You are on page 1of 11

Passive Earth Pressure Mobilization during Cyclic Loading

Ryan T. Cole1 and Kyle M. Rollins2

Abstract: The passive resistance measured in a series of full-scale tests on a pile cap is compared with existing theories. Four different
soils were selected as backfill in front of the pile cap and the load-deflection relationships under cyclic loading were investigated. The log
spiral theory provided the best agreement with the measured passive resistance. The Rankine theory significantly underestimated the
passive force, while the Coulomb theory generally overestimated the resistance. The displacement necessary to mobilize the maximum
passive force was compared with previous model and full-scale tests and ranged from 3.0 to 5.2% of the cap height. A hyperbolic model
provided the best agreement with the measured backbone passive resistance curve compared with recommendations given by Caltrans and
the U.S. Navy. However, this model overestimated the passive resistance for cyclic loading conditions due to the formation of a gap
between the pile cap and backfill soil and backfill stiffness reduction. Based on the test results, the cyclic-hyperbolic model is developed
to define load-deflection relationships for both virgin and cyclic loading conditions with the presence of a gap.
DOI: 10.1061/共ASCE兲1090-0241共2006兲132:9共1154兲
CE Database subject headings: Passive pressure; Pile groups; Earth pressure; Load tests; Lateral pressure; Cyclic loads.

Introduction lateral spread and this could lead to unsafe designs. Therefore, to
properly design pile caps and bridge abutments an accurate as-
Passive earth pressures against structures may help resist lateral sessment of passive pressure is critical.
forces generated from earthquakes, waves, wind, structural ex- Recently there has been a renewed interest in passive resis-
pansion, and impacts. In contrast, passive earth pressures can act tance, and several full-scale tests have been conducted 共Romstad
as a destabilizing force as a soil mass pushes into a structure, as et al. 1996; Gadre 1997; Duncan and Mokwa 2001; Rollins and
might occur during a landslide or a liquefaction-induced lateral Sparks 2002兲. The results from these studies indicate that passive
spread. In either case, an accurate estimate of the passive earth resistance may contribute significantly to the lateral resistance of
pressure is of significant importance to engineers in producing pile caps and abutments. Nevertheless, due to limited full-scale
test results, there is still much uncertainty regarding appropriate
safe yet economical structures such as pile caps and bridge
load-deflection relationships for use in design. Furthermore, it is
abutments.
unclear how the passive force will change when subjected to cy-
Several methods are available for estimating the maximum
clic loading, further complicating the application of passive resis-
passive pressure, including the Rankine, Coulomb, and log spiral
tance in soil-structure interaction problems.
theories. The passive forces computed using these theories may
To provide information useful in characterizing passive pres-
be contradictory and often vary by more than a factor of two. sure and pressure-deflection relationships during cyclic loading,
Additionally, the development of the passive force with deflection several cyclic lateral load tests were performed on a full-scale pile
is not well understood, adding further complications when con- cap for this study. Four different soils were selected as backfill—a
sidering the interaction between the structure and the soil. Given clean sand, silty sand, fine-grained gravel, and coarse-grained
the uncertainty regarding computation of the maximum passive gravel—covering a range of materials potentially used in practice.
resistance and its development with deflection, conservative as- Greater detail regarding the testing program is provided in a com-
sumptions are often employed, which may include using the panion paper 共Rollins and Cole 2006兲. The passive resistance re-
Rankine method or neglecting the passive force altogether. Al- sults from this study are presented and compared with existing
though the Rankine method may be conservative when the pas- theories and models. Finally, a method for computing the passive
sive force is resisting lateral forces, use of the Rankine method resistance during cyclic loading is presented.
will significantly underestimate the passive force produced by a

1
Senior Engineer, IGES, Inc., 4153 Commerce Dr., Salt Lake City, Passive Earth Pressure
UT 84107. E-mail: ryanc@igesinc.com
2
Professor, Dept. of Civil and Environ. Engrg., 368 CB, Brigham
Young University, Provo, UT 84602 共corresponding author兲. E-mail: Theories
rollinsk@byu.edu Many methods are available for computing the maximum passive
Note. Discussion open until February 1, 2007. Separate discussions earth pressure, with the more common based on limit-
must be submitted for individual papers. To extend the closing date by
equilibrium, including the Rankine 共1847兲, Coulomb 共1776兲, and
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos- log spiral 共Terzaghi 1943兲 theories. Others procedures include
sible publication on August 15, 2004; approved on December 20, 2004. plasticity theories, method of slices, empirical, and various finite-
This paper is part of the Journal of Geotechnical and Geoenvironmental element and finite-difference computer methods.
Engineering, Vol. 132, No. 9, September 1, 2006. ©ASCE, ISSN 1090- For a structure restrained against vertical movement, such
0241/2006/9-1154–1164/$25.00. as a pile cap, shear between the structure-soil interface will

1154 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006


Table 1. Movement Necessary for Development of the Maximum Passive Earth Pressure
Values of ⌬max / H

