS ⬍ 0.5, and where T / TS ⱖ 0.5:

1
共11兲
C1 =
1 + Ae
−B 冉 T
TS 冊
−0.5

1
To best fit the data and meet the average value at T / TS = 0.5, A = C1−ave − 1 and B are
varied such that the correlation coefficient between the fit curve and the data points is a
maximum.
Figure 5 illustrates a large range for the calculated values of C1 for R = 4.0, though a
discernible trend is evident. At small period ratios, approximately less than 1.0, C1 is
very large. At larger period ratios, C1 converges to unity as observed by others and rec-
ognized in design codes. This is consistent with the long period equal displacement prin-
ciple. Results for other simulations indicate that at smaller R-values 共R = 1.5兲, current
code recommendations result in a conservative estimate of C1 for most short period
cases 共T / TS ⱕ 1.0兲. However, as R is increased 共R = 2.0, 4.0, or 8.0兲, the recommenda-
tions suggested in current code documentation tend to be highly unconservative, indi-
cating rocking-induced transient displacement contributions may be very large. Overall,
the results indicate that the current displacement coefficient method 共based on recom-
mendations for C1-R relations兲 may be unreasonable for estimating rocking-induced dis-
placements. An inspection of the data may help assist in alternative forms of C1-R re-
lations for estimating these inelastic displacement contributions.
Although binning the data by basic ground motion characteristics such as PGA or
PGV does not reveal any consistent trends between varying values of R, the data appear
to be sensitive to the ratio of the spectral accelerations determined at the periods of the
nonlinear and elastic systems, termed SAnonlinear and SAelastic, respectively. The term
“nonlinear” is used to differentiate the elastic foundation allowed to uplift from the elas-
tic system not allowed to uplift. From the viewpoint of a “pushover” analysis, the load
versus displacement curve of the elastic system not allowed to uplift is a straight line,
while the elastic system allowed to uplift is linear until uplift is reached, after which the
stiffness degrades 共note that no hysteresis will be observed with repeated loading since
the Winkler springs are elastic兲. The spectral acceleration for each “elastic” and “non-
linear” system with a defined period T is defined as the peak horizontal acceleration re-
sponse from an input ground motion. A normalized parameter is defined as ␺SA
= SAnonlinear / SAelastic. The data are binned into three groups of equal number of samples,
where the first group all have values of ␺SA ⬍ X1, the second group X1 ⱕ ␺SA ⬍ X2, and
the third group ␺SA ⱖ X2. The values of X1 and X2, as well as the maximum and mini-
674 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

Table 2. Binning results

R X1 X2 Max ␺SA Min ␺SA

1.5 0.947 1.000 1.651 0.854

2.0 0.881 0.995 1.556 0.619
4.0 0.679 0.874 1.171 0.417
8.0 0.518 0.760 1.150 0.275

mum values of ␺SA, for all R cases are listed in Table 2.

Table 3 lists the calculated parameters A and B resulting in the highest correlation
coefficient for Equation 11, for each value of R and each bin of ␺SA.
Figure 6 shows the results of C1 versus normalized period in binned form with simu-
lation data points for R = 4.0. The recommended curve, as discussed in this paper is over-
lain on the data, as well as a logistic curve similar to the recommended curve but with a
regressed function of the form y = a / 共1 + be−cx兲, where a, b, and c are regressed coeffi-
cients. The former applies a limit on C1, termed “Regression Through Data 共Per Equa-
tion 11兲.” The latter applies no limit in the short-period ratio range, termed “Regression
Through Data.” The recommended curve is typically conservative with respect to the lo-
gistic curve for longer period structures. However, the recommended curve provides a
more realistic design displacement for shorter period structures 共T / TS ⬍ 0.5兲 since only
the mean value of C1 is considered in this range, where the logistic data fit would pre-
scribe an excessively conservative design.
Using the regression analysis presented in Equation 11 in conjunction with Table 2
and 3 allows one to estimate the lateral design displacement of a rocking-dominated sys-
tem using a design spectrum approach. Consider, for example, a system with a natural

Table 3. Regression coefficients for Equation 11

R Group A B

1.5 I, ␺SA ⬍ X1 −0.013 0.01

1.5 II, X1 ⱕ ␺SA ⬍ X2 −0.087 3.89
1.5 III, ␺SA ⱖ X2 −0.479 0.52
2.0 I, ␺SA ⬍ X1 0.051 0.45
2.0 II, X1 ⱕ ␺SA ⬍ X2 −0.259 3.05
2.0 III, ␺SA ⱖ X2 −0.888 0.56
4.0 I, ␺SA ⬍ X1 −0.503 0.01
4.0 II, X1 ⱕ ␺SA ⬍ X2 −0.881 0.12
4.0 III, ␺SA ⱖ X2 −0.967 0.07
8.0 I, ␺SA ⬍ X1 −0.930 0.07
8.0 II, X1 ⱕ ␺SA ⬍ X2 −0.984 0.02
8.0 III, ␺SA ⱖ X2 −0.971 0.10
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 675

