Vous êtes sur la page 1sur 16


discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/239357967

Analytical momentrotation curves for rigid

foundations based on a Winkler model
Impact Factor: 1.22 DOI: 10.1016/S0267-7261(03)00034-4





M.Hesahm El Naggar
The University of Western Ontarioprofessor a

All in-text references underlined in blue are linked to publications on ResearchGate,

letting you access and read them immediately.

Available from: M.Hesahm El Naggar

Retrieved on: 06 February 2016

Soil Dynamics and Earthquake Engineering 23 (2003) 367381


Analytical moment rotation curves for rigid foundations

based on a Winkler model
Nii Allotey, M. Hesham El Naggar*
Department of Civil and Environmental Engineering, Faculty of Engineering, Geotechnical Research Centre,
The University of Western Ontario, London, Ont., Canada N6A 5B9
Accepted 6 January 2003

Analytical equations for the moment rotation response of a rigid foundation on a Winkler soil model are presented. An equation is derived
for the uplift-yield condition and is combined with equations for uplift- and yield-only conditions to enable the definition of the entire static
moment rotation response. The results obtained from the developed model show that the inverse of the factor of safety, x; has a significant
effect on the moment rotation curve. The value of x 0:5 not only determines whether uplift or yield occurs first but also defines the
condition of the maximum moment rotation response of the footing. A Winkler model is developed based on the derived equations and is
used to analyze the TRISEE experiments. The computed moment rotation response agrees well with the experimental results when the
subgrade modulus is estimated using the unload reload stiffness from static plate load deformation tests. A comparison with the
recommended NEHRP guidelines based on the FEMA 273/274 documents shows that the choice of value of the effective shear modulus
significantly affected the comparison.
q 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Subgrade modulus; Foundation uplift; Soil yield; Momentrotation; Unloadreload stiffness; Winkler model; Bearing capacity; Backbone curve;

1. Introduction
The foundation rocking behavior could greatly contribute to the response of the supported structure to seismic
loading, and in some cases it may become the governing
factor when choosing a retrofitting scheme [1]. In the past,
seismic provisions in most codes accounted approximately
for the effects of foundation behavior on the structural
response (usually referred to as soil structure interaction
(SSI) effects) by adjusting the fundamental period and
damping ratio of the structure. The implementation of the
performance-based seismic design approach requires simple
and efficient cyclic load deformation models for different
structural elements [2]. Thus, the simplified approaches
used in older codes to model SSI effects are not appropriate,
and it is necessary to develop foundation models that can
capture the most important characteristics of the foundation
cyclic load deformation behavior. Therefore, new seismic
* Corresponding author. Tel.: 1-519-661-4219; fax: 1-519-661-3942.
E-mail address: naggar@uwo.ca (M. H. El Naggar).

design guidelines such as FEMA 273/274 [3] and ATC 40

[4] require explicit modeling of foundation elements when
determining both linear and nonlinear responses of
Foundation rocking contributes significantly to the
seismic response of a foundation for both tall slender
structures and medium-rise buildings [5]. The rocking mode
involves uplift of the foundation at one side and soil
yielding at the other side of the foundation, and generally
results in the permanent settlement of the footing. Many
researchers have investigated the nonlinear foundation
rocking action using rigorous finite element and boundary
element models [6 8]. However, finite element and
boundary element solutions are not efficient for nonlinear
time domain analysis since they require large computational
time and effort, and thus are not practical for regular design
The Winkler model is widely used in SSI analysis because
of its simplicity and ability to incorporate different nonlinear
aspects of the behavior at a reduced computational effort
compared to other approaches. The use of the Winkler model

0267-7261/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved.


N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

a set of equations to evaluate the moment rotation response

for both uplift- and yield-only conditions of a rigid
foundation on a Winkler soil model. The equations were
derived in the context of a retaining wall foundation, and as
such, they do not constitute all the necessary equations for the
complete static response of a rigid foundation. Siddharthan et
al. [13,14] assumed that the ultimate moment occurs when
uplift commences after soil yield has occurred. Their
assumption may represent an acceptable approximation for
the ultimate moment capacity of the foundation under certain
conditions but may not be valid under all conditions.

2. Objectives and scope of work

Fig. 1. Schematic of Winkler soil model.

has been extended to dynamic SSI applications by introducing the Beam-on-Nonlinear Winkler Foundation (BNWF)
models [9]. Filiatrault et al. [10] and Chaallal and Ghlamallah
[11] have used Winkler models to account for foundation
flexibility in their numerical analyses to study the effects of
SSI on both the linear and nonlinear responses of various
structures. A schematic for a rigid foundation on a Winkler
soil is shown in Fig. 1.
Bartlett [12] introduced a Winkler approach to model the
cyclic response of footings on clay. It was noted from the
results that foundation uplift occurs before soil yielding
when the static factor of safety (FS) is , 2. However, he
studied the response using a numerical approach and did not
provide any general equations to predict the response under
different footing conditions. The FEMA 273/274 guidelines
[3] for modeling foundations are based mainly on results of
Bartlett [12], which are depicted schematically in Fig. 2.
The figure shows the moment expressions for the two
extreme conditions for the soil underneath the foundation:
the ideal condition of an uplifted rigid footing supported on
elastic soil at only one corner; and the condition in which the
uplifted footing is supported by a fully developed plastic
block as a result of soil yielding. The moment expressions
corresponding to these two extreme conditions are easily
estimated from simple statics.
The backbone curve of the pseudo-static cyclic moment
rotation response forms an important part of the cyclic
response of a footing, and thus, has to be accurately modeled
when analyzing the seismic response of the supported
structure. Analytical solutions for the moment rotation
response of foundations are difficult to derive because of the
complex nonlinear foundation behavior. Therefore, either a
numerical technique is used for the analysis or simplifying
assumptions are introduced. Siddharthan et al. [13] presented