Type of backfill Sowers and Sowers 共1961兲 Canadian Geotechnical Society 共1992兲 Clough and Duncan 共1991兲
Dense sand 0.002 0.02 0.01
Medium-dense sand — — 0.02
Loose sand 0.006 0.06 0.04
Stiff cohesive — 0.02 —
Soft cohesive — 0.04 —
Compacted silt — — 0.02
Compacted lean clay — — 0.05
Compacted fat clay — — 0.05

fully-mobilize the interface friction. Therefore, only earth pres- provide an “effective width” for the wall 共U.S. Naval C.E.L.
sure theories that account for the interface 共wall兲 friction angle 1962; Brinch Hansen 1966; Krebs Ovesen and Stromann 1972兲.
共␦兲 are likely to provide accurate estimates of passive resistance. The results from these studies suggest the 3D effects may signifi-
The Rankine method assumes ␦ is equal to the slope of the cantly increase the resistance of “shorter” walls or pile caps. Each
backfill and tends to underestimate the passive resistance. Con- of these methods use the wall geometry to incorporate the 3D
versely, the Coulomb method accounts for ␦ but tends to over- influence; however, the Brinch Hansen formula also considers the
estimate the passive resistance for cases where the interface friction angle of the soil. Full-scale passive earth pressure tests
friction is greater than 0.4– 0.5 times the backfill soil friction performed by Duncan and Mokwa 共2001兲 found reasonable
angle 共␾兲 共Morgenstern and Eisenstein 1970; Clough and Duncan agreement with the Brinch Hansen 3D factor 共R兲. Therefore, the
1991; Duncan and Mokwa 2001兲. Brinch Hansen formula was used for this study.
Results from model and full-scale tests 共Mackey and Kirk
1967; Fang et al. 1994; Gadre 1997; Duncan and Mokwa 2001;
Development of Passive Earth Pressure
Rollins and Sparks 2002兲 suggest that the log spiral method may
with Deflection
be the more appropriate method for computing passive earth pres-
sure. Additionally, alternative methods for computing passive It is generally accepted that more displacement is required to
earth pressure, such as plasticity theory and finite-element analy- mobilize full passive conditions with respect to active conditions.
sis, generally agree with the log spiral theory 共Duncan and However, there is no consensus on the displacement needed to
Mokwa 2001兲. Although the theoretical and field test results typi- reach the ultimate passive state. Several recommendations and
cally confirm the reliability of the log spiral theory, it is not methods have been proposed to address this issue, including full-
commonly used in practice. scale and model tests, which provide useful information defining
how passive earth pressure develops with deflection.
The movement required to reach the full passive state 共⌬max兲 is
Pressure Distribution
generally expressed as a function of the height of the pile cap
The distribution of passive pressure with depth is typically as- or wall 共H兲. The movement required to reach the passive state
sumed to be linear for a uniform soil; however, there is conflicting has been investigated both experimentally and using numerical
literature regarding the actual shape of the distribution with depth. techniques. Clough and Duncan 共1991兲 suggest the movement
Several model tests indicate a linear pressure distribution 共Rowe required to reach the passive state is greater for loose, compress-
and Peaker 1965; Mackey and Kirk 1967; Fang et al. 1994兲, while ible soils, than for dense, relatively incompressible soils.
others show a nonlinear pressure distribution 共Narain et al. 1969; Table 1 presents the movement necessary for development
Mackey and Kirk 1967; James and Bransby 1970; Fang et al. of the maximum passive pressure as a function of wall height
1994兲. Narain et al. 共1969兲 and Fang et al. 共1994兲 observed linear 共⌬max / H兲 according to several sources. The Canadian Geotech-
and nonlinear 共parabolic兲 pressure distributions for cases of trans- nical Society 共1992兲 describes the passive earth movement as
lational and rotational wall movements, respectively. Despite the rotational, whereas Clough and Duncan 共1991兲 describe the
conflicting results, for design purposes, geometrically uncompli- movement as translational or rotational. These recommended val-
cated distributions are preferred to simplify calculations 共Clough ues of ⌬max / H necessary to obtain maximum passive earth pres-
and Duncan 1991兲. sure 共E p max兲 may vary by an order of magnitude for similar soil
type and density. While tables such as this can be useful for large
movements, they provide no information for estimating E p for
Three-Dimensional Effects
movements less than ⌬max, which is often of considerable interest
The theories previously discussed express the passive force per in soil–structure interaction problems.
unit length of wall 共E p兲. The derivations are based on a plane The U.S. Navy 共1982兲 and Canadian Geotechnical Society
strain assumption that considers a cross section from an infinitely 共1992兲 provide graphical relationship between the normalized
long structure. Therefore, to obtain the total force over the length wall movement 共⌬ / H兲 and the passive earth pressure coefficient
of the wall, one simply multiplies E p by the length of the wall. 共K p兲 for loose and dense sands. Similar curves are given by
However, for shorter walls or pile caps, the failure surface Clough and Duncan 共1991兲 for loose sands, dense sands, and
extends beyond the edge of the pile cap and mobilizes more re- compacted backfill. These curves can be normalized by the maxi-
sistance than would be expected based on the actual length of the mum passive earth pressure coefficient 共K p max兲 and scaled for
wall. Based on field and laboratory testing, several methods have different values of K p max to characterize the development of pas-
been suggested to account for the boundary or 3D effects and sive pressure with deflection.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006 / 1155


Fig. 1. Hyperbolic representation of passive resistance 共adapted from
Duncan and Mokwa 2001兲