Figure 6. Binned data—C1 versus normalized period for R equal to 4.0. Note, 6 data points
共6% of binned subset兲 of outliers with C1 greater than 20 not shown for 共c兲. The curve “Re-
gression Through Data 共Per Equation 11兲” is recommended and defined in conjunction with
Table 2 and 3.
676 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

period equal to 2.0 seconds and a design ground motion with a characteristic period of
0.4 seconds. First, from the intersection of the model pushover and the design spectrum,
say the SA of a nonlinear system is estimated as 0.5 g 共for example兲. Using either a
similar analytical model not allowed to uplift, or with a lateral stiffness defined by one
rotational spring in place of the Winkler model 共an alternative approach suggested in
code兲, an intersection with the design spectrum gives 0.6 g. Therefore, ␺SA
= SAnonlinear / SAelastic = 0.5/ 0.6= 0.83. Given a target design R value of 4.0, X1 = 0.68 and
X2 = 0.87 from Table 2. From Table 3, the system falls into group II under R = 4.0 be-
cause X1 ⱕ ␺SA ⬍ X2, therefore A = −0.881 and B = 0.12. Finally, from Equation 11,
C1 = 1 / 关1 − 0.881e−0.12*共共2/0.4兲−0.5兲兴 = 2.06.
In summary, for the range of R values studied it appears that at low period ratios and
for large values of ␺SA, where the spectral acceleration of the nonlinear system is greater
than the spectral acceleration of the linear system, large displacement ratios can result.
FEMA-356 uses a limit of C1 = 1.5 and ATC-40 gives a limit of C1 = 2.0; much larger
displacements are found in this study for the uplifting system for R values greater than
1.5, suggesting the current codes may be unconservative for the case of a shallow foun-
dation allowed to uplift.

ILLUSTRATIVE DESIGN EXAMPLE

An illustrative design example using recommended simplified design procedures is
explored, using the suggestions from the calculated C1-R relations for a rocking shallow
foundation allowed to uplift. Additionally, rocking-induced lateral displacements and the
level of damping are used to estimate accrued settlements based on empirical damping
versus distortion data.
The design example foundation structure is modeled as a beam on nonlinear Winkler
foundation 共BNWF兲, using nonlinear q–z springs based on the formulation presented in
Boulanger et al. 共1999兲. Salient Winkler spring system parameters are based on sugges-
tions in Harden et al. 共2005兲 and are graphically shown in Figure 7. Recommendations
for Winkler spring system modeling were based on evaluation against results from cen-
trifuge testing at U.C. Davis 共Gajan et al. 2003, Rosebrook and Kutter 2001兲 and other
1-g test data 共Bartlett 1976, Negro et al. 1998, Weissing 1979兲. The Winkler elements
degrade in stiffness prior to reaching ultimate capacity. A parabolic-shaped bearing ca-
pacity beneath the foundation is recommended to account for the experimentally ob-
served stress distribution. Finally, the study recommends a specific region at the ends of
the foundation receive a larger stiffness to account for observed stiffening due to densi-
fication of the soil medium associated with rotation of the foundation. Such an approach
is similar to procedures outlined in ATC-40 共1996兲.
For this design example, structure aspect ratios of foundation length L to width B of
3.2 共=L / B兲 and shear wall height H to wall length Lw of 3.0 共=H / Lw兲 are selected. The
geometric configuration shown in Figure 8 is selected to result in these aspect ratios,
where a five-story shear wall of dimensions 5 m long by 15 m high is supported on a
strip foundation of 8 m by 2.5 m. The design vertical factor of safety, FSV, for this foot-
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 677

Figure 7. Graphical representation of BNWF model and associated parameters for studying
SSI.

ing is selected as 4.0. A BNWF model is created to model this elastic shear wall resting
on a nonlinear soil, such that only the influence of soil compliance 共both uplift and soil
nonlinearity兲 is considered.
The soil medium is assumed to be sand at 90% relative compaction, which corre-
sponds to a relative density of 75% 共after Lee and Singh, 1971兲 and ␾⬘ = 33°. The soil
and modeling properties used are provided in Table 4. Assuming the shear wall supports
a tributary area of approximately 50 m2, with a distributed dead and live load of 100 kN

Figure 8. Dimensions of 5-story shear wall/shallow foundation system selected for design
example.
678 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

Table 4. BNWF parameters used for design example

SSI property Value Reference

Effective shear modulus ratio G / Go = 0.478 FEMA-356 共ASCE 2000兲

Shear wave velocity ␯s = 150 m / s Das 共1999兲
Initial shear modulus Go = 18100 kPa FEMA-356 共ASCE 2000兲
Global vertical stiffness Kz = 300 MN/ m Gazetas 共1991兲
Global rotational stiffness K␪ = 3200 MN-m Gazetas 共1991兲
End length ratio Le / L = 25% Harden et al. 共2005兲
Middle region subgrade reaction kmid = 15 MN/ m Harden et al. 共2005兲
End region subgrade reaction kend = 85 MN/ m Harden et al. 共2005兲
End tip resistance qi= 70% Harden et al. 共2005兲
Allowable bearing qa= 275 kPA FEMA-356 共ASCE 2000兲