The objective of the current study is to provide a

complete analytical solution for the static moment rotation
response of a rigid foundation resting on a Winkler soil
model. The solution by Siddharthan et al. [13] is extended to
provide additional equations in order to completely define
the entire moment rotation response curve. The different
parameters governing the moment rotation response are
examined using the derived equations. The moment
rotation response curves computed using the derived
equations are compared with experimental results found in
the literature. The analytical results are also compared with
code provisions used to estimate the moment rotation
response in order to shed some light on their level of

3. Derivation of state equations

The following assumptions are made in the derivation
of the state equations: the axial load is constant and acts
at the center of the footing; the moment acts about the
longitudinal axis of the footing and is computed about its
center; and the length of the footing is one unit. Fig. 3
shows a schematic of the assumed stress and displacement
conditions for various footing states. State 1 represents
elastic conditions, state 2 represents the initial foundation
uplift condition (uplift-only), state 3 represents the initial
soil yield condition (yield-only) and state 4 represents the
soil yield and foundation uplift condition. These states
correspond to different segments of the moment rotation
curve shown in Fig. 2 and are considered herein to derive
the foundation moment rotation response curve as
3.1. Elastic condition
This stress state, represented by segment (1) in Fig. 2, is
widely used in practice for the design of footings.
Considering the kinematics for this state yields (Fig. 3a)

d0 d1x d2x


N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381


Fig. 2. Schematic of different states of momentrotation response (after FEMA 273/274).

This is

qx qp 2 k v u x 2


capacity of the footing under vertical load, qu : The onset of

this situation occurs when (from Eq. (1b))


where B is the width of the footing, x the distance from the

left end of the footing to spring i (in the Winkler model), qx
is the pressure at point x; d0 is the vertical distance between
the footing base center after loading and its original level
(GL) before loading, d1x and d2x are vertical distances given
by qx =kv and x 2 B=Lu; respectively, u the footing
rotation and kv the subgrade modulus. Finally, the initial
stress qp P=B; where P is the axial load applied to the
footing. Summing moments about the center of the footing
gives a linear relation between moment and rotation, i.e.

k v B3 u

where M is the moment acting on the footing. From Eq. (2),

the commencement of foundation uplift with no initiation of
soil yield (point 2 in Fig. 2) occurs when




k v B2

If soil yield occurs before foundation uplift, the maximum

stress at one end of the footing reaches the static bearing


q 2 qp
6 u



q 2 qp
kv B u


k v B k v B2


and can also be expressed as


qu B2 PB



Noting that these two limiting conditions occur on

opposite sides of the footing, it can be easily shown
that soil yielding would occur before foundation uplift
when [12]
qp $ u
3.2. Initial uplift condition
This stress state is represented by segment (3) in the
moment rotation curve shown in Fig. 2. The lower limit of
this segment begins from the point represented by Eq. (3).
Referring to Fig. 3b and considering the kinematics of this
condition, Eq. (6) can be obtained

d0 d1x d2x d2v



N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

Fig. 3. Stresses and displacements for different footing states.

where d2v B1 2 h 2 1=2u; and h is the ratio

of footing part not in contact with soil (length of
uplift) to the footing width. Substituting into Eq. (6a)
qx kv uB1 2 h 2 x


Calculating the moment at the center of the footing, the

following equation is obtained
M M2l 1 2h

from which the limiting value of M for an infinitely strong

soil is PB=2. The equation of this segment is given by [13]
M M2l 3 2 2
k v B2 u
M M2l 3 2 2
Eqs. (7) and (8) show that the moment, M; is a linear
function of the uplift ratio, h; and a nonlinear function of

N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

the rotation, u: When foundation uplift occurs first, soil

yield is initiated when the stress at the extreme edge
reaches the yield value, i.e. before M 3M2l ; which is the
ultimate condition for an infinitely strong soil. This
condition occurs when [13]
M3l 3M2l 2






3.3. Initial yield condition

Eq. (5) shows that soil yield occurs before foundation
uplift when the initial stress is greater than half the ultimate
bearing capacity of the footing. This condition is represented by segment (4) of the moment rotation curve in
Fig. 2 and its lower limit is defined by Eq. (4). Based on the
kinematics of this condition (Fig. 3c)

where d1u qu =kv ; d2u Bj 2 1=2u and j is the ratio of

the footing portion on yielded soil (yielded length) to the
footing width. Substituting into Eq. (10a) gives
qx qu kv uBj 2 x


Similar to the approach used in the initial uplift condition, it

can be shown that
M M2u 1 2j


Eq. (11) shows that the moment is a linear function of the

yielded length. The expression for this segment of the
moment rotation curve is given by [13]
M M2u 3 2 4
k v B3 u
M M2u 3 2 2 2u


q2u B

12kv M2u


d0 d1x d2x d1u d2u

M3u M2u 3 2
q u B2


which can also be expressed as

which can also be expressed as


(10b) and (11) as [14]