Caltrans 共2001兲 recommends using a bilinear model with an


initial slope 共stiffness兲 of 11.5 kN/ mm/ m width of wall, which is
limited to an ultimate uniform passive pressure of 239 kPa. The
stiffness and uniform passive resistance are scaled linearly for
different abutment heights using

kN/mm H
Kabut = 11.5 ⫻w⫻ 共kN/mm兲 共1兲
m 1.7

h
E pH = Ae ⫻ 239 kPa ⫻ 共kN兲 共2兲
1.7

where H⫽wall height in meters; w⫽wall width in meters;


and Ae⫽effective wall area in meters square. This method is
based on judgment and large-scale abutment testing. 共Maroney Fig. 2. Pile cap instrumentation: 共a兲 plan; 共b兲 elevation view
1995; Romstad et al. 1996兲.
Duncan and Mokwa 共2001兲 approximate the passive resistance
development using a general hyperbolic equation 共see Fig. 1兲 Full-Scale Passive Earth Pressure Results

⌬ A series of lateral load tests were performed on a rectangular pile


P= 共3兲 cap 5.18 m in width, 3.05 m in length, with a height of 1.12 m.
1 ⌬ Fig. 2 shows a plan view and cross section of the testing facility.
+ Rf
Kmax Pult The tests were designed to obtain the passive resistance compo-
nent contributing to the lateral resistance of the pile cap. Four
where P⫽mobilized passive resistance 共units of force兲; different soils were used as backfill compacted in front of the
Pult⫽maximum passive resistance 共units of force兲; ⌬⫽deflection pile cap. The pile cap was loaded laterally in a cyclic manner to
共units of length兲; R f ⫽failure ratio; and Kmax⫽initial stiffness of define the nonlinear load versus displacement curve. The pile cap
the load-deflection curve 共units of force/length兲. Kmax is estimated was loaded to a specified deflection and the load was typically
using the elastic solution of a horizontal load on a vertical rect- reapplied and cycled 10– 15 times to approximately the same de-
angle after Douglas and Davis 共1964兲, which requires estimates of flection level. For subsequent loading sequences the pile cap was
the soils’ initial Young’s modulus 共Ei兲 and Poisson’s ratio 共␯兲. loaded to a larger deflection and the cycling was repeated. Further
Duncan and Mokwa 共2001兲 used an R f value of 0.85 for all of details of the study are discussed in a companion paper 共Rollins
their computed results. However, they state that the value can and Cole 2006兲.
range from 0.75 to 0.95. Alternatively, one could eliminate R f by The passive resistance versus deflection curves obtained for
solving Eq. 共3兲 for R f when P = Pult and ⌬ = ⌬max, and rewriting as the first cycle of each backfill test are given in Figs. 3 and 4. The
shape of the curves change as the pile cap is deflected further into
⌬ the backfill. Due to the cyclic loading, gaps developed between

冉 冊
P= 共4兲
1 Pult ⌬ the pile cap and backfill. For the sand tests, the width of the gap
+ 1− was consistent with depth and width along the pile cap, which
Kmax Kmax · ⌬max Pult
may explain the abrupt changes in the shape of the first cycle’s
Using this form of the equation, R f is eliminated and ⌬max can be passive resistance. However, the gap was not consistent and some
estimated using an appropriate value for ⌬max / H, chosen appro- backfill caving occurred for the gravel tests, which could explain
priately from Table 1. the initial concave up portion of the gravel curves. The passive

1156 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006


Fig. 3. First cycle and backbone passive resistance curves Fig. 4. First cycle and backbone passive resistance curves

Peak Passive Resistance


force begins to develop as the gap between the pile cap and back- The observed peak horizontal passive force 共Ph−ult兲 and corre-
fill soil is closed and contact is made. During reloading, the gap sponding deflection 共⌬max兲 were obtained from the passive force
results in a significant decrease in passive resistance compared versus deflection 共backbone兲 curves for the clean sand, fine
with the monotonic backbone curve for deflections less than the gravel, coarse gravel, and silty sand backfill tests 共see Figs. 3 and
gap width. The gaps were typically observed as the pile cap de- 4兲. Ph−ult was calculated using the soil properties and pile cap
flection exceeded 0.5% of the pile cap height. geometry

Ph−ult = E ph · B · R 共5兲
Passive Earth Comparisons where E ph⫽computed horizontal component of passive resistance
per unit length of wall 共units of force/length兲; B⫽width of the pile
The strength parameters for the backfill soils were determined cap 共units of length兲; and R⫽three-dimensional correction factor
using a combination of laboratory and in-situ direct shear testing, 共dimensionless兲. The computed maximum horizontal passive
and are presented in Table 2. Additional information regarding the force using the Rankine, Coulomb, log spiral, and Caltrans meth-
backfill soil properties is given by Rollins and Cole 共2006兲. ods are listed in Table 3 along with the observed values for each