共DL= LL= 50 kN兲 per floor 共total DL+ LL= 5*100 kN= 500 kN兲, in addition to the self-
weight of the wall and footing equal to 900 kN, the total vertical load is equal to
1400 kN, or approximately 3% of the compressive strength of the concrete wall.
Five simplified design methods are used to evaluate the displacement demand of the
system; the capacity spectrum approach 共termed Analysis Type 1兲, a method that incor-
porates rocking based on Housner’s 共1963兲 approach 共Analysis Type 2兲, the nonlinear
static procedure with conventional C1 values 共Analysis Type 3兲, the nonlinear static pro-
cedure with C1 values modified per this study 共Analysis Type 4兲, and the time history
method 共Analysis Type 5兲. Comparison of these design methods for estimating the maxi-
mum displacement demand is conducted. Each method is evaluated for the cases of an
inelastic subgrade allowed to uplift, an elastic subgrade allowed to uplift and an elastic
subgrade fixed against uplift, as well as the fully fixed-base case 共i.e., no consideration
for soil structure interaction兲. Note that the case of an elastic subgrade fixed against up-
lift is identical to a model with a single elastic rotational and vertical spring; this type of
analysis is recommended in several design codes and is termed method 1, after FEMA-
356 共2000兲. For the four analytical models used to evaluate the five simplified analysis
methods, Table 5 lists the natural period of each model from initial stiffness of the vari-
ous pushover curves. The pushover curves were developed from the model as shown in
Figure 7, incorporating the parameters as listed in Table 4 and the soil structure inter-
action constraints as mentioned above 共i.e., elastic or nonlinear, fixed or allowed to up-
lift, etc.兲. Note that the system becomes more flexible when elastic soil springs are
added, and even more flexible when nonlinear behavior is introduced into the springs.

Table 5. Natural period of model for simplified analysis methods

SSI Included 共uplift may occur兲 SSI Not Included

Natural Period, Nonlinear Soil Springs Elastic Soil Springs Fixed Base
T
0.56 0.42 0.03
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 679

Figure 9. Normalized base shear versus normalized top displacement for the various models
used to evaluate the simplified analysis methods.

This softening effect can also be seen in the normalized pushover curves of the various
models, Figure 9. Note that the elastic model and the elastic model with uplift allowed
are identical until uplift occurs.
The fixed modification factors 共Equation 1兲 for the structure described, for use with
the analysis are selected as C0 = 1.4, C2 = 1.0, and C3 = 1.0. Justification for these values
is as follows. The modification factor to relate spectral displacement of an equivalent
SDOF system to the roof displacement of the building MDOF system is defined in
FEMA-356 共2000兲 equal to 1.4 for a building of five stories with any load pattern. The
BNWF model reasonably displays the hysteresis and strength degredation observed in
model test data through empirically calibrated nonlinear foundation elements briefly de-
scribed above; therefore the value of C2 which ordinarily would account for these effects
in a linear model is set to one. Additionally, the value of C3 is taken as one because P-
delta effects are directly accounted for in this design example. The site-specific demand
response spectrum for a site in Southern California is created after FEMA-356 共Section
1.6.1.5兲, based on a 10% probability of exceedance in 50 years. The characteristic period
TS of this spectrum is 0.37 sec. Finally, an effective mass factor Cm, to account for
higher mode participation, is required when calculating base shear per Equation 3-10 of
FEMA-356. The effective mass factor Cm is taken as 0.8, as defined in FEMA-356 for a
concrete shear-wall system with three or more stories.
The five analysis types are described in the following section, with values explicitly
described for the BNWF as applicable.

CAPACITY SPECTRUM APPROACH „CSA…—ANALYSIS TYPE 1

A pushover curve is developed using the BNWF model with the values discussed,
and is converted to an SDOF capacity curve, using a modal mass coefficient Cm and the
680 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

Figure 10. Analysis Type I: peak displacement of substitute structure, using capacity spectrum
approach and the nonlinear soil model.

participation factor C0. The substitute structure Capacity Spectrum is therefore defined
HPO 1 1
by SAsub = W Cm g and ⌬sub = ⌬top C0 , where HPO is the base shear. Because the design re-
sponse spectrum is not reduced for system nonlinearities, the peak displacement is given
by the intersection of the design capacity spectrum and the substitute structure capacity
spectrum, as shown in Figure 10, considering a system with nonlinear and uplifting soil
behavior. This is found to be ⌬sub ⬵ 200 mm for the model with nonlinear soil springs.
The design displacement is then converted to the actual structure displacement by ⌬top
= ⌬subC0, which in this case is 280 mm for the BNWF substitute structure.
For this analysis exercise, the estimate of ⌬top is not based on iterating the percent of
critical damping ␰ 共Tsecant兲 as suggested in other documents. FEMA-356 does not pro-
vide critical damping curves, accounting for a rocking foundation in the Capacity Spec-
trum Approach. For this analysis the percent of critical damping is assumed to be 5%.

HOUSNER ROCKING BLOCK APPROACH—ANALYSIS TYPE 2

An alternative simplified method to estimate the peak displacement of a rocking
shallow foundation is based on the application of Housner’s 共1963兲 approach, with slight
modifications. For a rocking block as shown in Figure 11, the effective viscous damping
is a function of the block dimensions, the seismic mass, and the mass moment of inertia.
In addition, the rocking will be sensitive to the vertical factor of safety, since this affects
the point of rotation through which the block rocks, based on an assumed contact area.
Considering the demand displacement of a substitute structure ⌬sub,the rocking pe-
riod of the structure may be calculated as
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 681

Figure 11. Housner’s 共1963兲 rocking block 共after FEMA-356兲.