All previous work reported in the literature addresses

either elastic condition, uplift- and yield-only stress states. In
this study, the state of stress of combined soil yield and
foundation uplift is considered. An expression is derived to
describe this state represented by segment (5) of the
moment rotation curve shown in Fig. 2. This segment of
the moment rotation curve represents the foundation
response beyond the states defined by Eqs. (9a), (9b), (13a),
and (13b), regardless of which occurred first, uplift or yield.
Considering the kinematics of this stress state yields (Fig. 3d)

d0 d1x d2x d1u d2u d2v


Substituting into Eq. (14), the following equations are

qx kv uB1 2 h 2 x


qx qu kv uBj 2 x


Equating Eqs. (15a) and (15b) gives

h j 1 2 u
k v uB


Calculating forces and moments at the footing center gives

P q u Bj

k v u B2
1 2 h j2


q u B2
k u B3
1 2 h j2 21 2 2h 4j
jj 2 1 v

Substituting Eq. (16) into Eqs. (17a) and (17b) and

rearranging, the equation for the moment rotation curve
for this condition can be derived as

Eq. (12) shows that similar to the initial uplift condition, the
moment is inversely proportional to the square root of the
rotation. Assuming that progressive soil yield without
footing uplift continues until the ultimate condition is
reached (i.e. footing failure at which j 1), from Eq. (11),
M 3M2u : However, uplift would generally occur before
the condition of j 1 is reached. The onset of footing uplift
after initiation of soil yield can be obtained from Eqs. (10a),


3.4. Soil yield and foundation uplift condition



q B2
2 u

24kv u2


the limiting case for u ! 1 yields

lim M




Eq. (19) gives the maximum moment capacity of the

foundation, which can also be derived by considering the
foundation equilibrium (statics) when a fully plastic stress


N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

block is assumed. Eq. (18) shows that the moment, M; is

inversely proportional to the square of the rotation.
Substituting Eqs. (9a) and (13a) into Eq. (19) yields the
same expressions as Eqs. (9b) and (13b). This validates the
derived moment rotation expression for this stress state.

4. Properties of moment rotation curves

The equations derived above define the static moment
rotation response curve completely for any stress state.
These equations are used to evaluate the response of
different foundations under different loading conditions. To
enable a comparison between different footings under
different response conditions, nondimensional variables
(c; x; MqB ) are introduced as follows

kv B


qu B



qu B2


where c is a soil property and represents the ratio of the soil

stiffness to its strength; x is the inverse of the foundation
bearing capacity safety factor under vertical load, FS; and
MqB is a normalized (nondimensional) moment. Using these
nondimensional variables, the complete moment rotation
relation can be expressed as

For x # 12
< x
: x 2 x2 2
24c2 u2







and for x # 12
212 x
< 12 x
212 x 212 x
2c12 x
x 2 x2 2
2c12 x
24c u
Eqs. (21a) and (21b) show that the normalized moment,
MqB ; is a function of only c and x: Figs. 4 and 5 show
the moment rotation response curves for a range of values of
c and x:
Fig. 4 shows the moment rotation curves for x 0:2;
and a range of practical values of c (50 1200). Small
values of c (Fig. 4a) represent foundations supported on
strong soils such as stiff clays and dense sand, where the
soil strength is high compared to its stiffness. Such
foundations will usually have a small width. On the other
hand, large values of c (Fig. 4b) represent foundations of

Fig. 4. Computed momentrotation curves for x 0.2: (a) for small c; (b) large c:

N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381


Fig. 5. Computed momentrotation curves for different values of x for c 200:

large width supported on a relatively low bearing capacity

soil such as soft clay or loose sand. It is noted from Fig. 4
that the rotational stiffness of the foundation (manifested
by the slope of the moment rotation curve) increases with
an increase in c: The figure also reveals that the ultimate
rotation decreases as c increases, almost linearly, i.e. the
ultimate rotation decreased by an order of magnitude as c
increased by an order of magnitude. It is worth
mentioning here that Eurocode 7 [15] specifies 6 millirad
(mrad) as the relative rotation to cause an ultimate limit
state. From Fig. 4, it can be observed that rotation of
6 mrad represents an elastic response for the case of c
50 and represents a nonlinear response state for the case
of c 1200:
Fig. 5 shows the effect of x on the moment rotation
response of foundations with c 200: It can be clearly seen
from the figure that the moment response increases as x
increases until it reaches 0.5, and then declines as x
continues to increase. The insert in Fig. 5 shows that the
maximum value of MqB ; which does not depend upon c;
varies with x in a parabolic manner, and attains a maximum
value of 0.125 at x 0:5: This shows that x 0:5
represents a limiting condition on the moment rotation
response of a spread rigid footing based on the Winkler soil
5. Discussion
The FEMA 273/274 documents and Siddharthan et al.
[13] state that the significance of x is that its value, above