Table 2. Summary of Soil Parameters Used for Computing Passive Resistance


Moist unit weight Cohesion intercept Interface friction
共␥m兲 共c兲 Friction angle, ␾’ angle ␦’
Backfill sample 共kN/ m3兲 共kPa兲 共degrees兲 共degrees兲 ␦/␾
Clean sand 18.35 0 39 30 0.77
Fine gravel 20.76 3.8 34 26 0.75a
Coarse gravel 23.12 7.2 40 30 0.75a
Silty sand 18.99 27.3 27 20 0.75a
a
Estimated values, assuming no adhesion 共intercept兲.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006 / 1157


Table 3. Comparison of Measured and Predicted Peak Passive Force Development of Passive Resistance with Displacement
共Ph-ult兲
The monotonic passive force versus deflection curves from the
Peak passive force 共kN兲
four tests were compared with the U.S. Navy 共1982兲 curve,
Method Clean sand Fine gravel Coarse gravel Silty sand the Caltrans 共2001兲 method, and the hyperbolic model given by
Measured 1,090 774 1,997 1,428 Duncan and Mokwa 共2001兲. The normalized U.S. Navy curve was
Caltrans 914 914 914 914 multiplied by the maximum passive pressure computed using log
Coulomba 1,577 共1,577兲b 1,149 共824兲b 3,464 共2,224兲b 1,575 共351兲b spiral charts and the Brinch Hansen 共1966兲 R factor to account for
Log spirala 922 817 1,688 1,210 the 3D effects 共see Table 4兲. The cohesion intercept contribution
Rankinea 357 共357兲b 405 共300兲b 719 共474兲b 804 共194兲b was also included.
a
Method includes Brinch Hansen 共1966兲 3D 共R兲 correction factors 共see The Caltrans curve was constructed using the pile cap height
Table 4兲. 共H = 3.05 m兲 and width 共w = 5.18 m兲 with Eqs. 共1兲 and 共2兲 result-
b
Cohesion contribution computed using trial wedge for Coulomb ing in an initial stiffness of 39.2 kN/ mm, and limited to 914 kN
and 2c冑K p for Rankine; numbers in parenthesis neglect cohesion for the pile cap geometry used in this study. Using these param-
contribution. eters the ⌬max / H occurs at 2.08% of the wall height.
The hyperbolic curve given by Duncan and Mokwa 共2001兲
was constructed using Eq. 共4兲. The parameters used for this model
backfill soil. The Coulomb values were computed, including the are given in Table 4 and were selected based on soil type and
cohesion contribution along the surface of sliding. For the geom- density conditions according to recommendations provided by
etry of the pile cap and ␾ used for this study, the Brinch Hanson Duncan and Mokwa.
共1966兲 method gives effective pile cap widths ranging from 6.12 Figs. 5 and 6 compare the measured backbone passive re-
to 7.13 m, equivalent to R values of 1.180– 1.376 共see Table 4兲. sistance curves for each backfill soil with the passive earth pres-
Additionally, the computed passive resistance for the Rankine and sure methods and models previously discussed. The Rankine and
Coulomb methods neglecting the cohesion component have been Coulomb maximum passive force values, corrected for 3D ef-
included, as the cohesion intercept may fluctuate over time with
fects, are also plotted for comparison.
changes in the pore pressure and is often neglected in design for
long-term conditions. The U.S. Navy curve consistently underestimates passive re-
As expected, the Rankine method consistently underestimated sistance for deflections less than 3.6% of the wall height. The
Ph−ult and was 44– 67% lower than measured value. The Caltrans initial curve shape underestimates the measured passive resistance
method provided passive resistance estimates within 20% of the by 30– 50%. With increasing deflection the Navy curve ap-
observed results for the clean sand and fine gravel tests, but un- proaches the maximum passive force; however, the shape of the
derestimated the coarse gravel and silty sand tests by 54 and 36%, curve is significantly different. The Caltrans curve initially under-
respectively. The Coulomb method estimated Ph−ult within 10% estimates the passive resistance by 40– 70% for deflections less
for the silty sand test, but overestimated the passive resistance by than 21 mm 共⌬ / H ⬍ 1.0% 兲. However, with increasing pile cap
45– 73% for the clean sand, fine gravel, and course gravel tests. deflection 共⌬ / H ⬎ 2.0% 兲 this method provided a reasonable esti-
Neglecting the cohesion in the Coulomb method provides closer mate for the clean sand and fine gravel tests, but underestimates
estimates of the passive resistance for the fine and coarse gravel the passive resistance by 20 and 54% for the silty sand and coarse
but further underestimates the resistance provided by the silty
gravel tests, respectively. The initial stiffness, 39.2 kN/ mm, given
sand. Overall, the log spiral method provides the most consistent
by the Caltrans approach, significantly underestimates the ob-
estimate of the passive force for this study. The computed Ph−ult
is typically within 15% of the measured results. Results from served stiffness by a factor of 3 to 4. The hyperbolic model with
this study reinforce the validity of the log spiral method for cal- the log spiral estimate of passive resistance 共Ph−ult兲 generally pro-
culating the maximum passive resistance, and are consistent with vides a reasonable estimate of the shape of the monotonic passive
conclusions from other full-scale field tests 共Duncan and Mokwa resistance curve. The model consistently predicts the passive re-
2001; Rollins and Sparks 2002兲 and model tests 共Mackey and sistance within 15% and the shape of the curve more accurately
Kirk 1967; Fang et al. 1994兲. captures the observed behavior.