T=
冑
4
WR
IO
cosh−1
冉 冊
1−
1
␪
␣
, where ␪ =
⌬sub
R cos共␣兲
共12兲

Additionally, R and ␣ are functions of the block geometry shown in Figure 11, ␪ is
the block rotation, and IO is the mass moment of inertia about one corner. The design
displacement may be found through iteration of an assumed displacement, or 共perhaps
the more straightforward approach兲 at the intersection of a pushover type curve and the
design spectrum.
The effective viscous damping 共in %兲 of the rocking block may be calculated as

␰eff = 40共1 − 冑r兲 共13兲

where r is calculated as

冋
r= 1−
mR2
IO
共1 − cos共2␣兲兲 册 2
共14兲

For the shear wall considered, the effective viscous damping is calculated to be 10%,
therefore ⌬sub ⬵ 147 mm. The design displacement is then converted to the actual struc-
ture displacement by ⌬top = ⌬subC0, which for the BNWF model is 206 mm.

NONLINEAR STATIC PROCEDURE „NSP…—ANALYSIS TYPE 3

The nonlinear static procedure is investigated, which estimates the demand displace-
ment using Equation 1. From Equation 2, the demand spectral acceleration of the elastic
682 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

system is 0.9 g 共at the initial period兲, and the yield force from a bilinear approximation
of the pushover curve for the system allowed to uplift is 197 kN. Using these variables,
the strength ratio R was found to be 4.0 for the BNWF model; R
= 0.9g*1400kN / 共197kN*0.8兲 = 4.0. The effective period, 0.56 sec, is then calculated
through the elastic perfectly-plastic transition of a bilinear approximation to the push-
over curve, similar to that described in FEMA-356 Figure 3-1 with a zero post-yield
stiffness. With R and the effective period determined, the design displacement is calcu-
lated for the conventional definition of C1. For this case, C1 defined by FEMA-356 is
equal to one 共Equation 1兲. The design displacement is then calculated directly from
Equation 1 as ⌬top ⬵ 91 mm.

NONLINEAR STATIC PROCEDURE „NSP… WITH MODIFIED C1 VALUES—

ANALYSIS TYPE 4
The nonlinear static procedure is further explored with the modified value of C1 cal-
culated by the procedure presented in this paper. As discussed above, the strength ratio R
was found to be 4.0 for the BNWF model, and an effective period of 0.56 sec. The de-
mand spectrum in this example has a characteristic period of 0.37 seconds. The ratio of
the elastic period to the characteristic period is 0.56/ 0.37= 1.5. Estimating the force de-
mand from the intersection of the demand spectrum with both the elastic and nonlinear
models, the spectral acceleration ratio is calculated as ␺SA = SAnonlinear / SAelastic = 0.81.
Given that R is equal to 4.0, X1 = 0.68 and X2 = 0.87 from Table 2. From Table 3, the
system falls into group II under R = 4.0 because X1 ⱕ ␺SA ⬍ X2, therefore A = −0.881
and B = 0.12. Finally, C1 defined by an elastic shallow foundation allowed to uplift is
calculated from Equation 11, C1 = 1 / 关1 − 0.881e−0.12⫻共1.5−0.5兲兴 = 4.6; nearly five times
what would be recommended in conventional design codes. The design displacement
from Analysis type 3 was calculated to be ⌬top ⬵ 91 mm. Incorporating the modified C1
value in place of the original C1 value gives a design displacement of ⌬top
⬵ 91 mm共C1 / C1−FEMA兲 = 91 mm共4.6/ 1.0兲 ⬵ 414 mm.

TIME HISTORY „TH… METHOD—ANALYSIS TYPE 5

Conducting a nonlinear time history analysis should provide for the most realistic
estimation of the seismic demands. Using recorded ground motions to assess structural
demands, here, three ground motions are selected from the suite of 19 ground motions
used to investigate the relation between C1 and R 共See Table 1兲. The longitudinal com-
ponents of two ground motions from the 1994 Northridge earthquake 共Woodland Hills
and Van Nuys, Sherman Circle denoted “NR-whox” and “NR-vnsc,” respectively兲 and
the longitudinal component of one ground motion from the 1971 San Fernando earth-
quake 共Glendale, Muni Building, denoted “SF-glen”兲 are selected. The longitudinal
components of the selected ground motions are assumed as input parallel to the wall in
this design example. The characteristics of these motions best represent the design spec-
trum for three binned groups of short, medium, and long characteristic periods. Figure
12a shows the elastic acceleration response spectra for these motions along with the de-
sign spectrum. It is recommended to use at least three representative time histories with
accelerations scaled such that the average spectral acceleration is at least 1.4 times the
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 683

Figure 12. 共a兲 Demand and unscaled ground motion acceleration spectrum at 5% damping, and
共b兲 demand and scaled design ground motion acceleration spectrum at 5% damping.

design spectrum. Accelerations within the range of 0.2 to 1.5 times the natural period are
used to calculate the average spectral acceleration. Figure 12b shows the three spectra
scaled using the aforementioned criteria along with the design spectrum. Figure 13
shows the top displacement and settlement time history for the largest PGA ground mo-
tion, “NR-vnsc.” This ground motion’s characteristic period of 0.52 seconds is closest to
the BNWF structure period of 0.56 seconds. Top displacements of 385 mm and 102 mm
are calculated using these scaled time histories and the Winkler model for the cases of
nonlinear and elastic soil, respectively.