or below 0.5, indicates whether uplift of the foundation or

yielding of the soil would occur first. However, the main
significance of x 0:5 is that it defines the maximum
moment rotation response possible as shown in Fig. 5.
This is further illustrated in Fig. 6, which shows initiation
of different stress states for different x values. For x
0:5; the uplift-yield portion segment follows immediately
after the elastic segment (point R). This shows that for
this case x 0:5; a yield-only or uplift-only condition
does not occur. On the other hand, uplift- and yield-only
conditions occur after the elastic condition at u 1 mrad
(point P) for x 0:1 (e.g. foundations where conditions
other than bearing capacity demands govern the design)
and x 0:9 (e.g. foundations of existing structures that
need retrofitting because of increased loads as a result of
code revisions or change in the use of structure),
respectively. However, the yield-uplift condition occurs,
but at a large rotation of u 25 mrad (not shown on the
graph). For x 0:3 (typical foundation design) and x
0:7 (foundation designed to mobilize its ultimate capacity
under seismic conditions), uplift (for x 0:3) and yield
(x 0:7) initiate at u 3 mrad (point Q). The onset of
the yield-uplift condition occurs at u 8 mrad (point S).
It can thus be concluded that as x approaches 0.5 from
either side, the region where uplift-only or yield-only
occurs shrinks and the region where yield and uplift occur
expands. Based on this observation, three regions of
moment rotation responses can be postulated: upliftdominant region; uplift-yield region; and yield-dominant


N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

Fig. 6. Computed momentrotation curves showing points of change of state.

Based on the ensuing discussion, the following important

observations can be made
1. The moment rotation curve included in the FEMA
273/274 documents (Fig. 2) presents a seemingly different
picture to some of the inferences drawn in this section. First,
the curves for the initial uplift or initial yield conditions are
shown as two separate curves (1 3 5 6 and 1 4 5 6,
respectively), with the initial uplift curve lying above that
for the initial yield. This implies that for the same c value, a
foundation design with x , 0:5 (which leads to initial
uplift) would result in a larger moment response than the
case where x . 0:5 (which leads to initial yield). However,
both curves are similar and the moment rotation response
is rather influenced by the absolute difference between x and
x 0:5: Secondly, Fig. 2 shows that segment (5) of the
curve, which represents the uplift-yield condition is
asymptotic to the ultimate elastic condition, M PB=2:
This is incorrect, since no yielding occurs by definition
(infinitely strong soil).
2. Siddharthan et al. [13,14] stated that the rocking
response could be grouped into either the initial uplift
condition, or the initial yield condition. It has been shown
that the equations for ultimate moment for both conditions
(Eqs. (9b) and (13b)) are the same if formulated in terms
of x: The results presented herein show that based on the
value of x; the moment rotation response can rather be
grouped into three categories based on dominating behavior
and not two categories based on the initiation of uplift of

yield. The correct expression for the ultimate moment has

also been derived.
6. Comparison with experimental results
6.1. Foundation rocking experiments
It is important to validate and corroborate any analytical
or numerical model by experimental evidence. Bartlett [12]
and Wiessing [16] conducted rocking tests on foundations
installed in clay and sand, respectively. The experimental
results agreed, in general, with those obtained numerically
from an elastic perfectly plastic cyclic Winkler model.
The ability of the Winkler approach developed in this
study to evaluate the moment rotation behavior of a
foundation is verified using available experimental results.
The model developed is used to analyze the moment
rotation response of foundations subjected to rocking action
in a laboratory testing program and the results are compared
with the measured values.
The European Commission (EC) sponsored the project
TRISEE (3D Site Effects of Soil Foundation Interaction in
Earthquake and Vibration Risk Evaluation), which included
large-scale model testing to examine the response of rigid
footings to dynamic loads. The results of these tests are of
high quality and are readily available [17]. Therefore, these
tests are analyzed using the developed model and the results
are compared with the measured values.

N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

6.2. TRISEE experiment

The experiments involved a 1 m square footing model
embedded to a depth of 1 m, in a 4.6 m 4.6 m 3 m deep
sample of saturated Ticino sand. Ticino sand is a uniform
coarse-to-medium silica sand. The properties of the sand are
as follows: D50 0:55 mm; coefficient of uniformity, Cu
1:6; specific gravity, Gs 2:684; emin 0:579; and emax
0:931 [18]. Two series of tests were performed on the model
foundation installed in two different soil samples with
relative density of 45% (low density, LD) and 85% (high
density, HD).
A vertical load of 100 and 300 kN was applied to the LD
and HD samples, respectively, before the application of the
horizontal cyclic loading phase. The imposed pressures of
100 and 300 kPa represent typical design pressures for
foundations in medium to dense sands, where the design is
usually governed by admissible settlement, and not bearing
capacity requirements. The resulting static FS under vertical
load only was found to be about five in both cases. The
cyclic loading involved three phases: Phase Ithe application of small-amplitude force-controlled cycles; Phase
IIthe application of a typical earthquake-like time history;
and Phase IIIsinusoidal displacement cycles of increasing
amplitude. Only relevant sections of the results of the tests
would be presented for comparison purposes. Further
information on the experiments can be found in Refs.
[17 20].
6.3. Comparison with TRISEE experiments
A reasonably accurate estimate of the soil subgrade
modulus, kv ; is required for the analytical model to