Table 4. Summary of Parameters Used in Hyperbolic Model


Young’s
Dr modulus, Ei Poisson’s ratio, Kmax ⌬max 3D factor,
Backfill sample 共%兲 共MPa兲a va 共kN/mm兲b 共mm兲 Rc
Clean sand 63 37.1 0.27 246 38 1.360
Fine gravel 54d 33.8 0.31 226 34 1.260
Coarse gravel 69d 39.0 0.26 259 39 1.376
Silty sand 67d 38.3 0.35 260 58 1.180
a
Interpolated from the range of values recommended by Duncan and Mokwa 共2001兲.
b
Elastic solution of a horizontal load on a vertical rectangle 共Douglas and Davis 1964兲.
c
Brinch Hansen 共1966兲 3D correction factor.
d
Estimated using correlations; Lee and Singh 共1971兲.

1158 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006


Fig. 5. Backbone passive resistance comparisons Fig. 6. Backbone passive resistance comparisons

Conclusions Based on Analysis of Passive approach is necessary to model the development of passive resis-
Pressure Tests tance for cyclic loading conditions.

The results of this study indicate that the log spiral passive earth
pressure theory provides the more accurate estimates of the mea- Cyclic Passive Earth Resistance Model
sured passive force. This method is particularly robust, because it
accounts for the cohesion of the soil and soil–structure interface This section presents a passive force versus deflection model that
friction 共␦兲. The normalized deflection 共⌬max / H兲 required to fully accounts for the reduction in force due to cyclic loading in which
mobilize the passive earth pressure ranged from 0.03 to 0.052 for a gap forms between the pile cap and backfill soil and the initial
the backfill material tested and is within the range of values ob- backfill stiffness decreases due to repeated loading. The model
served from the model and full-scale tests previously performed employs the hyperbolic relationship given by Eq. 共4兲 as proposed
under translational movement. The log spiral passive earth pres- by Duncan and Mokwa 共2001兲 to characterize the backbone
sure theory corrected for 3D shear effects used together with the passive resistance curve combined with an empirical procedure
hyperbolic model yields reliable estimates between predicted and describing the soil movement 共gap兲 and reduction in stiffness as a
observed backbone curves for each of the backfill soils investi- function of the previous deflection of the pile cap. This method
gated in this study. However, the model only predicts the peak provides a means of modeling the passive force developed as a
force versus deflection curve the soil would experience if loaded pile experiences cyclic loading, and is termed the cyclic-
monotonically. hyperbolic model.
The passive resistance curves 共see Figs. 3 and 4兲 show an
increasing reduction in passive resistance compared with the
Proposed Model
monotonic curve, as the pile cap is cyclically loaded beyond 0.5%
of the pile cap height. The reduction is both a function of the gap A bilinear model was constructed to characterize the passive re-
width between the pile cap and soil and a reduction in backfill sistance reload curves of Figs. 3 and 4. The proposed model and
stiffness due to repeated loading. None of the existing methods terminology are presented in Fig. 7. The initial stiffness of the
consider the effects of cyclic loading and therefore consistently virgin passive force versus deflection curve is defined as
overestimate the load for repeated load conditions and do not Kmax ; Pult defines the maximum passive force 共from log spiral
capture the observed load-deflection behavior. Therefore, a new theory兲 and the passive force versus deflection curve approaches

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006 / 1159


Fig. 7. Cyclic-hyperbolic terminology

this value asymptotically. As the load is released, the soil relaxes


and a gap forms between the pile cap and backfill soil.
Based on the load testing, the subsequent reloaded stiffness of
the passive force-deflection curve decreases relative to the initial
stiffness and does not begin to develop until the pile cap closes Fig. 8. Kmax, Kr, and ⌬s obtained for cyclic-hyperbolic model
the gap. A linear approximation of the stiffness of the reloaded
force-deflection curve is defined as Kr and the horizontal offset
for the linear load versus deflection approximation is called the
apparent soil movement 共⌬s兲, as shown in Fig. 7. The apparent
soil movement is somewhat less than the maximum previous Plots in Fig. 9 provide Kmax , Kr, and ⌬s values for the fine
backfill deflection 共⌬ p兲 due to rebound and relaxation of the gravel and coarse gravel backfill materials. The shapes of these
backfill when the load is removed. For deflections less than ⌬s curves have much more initial curvature than those for the clean
the passive resistance is assumed to be zero. The intercept and silty sand curves. Consequently, more judgment was neces-
between the linear reloaded force-deflection curve and the sary to select the linear stiffness values. With the 共initial兲 soil
hyperbolic load-deflection curve is defined by the coordinate stiffness 共Kmax兲, reloaded soil stiffness 共Kr兲, and apparent soil
共⌬int, Pint兲. For deflections beyond ⌬int, the passive load- movement 共⌬s兲 obtained for each backfill soil, the data was nor-
deflection follows the hyperbolic curve until the load is released. malized and two relationships were developed.