DESIGN EXAMPLE RESULTS AND DISCUSSION

Displacement demands predicted for the three model types and four analysis meth-
ods are summarized in Table 6. For all analysis methods, the CSA, Housner’s model, and
NSP with C1 conventional definition are less conservative than either the time history
method or the NSP with modified C1 values. The NSP with C1 based on an uplifting
elastic foundation results in a conservative estimate when compared with the time his-
tory method. This is positive in the sense that using the C1 values from this study allows
684 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

Figure 13. 共a兲 Story drift and 共b兲 settlement time history for nonlinear soil springs case.

a design engineer to include the effects of foundation uplift with current code simplified
design methods without performing an overly rigorous analysis, while still resulting in a
conservative demand estimate.
With respect to base shear demand, the NSP with C1 conventional definition, NSP
with modified C1 values, and time history method with a nonlinear subgrade resulted in
the lowest base shear estimates 共Table 7兲. This result is to be expected given the larger
displacements observed for these cases, and is a significant benefit of allowing perma-
nent 共where nonlinear springs are used兲 and transient displacements of the foundation.

Table 6. Design displacement values for the various analysis methods

Soil Structure Interaction 共SSI兲

Included Not Included
Uplift Allowed — Fixed Base
Nonlinear Soil Elastic Soil
Springs Springs Method 1
Analysis Method Type ⌬top 共mm兲 ⌬top 共mm兲 ⌬top 共mm兲 ⌬top 共mm兲
TYPE 1—CSA 280 178 38 0.14
TYPE 2—Housner 206 202 202 NA
TYPE 3—NSP 共conventional C1兲 91 57 57 0.27
TYPE 4—NSP 共modified C1兲 414 308 62 0.27
TYPE 5—TH 385 102 237 0.0003
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 685

Table 7. Design base shear demand for the various analysis methods

Soil Structure Interaction 共SSI兲

Included Not Included
Uplift Allowed — Fixed Base
Nonlinear Soil Elastic Soil
Springs Springs Method 1
Analysis Method Type V 共kN兲 V 共kN兲 V 共kN兲 V 共kN兲
TYPE 1—CSA 211 331 1181 728
TYPE 2—Housner 284 291 291 NA
TYPE 3—NSP 共conventional C1兲 162 240 1024 793
TYPE 4—NSP 共modified C1兲 170 265 1108 793
TYPE 5—TH 169 251 4245 1341

The fixed-base 共no soil springs兲 analysis provides a valuable contrast to the analyses in-
corporating some form of soil structure interaction. The design shear is significantly
larger than those cases where uplift is allowed, and the target displacements are much
smaller—both prohibitive to economic design.
Another useful comparison can be made to the Uniform Building Code 共UBC兲, 1997
edition, which allows for a direct calculation of the base shear and maximum displace-
ment, independent of soil structure interaction. Note that a similar comparison can be
made with the International Building Code 共IBC兲, 2000 edition. Assuming worst-case
conditions 共a fault proximity of less than 2 km and soil type “D”兲 in a Zone 4 location,
the UBC Static Force Procedure 共Chapter 16兲 gives a base shear of 514 kN. This is ap-
proximately twice the value found when using the simplified analysis methods and with
elastic soil springs 共recall that the UBC is generally understood to be a conservative de-
sign document兲. Based on the height of the building and lateral force–resisting system,
the maximum elastic deflection at the top of the building is calculated as 119 mm, while
the maximum inelastic deflection is calculated as 375 mm. Note that for both elastic and
nonlinear spring models, all of the analysis methods except the NSP with C1 conven-
tional definition allow for similar or greater displacements than the UBC results 共a con-
servative baseline兲. This highlights to a greater extent the need to account for soil struc-
ture interaction in the nonlinear static procedure.

DESIGN EXAMPLE SETTLEMENT ESTIMATION

The displacement of the system evaluated by any of the simplified methods allows
one to estimate the settlement using empirical energy dissipation data and the calculated
number of cycles from an initial displacement using Housner’s 共1963兲 model.
Housner 共1963兲 gives the angle of rotation of a foundation after the nth impact, due
to an initial rotation,␪o, as
686 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

␪n = 1 − 冑1 − rn共1 − 共1 − ␪o兲2兲 共15兲

The initial rotation ␪o is estimated from the results for each of the simplified design
methods using the top lateral design displacement demand 共Table 6兲 or using the direct
values from the time history analysis 共for Method 5兲, and where r was previously de-
fined in Equation 14.
Using a suite of experimental data, Harden et al. 共2005兲 synthesize results in terms
of equivalent viscous damping and distortion level. A sample of these results, for stati-
cally moment-loaded footings is shown in Figure 14. Distortion here is defined as the
settlement normalized by the foundation length per cycle. These data fall both below and
above typical foundation design distortion values, as suggested by Duncan and Buchi-
gnon 共1987兲, for steel frame and load-bearing wall structures. Note, however, that a de-
pendency on the static factor of safety FSV is observed, which is not provided in design
recommendations. Figure 15 shows the magnitude of equivalent viscous damping versus
the half amplitude of rotation 共i.e., not averaged over cycles of loading兲 of a footing per
cycle. This information will be used to estimate the amount of equivalent viscous damp-
ing mobilized during foundation settlement due to the demand lateral displacement es-
timated from the simplified design methods.
Nonlinear regression through the data shown in Figure 14 results in equivalent vis-
cous damping versus distortion per cycle, for FSV ⱖ 3 as
ln共␰兲−4.451
s=e 0.248 共16兲
Similarly, incorporating the results of shallow foundation model tests shown in Figure
15, the relationship for equivalent viscous damping versus half-amplitude of rotation is
found through regression for FSV ⱖ 3 as