accurately evaluate the foundation rocking response during

the loading tests. The load deformation results obtained
during the application of the static vertical load only (similar
to a plate loading test) were used to back figure the soil
subgrade modulus. Atkinson [21] recommended that results
of plate loading tests, when expressed in terms of an
average strain given by the settlement/width ratio, e a
rs =B; are similar to the results of a triaxial test but scaled up
by a factor of 2 3. Briaud and Gibbens [22] Ismael [23]
made similar observations regarding the relationship of P
versus rs =B in plate loading tests.
Fig. 7 shows the load deformation results of the
TRISEE experiments plotted in terms of e a ; along with
the stiffness values (subgrade modulus) for initial, secant
and unload reload loading conditions as evaluated from the
test results. These values of the subgrade modulus are used
in the developed Winkler model to calculate the moment
rocking response of the foundation and the results are
compared with the measured response in Fig. 8. It should be
noted that the measured response represents the envelop of
the loading cycles with gradually increasing peak amplitude
(i.e. obtained by connecting the tips of the hysteretic loops)
[24,25]. Lo Priesti et al. [26] noted that because of the
different impact of plastic strains under different loading
conditions, it is impossible to obtain the same backbone
curve for both monotonic and cyclic loading. Fig. 8 shows
that the results computed using the unload reload stiffness
gives the best agreement with the experimental results. The
initial stiffness is slightly underestimated, but the overall
response is generally satisfactory. This is expected since
cyclic loading represents an unload reload action, and the
unload reload stiffness is more representative of the smallstrain stiffness.

Fig. 7. Loaddeformation results from TRISEE experiments: (a) for HD tests; (b) for LD tests.


N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

Fig. 8. Comparison between predicted and experimental momentrotation curves.

Soil nonlinearity and creep effects significantly influence the initial stiffness. For example, the experimental
results for the LD specimen showed that the creep
settlement accounted for about 40% of the observed
settlement [18]. The comparison in Fig. 8 of moment
rotation curves (rocking stiffness) calculated using different subgrade moduli with the experimentally determined
curve shows that rocking stiffness is grossly underestimated when the secant subgrade modulus is used in
the calculations. This result shows that care and judgment
are required to select an appropriate subgrade modulus.

The issue of an appropriate subgrade modulus is discussed

in Section 7.

7. Comparison with code recommendations

7.1. Code recommendations
The NEHRP FEMA 273/274 provisions for SSI in a
seismic response analysis vary according to the type of
analysis performed. For the linear static procedure (LSP),

N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

a simplified method given in the FEMA 302/303 documents

[27] should be used. For the linear dynamic procedure
(LDP), equivalent elastic foundation stiffnesses are used.
For the nonlinear static procedure (NSP), static force
deformation curves are used to model the foundation
response. Finally, complete cyclic force deformation
curves are used to model the foundation in the nonlinear
dynamic procedure (NDP).
The recommended force deformation curve is a bilinear
moment rotation relationship, and will be referred to as the
bilinear model herein. The bilinear model accounts for soil
variability and difficulty in accurately determining the
foundation loads and other factors by introducing upper and
lower bounds. The upper and lower bounds are specified as
twice and one half the best estimates of stiffness and
strength. The best stiffness estimates are based on the elastic
halfspace solutions [28], the best bearing capacity estimate
is based on bearing capacity of a shallow footing under
vertical load [29] and the ultimate moment capacity is
evaluated using Eq. (19).
7.2. Discussion of factors affecting estimation of the shear
7.2.1. Mean effective pressure
The FEMA 302/303 documents state that the mean
effective pressure, sm ; should be estimated using both the
overburden and applied pressures, while the FEMA 273/274
documents stipulate the use of the overburden pressure only.
The assumed value sm may have a significant effect on the
small-strain soil modulus, E0 ; calculated from expressions
that relate the soil modulus to the mean effective pressure,
e.g. [30]
E0 1510

2:17 2 e2 0
sm 0:53p0:47


It should be noted that Eq. (22) was developed based on

triaxial tests performed on Ticino sand that was used in the
TRISEE experiments. The variation of the calculated shear
modulus with the assumption used to calculate sm for the
soil sample used in the TRISEE experiment is shown in
Table 1. The difference between G0 (calculated from E0
assuming isotropic elastic conditions) based on the two
assumptions within 1 m below the foundation (i.e. one B;
where most of stress increase and deformation occur due to
applied vertical load) is around 50% for the HD case. Since
the FEMA 273/274 specifies a lower bound based on one
half of the best estimate of stiffness and bearing capacity to
cover all sources of uncertainty, care must be exercised
when calculating the mean effective pressure.
7.2.2. Shear modulus reduction curve
Equivalent linear methods that make use of the secant
stiffness (i.e. reduced modulus) estimated from load
deformation curves [21] are usually employed when
analyzing the nonlinear response of foundations subjected


Table 1
Variation of G0 with depth for different assumptions of the mean effective
pressure, sm
Depth G0 overburden applied stress G0 overburden only Difference
HD test