Development of Model Apparent Soil Movement Relationship


The model was constructed to define the apparent soil movement The first relationship was developed by normalizing the apparent
and the reduction in backfill soil stiffness as a function of the soil movement by the previous deflection 共⌬s / ⌬ p兲 and plotting
preceding deflection. To define these relationships, the continuous this value as a function of the previous pile cap deflection nor-
passive resistance curves for each backfill soil type were plotted malized by its height 共⌬ p / H兲 in Fig. 10共a兲. The data indicate that
and the initial stiffness and reloaded stiffness lines were fit to there is a nonlinear relationship between the maximum previous
each of the continuous passive resistance versus deflection curves. pile cap deflection and the position of the backfill once the pile
Fig. 8 is a plot of the linear stiffness approximations 共Kmax and cap is unloaded and moves away from the soil mass. This rela-
Kr兲, and apparent soil movement 共⌬s兲 for the clean sand and silty tionship did not appear to be dependent on soil type for the range
sand backfill materials. At small deflections, the measured passive of soils investigated. At small normalized deflections, the soil
resistance curves correlate well with the chosen linear reload stiff- behaves more elastically so that the apparent soil movement is a
ness values over the entire deflection range. At larger deflections, small fraction of the previous pile cap deflection after rebound. At
partial collapse and in-filling of the gap creates a concave upward greater normalized deflections, the soil behaves more plastically,
shape to the passive resistances during the initial portion of the which decreases the rebound, so that the apparent soil movement
reload curve. In these cases, the reload stiffness values were se- is a larger fraction of the previous pile cap deflection.
lected to match the upper deflection range for each curve so that The nonlinear relationship defining the normalized apparent
the passive resistance would be somewhat underestimated and soil movement was approximated using the general form of a
generally conservative in the lower deflection range. hyperbolic equation

1160 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006


Fig. 9. Kmax, Kr, and ⌬s obtained for cyclic-hyperbolic model
Fig. 10. Cyclic-hyperbolic recommended design curves; 共a兲 values
of ⌬s / ⌬ p versus ⌬ p / H; 共b兲 Kr / Kmax versus ⌬s / H
x
y= 共6兲
a + bx
Values for the coefficients a and b were obtained by transforming
the data into linear form and fitting the data using a least-squares The soil stiffness reduction relationship was approximated using
linear regression where b⫽slope and a⫽intercept. The a and b the nonlinear hyperbolic relationship given in Eq. 共6兲. The coef-
values obtained from this regression were used to construct a ficients were obtained by transferring the data into linear form
hyperbolic equation for the data given by and using a least squared regression to fit the data, thus obtaining
the slope 共b兲 and intercept 共a兲. The equations for this relationship
⌬s 共⌬ p/H兲 are presented as an upper-bound dense gravel, dense sand, and
= 共7兲
⌬ p 0.0095 + 1.23 · 共⌬ p/H兲 lower-bound medium dense gravel, respectively
In addition, results from the regression were used to construct
95% confidence intervals as shown in Fig. 10共a兲. Kr 共⌬s/H兲
=1− 共8兲
Kmax 0.0021 + 2.0 · 共⌬s/H兲
Soil Stiffness Reduction Relationship
The second relationship expresses the reloaded soil stiffness nor-
malized by the maximum 共initial兲 soil stiffness 共Kr / Kmax兲 as a Kr 共⌬s/H兲
=1− 共9兲
function of the apparent soil movement normalized by the pile Kmax 0.0013 + 1.40 · 共⌬s/H兲
cap height 共⌬s / H兲. The normalized values of 共Kr / Kmax兲 as a func-
tion of 共⌬s / H兲 are plotted in Fig. 10共b兲. There is reasonable
agreement between the data points for the two sands; however, Kr 共⌬s/H兲
there is a significant difference between those for the fine and =1− 共10兲
Kmax 0.0012 + 1.24 · 共⌬s/H兲
coarse gravel. The difference in density is one possible explana-
tion for the differences in the shape of the curves shown in The remaining inflection point of the proposed model 共⌬int, Pint兲
Fig. 10共b兲. As indicated in Table 4, the relative compaction of the defines where the reduced stiffness 共Kr兲 intercepts the predefined
fine gravel material was lower than that for the other materials, hyperbolic shape 共see Fig. 7兲. This point is found by setting the
while the relative compaction of the coarse gravel was the high- equation of the reloaded soil stiffness and hyperbolic equation
est. In addition the gap between the pile cap and backfill soil was equal to one another and solving for ⌬int using the positive solu-
not consistent with depth for the gravels as it was for the sands. tion of the quadratic equation

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006 / 1161


Table 5. Summary of Parameters Used in Cyclic-Hyperbolic Model
Backfill Ph-ult Kmax ⌬max
sample 共kN兲 共kN/mm兲 共mm兲 Rf
Clean sand 1,081 245 38 0.88
Fine gravel 774 245 34 0.91
Coarse gravel 1,824 210 39 0.78
Silty sand 1,334 245 58 0.91
Note: Adjusted parameters for comparative purposes.