␰eq = e0.336 ln共␪兲+3.137 共17兲

Using Equations 15–17, an iterative approach can be applied until the number of
cycles 共with two impacts per cycle兲 forces the rotation amplitude to zero 共at rest兲. The
procedure begins with the initial displacement demand calculated by any of the simpli-
fied methods previously discussed. The last step of the procedure is to sum the settle-
ments calculated for all cycles. The accumulated settlement values are listed in Table 8,
calculated using the maximum estimated top displacement values from the simplified
procedures.
The accumulated settlement values are sensitive to the starting value of top displace-
ment. Larger initial horizontal displacements accrue larger settlements. This is an intui-
tive result; as a higher level of performance is demanded of the system, a larger cost is
accrued in settlement. Similarly, using the nonlinear static procedure with values of C1
calculated from this study gives a conservative result with respect to the CSA, the Hous-
ner model, and the traditional NSP, though much more comparable to the time history
method. Recall from Figure 13b the maximum settlement from the largest PGA of the
three motions considered using the time history method, was 80 mm. Most notably, the
NSP method, using conventional C1 values grossly underpredicts the accumulated settle-
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 687

Figure 14. Equivalent viscous damping versus maximum footing distortion 共settlement normal-
ized by footing length per cycle for slow cyclic tests: 共a兲 footings on sand and 共b兲 footings on
clay. Experimental data sets courtesy of Bartlett 共1976兲, Gajan et al. 共2003兲, Negro et al. 共1998兲,
Rosebrook and Kutter 共2001兲, and Wiessing 共1979兲.

ment when compared to the time history method, which is approximately seven times the
Type 3 method. The simplified settlement analysis using the initial displacement from
the modified NSP gives a value close to the settlement estimated using the initial dis-
688 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

Figure 15. Equivalent viscous damping versus half amplitude of rotation for slow cyclic tests:
共a兲 footings on sand and 共b兲 footings on clay. Experimental data sets courtesy of Bartlett 共1976兲,
Gajan et al. 共2003兲, Negro et al. 共1998兲, Rosebrook and Kutter 共2001兲, and Wiessing 共1979兲.

placement from the time history method, within 20% on the conservative side. These
reasonably close values add credibility to the proposed empirical method of estimating
settlement, combined with Housner’s rocking block approach, and using the experimen-
tally regressed ␰eq–distortion curves.
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 689

Table 8. Calculated accumulated permanent settlement values

Total Distortion
Analysis Method Type ⌬top 共mm兲 s 共mm兲 共s / L兲 共%兲
TYPE 1a—CSA 280 56 0.7
TYPE 2a—Housner 206 38 0.5
TYPE 3a—NSP 共conventional C1兲 91 12 0.2
TYPE 4a—NSP 共modified C1兲 414 96 1.2
TYPE 5b—TH 385 86 1.1
a
Estimated based on empirical simplified settlement analysis method
b
Accrued settlement from BNWF model 共time history analysis兲 with nonlinear soil springs

CONCLUSIONS
The relationship between the strength ratio R and the displacement ratio C1 was in-
vestigated in this paper for shallow foundations allowed to uplift. Compared to current
design codes, the displacement ratio was found to be significantly larger in the accelera-
tion sensitive range of the response spectrum 共T / TS ⱕ 0.5兲, indicating that rocking shal-
low foundations may be a case where special attention is required when using simplified
design procedures. Suggestions for improved C1-R relations are provided, based on re-
gression through the nonlinear time history data collected in this study.
To illustrate the approach for using the improved C1-R relation, a design example
was presented considering five different methods of analysis. Using the C1 value from
the uplift study provided for a relatively close estimate of peak displacement demand
when compared to the time history method. Both methods are conservative with respect
to the displacements calculated from the LSP, rocking block analogy, and CSA proce-
dures commonly used in design codes. A procedure to estimate settlement based on em-
pirical rocking foundation data was presented and used for each of the simplified design
methods. The simplified settlement estimate using an initial displacement from the
modified NSP procedure compared well with the accrued settlement calculated from
time history analysis using a Winkler foundation with nonlinear soil springs allowed to
permanently deform.