LD test



to static loads. On the other hand, the modulus reduction

curves approach was developed mainly for the dynamic
loading case, but has been used for both static and dynamic
loading cases. This causes some confusion regarding the
definition of modulus reduction curves, which has been the
focus of some research [21,31,32].
The maximum, small-strain soil shear modulus, G0 ; has
the same value for static monotonic, static cyclic and
dynamic loading cases [31,32]. On the other hand, the
modulus reduction curve is the same for static cyclic and
dynamic loading cases, but different for static monotonic
loading. For static monotonic loading, the modulus
reduction curve is defined as a reduction in the secant
stiffness from the origin to a specified point on the
monotonic curve with an increase in static shear strain;
while for cyclic (dynamic) loading, it is defined as a
reduction in the peak-to-peak secant stiffness of the
unload reload hysteretic loops with an increase in cyclic
shear strain. For dynamic analysis, G0 can thus be
evaluated using monotonic loading tests, however, modulus of reduction curves must be obtained from cyclic
loading tests [33]. This further explains why the unload
reload stiffness gave a better comparison with the TRISEE
experimental results.
Fig. 9 shows the modulus reduction curves established
from the TRISEE test results during the initial static loading
phase [18] by normalizing the soil elastic modulus, E; backcalculated from the test results using the small-strain soil
modulus, E0 : Fig. 9 shows that the unload reload elastic
modulus represents 70% of the small-strain modulus
calculated by Eq. (22). A better agreement between the
two values of the soil modulus would be expected.
However, the difference is mainly due to the assumptions
used when estimating the secant Youngs modulus, and the
effect of soil nonlinearity on the modulus obtained from
only one unload reload cycle [31,34]. It is thus expected
that the correct value of the unload reload modulus would
be . 0:7E0 : During the cyclic phase of the experiment, an
increase in the cyclic shear strain as a result of the
interaction between the foundation and the soil (commonly


N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

Fig. 9. Modulus ratio versus average strain for HD and LD tests.

termed secondary nonlinearity) would result in a reduction

in the unload reload modulus. This phenomenon is
localized and when its effect is averaged over the depth of
the soil medium, the reduction in G is not expected to be
large. Therefore, it is assumed that the opposing effects
would cancel out and G can be taken as 0:7G0 :
7.2.3. Assumptions for G
Table 2 shows values of the rocking and vertical stiffness
computed using generalized uncoupled static stiffness

expressions [35] (no consideration of loading frequency)

given in FEMA 302/303 based on different assumptions for
G: The table also shows the equivalent subgrade modulus
estimated from either the rocking or vertical stiffness
(according to the approach given in the FEMA 273/274
guidelines). The FEMA 302/303 recommendations specify
the depth of influence for rocking and horizontal/vertical
motions as 1:5ru and 4ru ; respectively, where ru and ru are
the equivalent radii of a circular foundation with the same
moment of inertia and area. For the square foundation, this
results in a depth of influence of 0.6 m for rocking motion
and 2.2 m for translational motion. Representative values of
the small-strain shear modulus, G0 ; are thus computed at 0.3
and 1 m (representing one B) below the footing for both
motion types. The different assumptions used for G are as
1. G back-calculated from the unload reload modulus
shown in Fig. 9. As discussed earlier, this value is
assumed to be representative of the best estimate of G:
2. G G0 ; i.e. no reduction in the soil shear modulus,
assuming that the cyclic strain is relatively small. This
assumption is expected to result in a slight overestimation of the stiffness and thus underestimates the predicted
3. G back-calculated from the rocking stiffness expression,
using the measured small displacement rocking stiffness
obtained during Phase I of the TRISEE experiment.

Table 2
Computed foundation stiffnesses based on the different assumptions for G

HD test
G; taken as equal to
G0 (MPa)
G; from unloadreload
stiffness (MPa)
G; from Phase I experimental
M 2 u stiffness (MPa)a
G; from guideline G reduction
factors (MPa)
Winkler-based on unload
reload stiffness
LD test
G; taken as equal to
G0 (MPa)
G; from unloadreload
stiffness (MPa)
G; from Phase I experimental
M 2 u stiffness (MPa)a
G; from guideline G reduction
factors (MPa)
Winkler-based on unload
reload stiffness

Uncoupled rocking
stiffness (MN/m)

Uncoupled vertical
stiffness (MN/m)

Subgrade modulus estimated

from rocking stiffness (MN/m2)

Subgrade modulus estimated from

vertical stiffness (MN/m2)



































G back-calculated from measured rocking stiffness using rocking stiffness equation.

Value of Winkler spring stiffness estimated using measured unload reload stiffness.

N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

The estimated values of G in this case are larger than the

calculated values of G0 : This value is considered to
represent an upper bound for G:
4. G estimated using the reduction factors specified in the
FEMA 273/274 guidelines. This value is estimated using
the maximum base shear of 0:4P for Phase III of the
TRISEE tests, which leads to an effective shear modulus
ratio of 0.4. This estimate is not expected to produce a
good comparison with the measured response because
the cyclic loading is through the foundation, while the
reduction factor used is proposed for soil primary
nonlinearity associated with cyclic loading through the


soil. Nonetheless, this value could be considered as a

lower bound for G:
7.3. Comparison between Winkler approach and code
bilinear model
The force deformation curves are established employing
the bilinear model and using stiffness constants based on the
different G values as shown in Table 2 for both HD and LD
tests. The results are compared in Fig. 10 with force
deformation curves computed based on the Winkler model
and assuming two different values for the subgrade modulus:

Fig. 10. Comparison of code bilinear approximation and computed curves based on the rocking and vertical stiffnesses with experimental curves.