⌬ p / H⫽0.01, and enter the abscissa of Fig. 10共a兲. Using the


best-fit curve, a value of 0.45 is obtained from the ordinate
共⌬s / ⌬ p兲, resulting in an apparent soil movement 共⌬s兲 of
5.0 mm.
3. Normalize the apparent soil movement 共⌬s兲 by the pile cap
Fig. 11. Example calculation for the 26-mm continuous passive height 共H兲 to obtain ⌬s / H = 0.0045. Using this value enter
resistance the abscissa of Fig. 10共b兲. Using the average sand curve, a
value of 0.41 is obtained from the ordinate 共Kr / Kmax兲, result-
ing in a reduced soil modulus 共Kr兲 of 101 kN/ mm.
− B + 冑B2 − 4AC 4. Calculate ⌬int using Eq. 共11兲 to obtain 14.2 mm and Pint
⌬int = 共11兲 using Eq. 共12兲 to obtain 905 kN.
2A
5. Construct the cyclic passive resistance-deflection curve for
where A = KrR f / Pult; B = Kr / Km − KrR f ⌬s / Pult − 1; and this particular case as shown in Fig. 11.
C = −Kr / Km⌬s. With ⌬int known, Pint can be computed using
the equation
Passive Resistance Comparisons
Pint = 共⌬int − ⌬s兲 · Kr 共12兲
with the Cyclic-Hyperbolic Model
Within the deflection interval from ⌬s to ⌬int the stiffness Kr can
be used to calculate the available passive force accounting for the For the subsequent analyses, the input parameters for the hyper-
reloading effects and from 0 to ⌬s the passive force is taken as bolic model were adjusted to more closely match the observed
zero due to gaps produced by cyclic loading. backbone passive resistance curve shape. This adjustment facili-
The cyclic-hyperbolic model is developed based on the previ- tated comparisons between the observed cyclic passive resistance
ous deflection 共⌬ p兲. Beginning from an initial zero load state, the and the computed values using the cyclic-hyperbolic model. The
pile cap load-deflection behavior initially follows the hyperbolic adjusted values were within 10% of the observed values and 15%
curve shape. Once the load is released, the soil relaxes slightly, of the estimated log spiral values. This is equivalent to a change
and the previous maximum deflection 共⌬ p兲 will have been estab- in friction angle of 1 – 2° or the addition of 3 – 5 kPa of cohesion.
lished. Then ⌬ p is normalized by the pile cap height 共H兲 and the This adjustment seems reasonable considering the potential varia-
value of ⌬ p / H is used to enter the abscissa of Fig. 10共a兲 to obtain tion in the soil parameters. Table 5 provides the hyperbolic pa-
⌬s / ⌬ p. This ratio is then multiplied by ⌬ p to obtain ⌬s. Next, ⌬s rameters used for the comparisons. The major difference between
is normalized by H, and the value of ⌬s / H is used to enter the each analysis performed was the 共Kr / Kmax兲 curve selected to
abscissa of Fig. 10共b兲 to obtain the reload stiffness 共Kr兲 from the determine the reduction in the reloaded soil stiffness 共Kr兲 based
curve appropriate for the backfill soil type of interest. Eqs. 共11兲 on the backfill soil type. For the clean and silty sand tests,
and 共12兲 are used to define the inflection point in the backbone the best-fit sand 共Kr / Kmax兲 curve was used, while the coarse
curve defined by the hyperbolic passive resistance given by gravel test was modeled using the dense-gravel curve, and the fine
Eq. 共4兲. Alternatively, one can use Eqs. 共6兲–共9兲 in place of Fig. 10. gravel test was modeled using the medium-dense gravel curve
The next section provides an example problem using the proposed 关Fig. 10共b兲兴.
model. Figs. 12 and 13 plot the comparison of the measured and com-
puted passive resistance curves using the cyclic-hyperbolic
model. Overall, the differences between the measured and com-
Cyclic Passive Resistance Example
puted passive resistance are reasonable. The computed curves are
For this particular example, the clean sand passive resistance within 10% of the observed curves for deflections less than
curve for the 26-mm push will be calculated using the maximum 40 mm 共⌬ / H ⬍ 0.036兲 and typically within 15% for deflections
previous deflection of 11 mm and the observed values of Pult and greater than 40 mm, with the exception of the clean sand at de-
Kmax 共refer to Fig. 11兲. flections greater than 50 mm. Overall, the proposed model pro-
1. Define the hyperbolic equation. This is accomplished by in- vides an improvement on the hyperbolic model for predicting the
serting the values into the hyperbolic equation provided passive resistance under cyclic loading with sufficient lateral
below. For the clean sand test, Kmax was equal to 246 kN/ forces to form gaps between the pile cap and backfill.
mm, ⌬max / H was equal to 38 mm/ 1120 mm or 0.034, and
Pult observed was equal to 1 , 090 kN 共see Tables 3 and 4兲.
With these values, the hyperbolic Eq. 共4兲 can be used to Conclusions
define the backbone monotonic curve.
2. Normalize the previous maximum deflection 共⌬ p兲 The passive resistance results from four full-scale tests, with dif-
of 11 mm by the pile cap height of 1,120 to obtain ferent backfill soil types, were compared with several of the more