ACKNOWLEDGEMENTS
Support of this work was provided by the Pacific Earthquake Engineering Research
Center’s Program of the National Science Foundation under Award Number EEC-
9701568 and PEER project number 2272001.2. The above support is greatly appreci-
ated. Helpful suggestions of Professors Geoff Martin and Bruce Kutter are greatly ap-
preciated. Experimental data used to construct the plots in Figure 14 and 15 were
provided by Prof. Kutter 共University of California, Davis兲, Prof. Negro 共ELSA兲, and
digitized from the M.S. work conducted by Wiessing 共1979兲 and Bartlett 共1976兲, which
was conducted under the guidance of Prof. Taylor at the University of Auckland, New
Zealand. These experimental data sets are greatly appreciated.
690 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

REFERENCES
Allotey, N., and Naggar, M., 2003. Analytical moment-rotation curves for rigid foundations
based on a Winkler model, Earthquake Eng. Struct. Dyn. 23, 367–381.
Al-Sulaimani, G. J., and Roessett, J. M., 1984. Design spectra for degrading systems, J. Struct.
Eng. 111 共12兲, 2611–2623.
American Society of Civil Engineers 共ASCE, 2000. Prestandard and Commentary for the Seis-
mic Rehabilitation of Buildings, prepared for the Federal Emergency Management Agency,
FEMA-356, Washington, D.C.
Applied Technology Council 共ATC, 1996. Seismic Evaluation and Retrofit of Concrete Build-
ings 共ATC-40兲, Volumes 1 and 2, Redwood City, CA.
Bartlett, P. E., 1976. Foundation Rocking on a Clay Soil, M.E. thesis, University of Auckland,
New Zealand.
Boulanger, R. W., Curras, C. J., Kutter, B. L., Wilson, D. W., and Abghari, A., 1999. Seismic
soil-pile structure interaction experiments and analyses, J. Geotech. Geoenviron. Eng. 125
共9兲, 0750–0759.
Chopra, A. K., and Goel, R. K., 1999. Capacity-Demand-Diagram Methods for Estimating
Seismic Deformation of Inelastic Structures: SDF Systems, Report PEER 1999/02, Pacific
Earthquake Engineering Research Center, University of California, Berkeley.
Cuesta, I., and Aschheim, M. A., 2000. Waveform independence of R factors, Proceedings, 12th
World Conf. on Earthquake Eng., Auckland, New Zealand, Paper No. 1246.
Cuesta, I., and Aschheim, M. A., 2001a. Using Pulse R-Factors to Estimate Structural Response
to Earthquake Ground Motions, MAI Center Report Series CD release 01-03, University of
Illinois at Urbana-Champaign, IL, March.
———, 2001b. Isoductile strengths and strength reduction factors of elasto-plastic SDOF sys-
tems subjected to simple waveforms, Earthquake Eng. Struct. Dyn. 30 共7兲, 1043–1059.
———, 2001c. Inelastic response spectra using conventional and pulse R-factors, J. Struct.
Eng. 127 共9兲, 1013–1020.
Cuesta, I., Aschheim, M., and Fajfar, P., 2003. Simplified R-factor relationships for strong
ground motions, Earthquake Spectra 19 共1兲, 25–45.
Das, B. M., 1999. Principles of Foundation Engineering, Brooks Cole Publishing Company,
Pacific Grove, CA.
Elghadamsi, F. E., and Mohraz, B., 1987. Inelastic earthquake spectra, Earthquake Eng. Struct.
Dyn. 15, 91–94.
Fajfar, P., 2000. A nonlinear analysis method for performance-based seismic design motions,
Earthquake Spectra 16 共3兲, 573–592.
Gajan, S., Phalen, J. D., and Kutter, B. L., 2003. Soil-Foundation Structure Interaction: Shallow
Foundations, Centrifuge Data Report for the SSG02/03 Test Series, Center for Geotechnical
Modeling Data Reports UCD/CGMDR-03/01 and 02, University of California, Davis.
Gazetas, G., 1991. Foundation Engineering Handbook, edited by H. Y. Fang, Van Nostrand Re-
inhold, New York, 40 pp.
Harden, C. W., 2003. Numerical Modeling of the Nonlinear Cyclic Response of Shallow Foun-
dations, M.S. thesis, University of California, Irvine.
INVESTIGATION OF EFFECTS OF FOUNDATION UPLIFT ON SIMPLIFIED SEISMIC DESIGN PROCEDURES 691