N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

one established from the vertical stiffness, and another from

the rocking stiffness as described in the NEHRP FEMA 273/
274 guidelines. The force deformation curves measured
during the tests are also shown in the figure.
Fig. 10 shows that both the code bilinear approximation
and the computed Winkler curves agree quite well with the
measured responses for HD and LD tests when G values are
obtained from cases 1 or 2 (i.e. G back-calculated from the
unload reload modulus or G G0 ). The prediction for case
2 was better than case 1, even though the opposite was
expected. It is interesting to note from Table 2 that the
computed stiffnesses for case 2 are very close to the
computed Winkler values, which were obtained based on
the measured unload reload stiffness (i.e. the curves for this
case are almost identical to those shown in Fig. 8). For
almost all cases, the initial stiffness is lower than the
measured for both the bilinear and Winkler models.
However, the overall response prediction is generally
satisfactory. Using the G value for case 3 (i.e. Phase III in
the TRISEE experiment) the initial stiffness is accurately
predicted, but as expected, both models, especially the
bilinear model, overestimate the overall response. Finally,
using the code G reduction factors resulted in a large
underestimation of the initial stiffness and in general a poor
prediction of the response for both models. This case
represents a lower bound for stiffness.
It may be inferred from the results that using a lower
bound stiffness estimate in combination with a best ultimate
moment estimate could lead to a significant overestimation
of the yield displacement, which for bilinear models defines
the onset of nonlinear effects. In this case, the permanent
displacement of the foundation may be underestimated. It
can also be concluded that a good estimate of the value of G
is important to accurately predict the moment rotation
response of the foundation using the code bilinear or
Winkler models.
Fig. 10 also shows that using a value of the subgrade
modulus back figured from either the vertical or rocking
uncoupled stiffness did not have a significant effect on the
computed response. For situations where the depth of
influence for both rocking and translation are assumed to be
the same [36,37], the vertical response would be only
slightly larger than that of the rocking, and could be ignored
when compared to the effect of the value of G: Thus, the
selection of the value of G is more crucial regardless of
which approach is used to evaluate the subgrade modulus.

8. Conclusions
The moment rotation response of rigid spread footings
has been investigated. A solution for the uplift-yield
foundation condition has been derived. The developed
solution, along with the solutions for uplift- and yield-only
conditions enables the full definition of the entire static
moment rotation response. The developed model was used

to analyze the TRISEE experiments and the following

conclusions are made.
1. x 0:5 represents the condition for the maximum
moment response based on the Winkler soil model.
2. The moment rotation response is rather influenced by
the absolute difference between x and x 0:5; which
dictates whether uplift or yield is dominant. Based on the
value of x; the moment rotation response can be
grouped into three categories: uplift-dominant, upliftyield and yield dominant.
3. The unload reload stiffness should be used to estimate
the value of the subgrade modulus in dynamic analysis.
The secant stiffness and initial stiffness lead to underestimating the subgrade modulus.
4. The value of G used in the analysis has a significant
effect on the calculated response. The response calculated using G with no reduction factor in both the code
bilinear model and the Winkler model agreed well with
the measured response.
5. Using either the rocking stiffness or vertical stiffness to
estimate the value of the subgrade modulus has a
minimal effect on the calculated response.

This research was supported by financial support from
the Institute for Catastrophic Loss Reduction at the
University of Western Ontario to the senior author and a
graduate scholarship to the first author from the National
Science and Engineering Research Center (NSERC). The
authors would also like to thank Dr P. Negro of the
European Laboratory for Structural Assessment (ELSA) for
making available the test results of the TRISEE

[1] Comartin CD, Keaton JR, Grant PW, Martin GR, Power MS.
Transitions in seismic analysis and design procedures for buildings
and their foundations. Proceedings of the Sixth Workshop on the
Improvement of Structural Design and Construction Practice in the
US and Japan, Applied Technology Council Report, ATC 15-5,
Victoria, BC; 1996.
[2] Ghobarah A. Performance-based design in earthquake engineering:
state of development. Engng Struct 2001;23(8):87884.
[3] Building Seismic Safety Council (BSSC). FEMA 273/274, NEHRP
guidelines for the seismic rehabilitation of buildings; Iguidelines,
IIcommentary, Washington, DC; 1997.
[4] Applied Technology Council, The seismic evaluation and retrofit of
concrete buildings. ATC 40, Vols. I and II. Palo Alto, CA: Redwood;
[5] Meek JW. Dynamic response of tipping core buildings. Earthquake
Engng Struct Dyn 1978;6(4):4534.
[6] Karabalis DL, Beskos DE. Dynamic response of 3-D rigid surface
foundations by time domain boundary element method. Earthquake
Engng Struct Dyn 1984;12(1):7393.

N. Allotey, M. H. El Naggar / Soil Dynamics and Earthquake Engineering 23 (2003) 367381

[7] Yan LP, Martin GR, Simulation of momentrotation behavior of a
model footing. Proceedings of the International FLAC Symposium on
Numerical Modeling in Geomechanics, Minneapolis, MN; 1999. p.
[8] Cremer C, Pecker A, Davenne L. Cyclic macro-element for soil
structure interaction: material and geometrical nonlinearities. Int J
Numer Meth Geomech 2001;25(13):125784.
[9] Gazetas G, Mylonakis G. Seismic soil structure interaction: new
evidence and emerging issues. Dakoulas P, Yegian M, Holtz RD,
editors. Geotechnical Special Publication No. 75 1998;111974.
[10] Filiatrault A, Anderson DL, DeVall RH. Effect of weak foundation on
the seismic response of core wall type buildings. Can J Civil Engng
[11] Chaallal O, Ghlamallah N. Seismic response of flexibly supported
coupled shear walls. J Struct Engng, ASCE 1996;122(10):118797.
[12] Bartlett PE. Foundation rocking on a clay soil. ME Thesis. University
of Auckland, School of Engineering Report No 154; 1979.
[13] Siddharthan RV, Ara S, Norris GM. Simple rigid plastic model for
seismic tilting of rigid walls. J Struct Engng, ASCE 1992;118(2):
[14] Siddharthan RV, El Gamal M, Maragakis EA. Stiffnesses of
abutments on spread footings with cohesionless backfill. Can Geotech
J 1997;34(4):68697.
[15] CEN, Eurocode 7geotechnical design part 1: general rules.
Bruxelles, Belgium: European Committee for Standardization
(CEN); 1994.
[16] Wiessing PR. Foundation rocking on sand. ME Thesis. University of
Auckland, School of Engineering Report No. 203; 1979.
[17] Negro P, Verzeletti G, Molina J, Pedretti S, Lo Presti D, Pedroni S.
Large-scale geotechnical experiments on soilfoundation interaction.
Special publication No. I.98.73, Ispra, Italy: Joint Research Center,
European Commission; 1998.
[18] Jamiolowski M, Lo Presti D, Puci I, Negro P, Verzeletti G, Molina J,
Faccioli E, Pedretti S, Pedroni S, Morabito P. Large scale geotechnical
experiments on soilfoundation interaction. In: Jamiolkowski M,
Lancellota R, Lo Presti D, editors. Pre-failure deformation characteristics of geomaterials. Rotterdam: Balkema; 1999. p. 749 57.
[19] Negro P, Paolucci R, Pedretti S, Faccioli E. Large scale soilstructure
interaction experiments on sand under cyclic load. Paper No. 1191,
12th World Conference on Earthquake Engineering, Auckland, New
Zealand; 2000.
[20] Pedretti S. Nonlinear seismic soil foundation interaction. PhD
Thesis. Department of Structural Engineering, Politecnico di Milano,
Italy; 1998.
[21] Atkinson JH. Nonlinear soil stiffness in routine design. Geotechnique
[22] Briaud JL, Gibbens R. Behavior of five large spread footings on sand.
J Geotech Geoenviron Engng, ASCE 1999;125(9):78796.


[23] Ismael NF. Allowable pressure from loading tests on Kuwaiti soils.
Can Geotech J 1985;22(2):1517.
[24] Puzrin AM, Shiran A. Effects of constitutive relationship on seismic
response of soils. Part I. Constitutive modeling of cyclic behavior of
soils. Soil Dyn Earthquake Engng 2000;19(5):30518.
[25] Bolton MD, Wilson JMR. Discussion: an experimental and theoretical
comparison between static and dynamic torsional soil tests.
Geotechnique 1990;40(4):662 4.
[26] Lo Priesti DCF, Pallara O, Cavallaro A, Maugeri M. Nonlinear
stressstrain relations of soils for cyclic loading. Proceedings of XI
European Conference on Earthquake Engineering, Paris; 1998.
Amsterdam: Balkema (Abstract volume CD-ROM, 187).
[27] Building Seismic Safety Council (BSSC). FEMA 302/303, NEHRP
recommended provisions for seismic regulations for new buildings
and other structures; Iguidelines, IIcommentary, Washington,
DC; 1997.
[28] Gazetas G. Foundation vibrations. In: Fang HY, editor. Foundation
engineering handbook. New York: Van Nostrand, Reinhold; 1991. p.
[29] Vesic AS. Bearing capacity of shallow foundations. In: Winterkorn
HF, Fang HY, editors. Foundation engineering handbook. New York:
Van Nostrand, Reinhold; 1975. p. 12147.
[30] Hoque E, Tatsuoka F, Sato T, Kohata Y. Inherent and stress-induced
anisotropy in small strain stiffness of granular materials. Proceedings
of the First International Conference of Earthquake Geotechnical
Engineering, Tokyo, Japan, vol. 1; 1995. p. 277 82.
[31] Lo Presti DCF, Pallara O, Lancellota R, Armandi M, Maniscalco R.
Monotonic and cyclic loading behavior of two sands at small-strains.
Geotech Test J 1993;16(4):40924.
[32] Shibuya S, Tatsuoka F, Teachavorasinskun S, Kong XJ, Abe F, Kim
YS, Park CS. Elastic deformation properties of geomaterials. Soil
Found 1992;32(3):2646.
[33] Tatsuoka F, Teachavoransinskun S, Park CS. Discussion: An
experimental and theoretical comparison between static and dynamic
torsional soil tests. Geotechnique 1990;40(4):65962.
[34] Alarcon-Guzman A, Chameau JL, Leonards GA, Frost JD. Shear
modulus and cyclic undrained behavior of sands. Soil Found 1989;
29(4):105 19.
[35] Elsabee F, Morray JP. Dynamic behavior of embedded foundations.
Report No. R77-33, Massachusetts: Department of Civil Engineering,
Massachusetts Institute of Technology; 1977.
[36] Roesset JM. A review of soilstructure interaction. In: Johnson JJ,
editor. Soil structure interaction: the status of current analysis
methods and research. Report No. NUREG/CR-1780, Washington,
DC: US Nuclear Regulatory Commission; 1980.
[37] Stewart JP, Kim S. Empirical verification of soilstructure interaction
provisions in building codes. Dakoulas P, Yegian M, Holtz RD,
editors. Geotechnical Special Publication No. 75 1998;125969.