Harden, C. W., Hutchinson, T. C., Martin, G. R., and Kutter, B. L., 2005. Numerical Modeling
of the Nonlinear Cyclic Response of Shallow Foundations, Report PEER 2005/04, Pacific
Earthquake Engineering Research Center, University of California, Berkeley.
Hidalgo, P. A., and Arias, A., 1990. New Chilean code for earthquake-resistant design of build-
ings, Proceedings, 4th U.S. Nat. Conf. Earthquake Engrg, Palm Springs, Calif., Vol. 2, pp.
927–936.
Housner, G. W., 1963. The behavior of inverted pendulum structures during earthquakes, Bull.
Seismol. Soc. Am. 53 共2兲, 403–417.
Lai, S., and Biggs, J., 1980. Inelastic response spectra for aseismic building design, J. Struct.
Div. ASCE 106 共ST6兲, 1295–1310.
Lee, P. Y., and Singh, A., 1971. Relative density and relative compaction, J. Soil Mech. Found.
Div. ASCE 97 共SM7兲, 1049–1052.
Makris, N., and Konstantinidis, D., 2003. The rocking spectrum and the limitations of practical
design methodologies, Earthquake Eng. Struct. Dyn. 32, 265–289.
Miranda, E., 1993. Site-dependent strength reduction factors, J. Struct. Eng. 119 共12兲, 3503–
3519.
Miranda, E., and Bertero, V. V., 1994. Evaluation of strength reduction factors for earthquake-
resistant design, Earthquake Spectra 10 共2兲, 357–379.
Nassar, A. A., and Krawinkler, H., 1991. Seismic Demands for SDOF and MDOF Systems, Re-
port No. 95, The John A. Blume Earthquake Engineering Center, Stanford University, Stan-
ford, CA.
Negro, P., Verzeletti, G., Molina, J., Pedretti, S., Lo Presti, D., and Pedroni, S., 1998. Large-
Scale Geotechnical Experiments on Soil-Foundation Interaction 共TRISEE Task 3兲, European
Commission, Joint Research Center, Special Publication No. I.98.73.
Newmark, N., and Hall, W., 1973. Seismic Design Criteria for Nuclear Reactor Facilities, Re-
port No. 46, Practices for Disaster Mitigation, National Bureau of Standards, U.S. Depart-
ment of Commerce, pp. 209–236.
Newmark, N., and Hall, W., 1982. Earthquake Spectra and Design, EERI Monograph Series,
Earthquake Engineering Research Institute, Oakland, CA.
OpenSees 共Open System for Earthquake Engineering Simulation platform, developed by the
Pacific Earthquake Engineering Research Center 共PEER兲, at the University of California,
Berkeley. http://opensees.berkeley.edu/
Ordaz, M., and Pérez-Rocha, L. E., 1998. Estimation of strength-reduction factors for elasto-
plastic systems: A new approach, Earthquake Eng. Struct. Dyn. 27, 889–901.
Peng, M.-H., Elghadamsi, F. E., and Mohraz, B., 1988. A stochastic procedure for nonlinear
response spectra, Proceedings, 9th World Conf. on Earthquake Engrg., Tokyo-Kyoto, Japan,
Vol. V, pp. 1069–1074.
Psycharis, I. N., 1981. Dynamic Behavior of Rocking Structures Allowed to Uplift, Ph.D. the-
sis, California Institute of Technology, Pasadena, CA.
Psycharis, I. N., 1983. Dynamics of flexible systems with partial lift-off, Earthquake Eng.
Struct. Dyn. 11, 501–521.
Riddell, R., 1995. Inelastic design spectra accounting for soil conditions, Earthquake Eng.
Struct. Dyn. 24, 1491–1510.
Riddell, R., Hidalgo, P., and Cruz, E., 1989. Response modification factors for earthquake re-
sistant design of short period structures, Earthquake Spectra 5 共3兲, 571–590.
692 C. HARDEN, T. C. HUTCHINSON, AND M. MOORE

Riddell, R., and Newmark, N., 1979. Statistical Analysis of the Response of Nonlinear Systems
Subjected to Earthquakes, Structural Research Series No. 468, Dept. of Civ. Engrg., Univer-
sity of Illinois, Urbana.
Rosebrook, K. R., and Kutter, B. L., 2001. Soil-Foundation Structure Interaction: Shallow
Foundations, Centrifuge Data Report for the KRR01/02/03 Test Series, Center for Geotech-
nical Modeling Data Reports UCD/CGMDR-01/01, 02, and 03.
Somerville, P., and Collins, N., 2002. Ground Motion Time Histories for the Van Nuys Build-
ing, prepared for the PEER Methodology Testbeds Project. http://www.peertestbeds.net/
van%20nuys.htm
Takada, T., Hwang, H.H.M., and Shinozuka, M., 1988. Response modification factor for
multiple-degree-of-freedom systems, Proceedings, 9th World Conf. on Earthquake Engrg.,
Tokyo-Kyoto, Japan, Vol. V, pp. 129–134.
Veletsos, A. S., 1969. Maximum deformations of certain nonlinear systems, Proceedings, 4th
World Conf. Earthquake Eng., Santiago Chile, Vol. 2, pp. 155–170.
Veletsos, A. S., and Newmark, N. M., 1960. Effects of inelastic behavior on the response of
simple systems to earthquake ground motions, Proceedings, 2nd World Conf. Earthquake
Engrg., Japan, Vol. II, pp. 895–912.
———, 1964. Design Procedures for Shock Isolation Systems of Underground Protective
Structures, Air Force Weapons Laboratory, New Mexico, Technical Documentary Report No.
RTD TDR-63-3096, III.
Veletsos, A. S., Newmark, N. M., and Chelapati, C. V., 1965. Deformation spectra for elastic
and elastoplastic systems subjected to ground shock and earthquake motions, Proceedings,
3rd World Conf. Earthquake Engrg., Wellington, New Zealand, Vol. 2, pp. 663–680.
Veletsos, A. S., and Vann, W. P., 1971. Response of ground-excited elasto-plastic systems, J.
Struct. Div. ASCE 97 共ST4兲, 1257–1281.
Vidic, T., Fajfar, P., and Fischinger, M., 1992. A procedure for determining consistent inelastic
design spectra, Proceedings of Workshop on Nonlinear Seismic Analysis of RC Structures,
Bled, Slovenia.
Wiessing, P. R., 1979. Foundation Rocking on Sand, School of Engineering Report No. 203,
University of Auckland, New Zealand.
共Received 28 December 2004; accepted 5 October 2005兲