1162 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006


Fig. 12. Predicted passive resistance using cyclic-hyperbolic model Fig. 13. Predicted passive resistance using cyclic-hyperbolic model

consider the effects of cyclic loading and therefore overesti-


commonly used passive pressure theories and force-deflection mate the resistance for cyclic conditions.
relationships. As a result of this study, the following conclusions 5. A new model has been developed to account for the
are made: passive force-deflection relationship for cyclic loading con-
1. The log spiral theory provided the most reasonable estimates ditions where the stiffness of the backfill soils was degraded
of the passive resistance and was typically within 15% of the by repeated loading and a gap formed between the pile cap
measured value. The Rankine method significantly underes- and the backfill soil. This method uses a linear reloading
timated the passive resistance by 44– 67%; the Coulomb curve in combination with the hyperbolic model for virgin
method generally overestimated the lateral resistance by loading.
45– 73%. 6. The cyclic-hyperbolic model, which uses normalized curves
2. The deflection required to reach the maximum passive resis- to account for stiffness degradation and gap formation as a
tance ranged from 3.0– 5.2% of the pile cap height, which function of maximum previous loading, was reasonably suc-
agreed with previous studies for dense soils. The larger value cessful in modeling the measured passive force-deflection
of 5.2% was for the silty sand material. behavior. Although better agreement could potentially be ob-
3. The hyperbolic model proposed by Duncan and Mokwa tained with a more refined model, the accuracy of the model
共2001兲 for computing the passive force-deflection relation- is very good considering its simplicity.
ship provided the best agreement with the measured load-
deflection behavior for monotonic loadings. The Caltrans
curve initially underestimates the passive resistance by Limitations
40– 70% for deflections less than 21 mm 共⌬ / H ⬍ 1.0% 兲.
With increasing pile cap deflection 共⌬ / H ⬎ 2.0% 兲 this The cyclic-hyperbolic model developed as part of this study is
method provided a reasonable estimate for the clean sand and dependent on empirical relationships for gapping and modulus
fine gravel tests, but underestimated the passive resistance by reduction observed from the full-scale testing. Therefore several
20 and 54% for the silty sand and coarse gravel tests, respec- limitations should be considered:
tively. The U.S. Navy method typically underestimated the 1. It is likely that the passive resistance will continue to degrade
development of passive resistance relative to the measured beyond the 10– 15 cycles 共N兲 used in this study and the
behavior. model presented does not account for passive resistance deg-
4. The existing methods investigated for this study do not radation as a function of N.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006 / 1163


2. The full-scale tests were conducted using one-way loading Gadre, A. 共1997兲. “Lateral response of pile-cap foundation system and
and therefore may have experienced more degradation for seat-type bridge abutments in dry sand.” Ph.D. thesis, Rensselaer
the soils that partially filled the gap; two-way loading condi- Polytechnic Institute, Troy, New York.
tions may result in more soil flowing into the gap. James, R. G., and Bransby, P. L. 共1970兲. “Experimental and theoretical
3. The soils exhibited some apparent cohesion during this study investigations of a passive earth pressure problem.” Geotechnique,
20共1兲, 17–37.
resulting in significant passive resistance degradation due to
Krebs Ovesen, N., and Stromann, H. 共1972兲. “Design method for vertical
gap formation. The gap formation may be less pronounced
anchor slabs in sand.” Proc., Spec. Conf. Perf. Earth and Earth-Supp.
with decreasing cohesion due to increased moisture contents
Struct., 2.1, ASCE, New York, 1481–1500.
and/or a rise in groundwater levels.
Lee, K. W., and Singh, A. 共1971兲. “Relative density and relative compac-
tion.” J. Soil Mech. Found. Div., 97共7兲, 1049–1052.
Mackey, R. D., and Kirk, D. P. 共1967兲. “At rest, active and passive earth
Acknowledgments pressures.” Proc., Southeast Asian Conf. on Soil Mechanics and
Foundation Engineering, Asian Institute of Technology, Bangkok,
The Utah Department of Transportation and the Federal Highway Thailand, 187–199.
Administration provided funding for this project. Their support is Maroney, B. H. 共1995兲. “Large-scale abutment tests to determine stiffness
gratefully acknowledged. The views and recommendations ex- and ultimate strength under seismic loading.” Ph.D. thesis, Univ. of
pressed in this paper do not necessarily reflect the views of the California, Davis, Calif.
sponsors. The writers also wish to thank David Anderson, the Morgenstern, N. R., and Eisenstein, Z. 共1970兲. “Methods of estimating
Civil Engineering Department Technician who provided assis- lateral loads and deformations.” Proc., Specialty Conf. on Lateral
tance with the testing setup and instrumentation for this study. Stresses and Earth—Retaining Structures, 51–102.
Narain, R. J., Saran, S., and Nandakumaran, P. 共1969兲. “Model study of
passive pressure in sand.” J. Soil Mech. Found. Div., 95共4兲, 969–983.
References Rankine, W. J. M. 共1847兲. “On the stability of loose earth.” Philosophical
transactions of the Royal Society of London.
Rollins, K. M., and Cole, R. T. 共2006兲. “Cyclic lateral load behavior
Brinch Hansen, J. 共1966兲. “Resistance of a rectangular anchor slab.”
of a pile cap and backfill.” J. Geotech. Geoenviron. Eng., 132共9兲,
Bulletin No. 21., Danish Geotechnical Institute, Copenhagen, 12–13.
1143–1153.
Caltrans. 共2001兲. “Seismic design criteria version 1.2.” California Dept.
Rollins, K. M., and Sparks, A. 共2002兲. “Lateral resistance of full-scale
Transportation, Sacramento, Calif.
pile cap with gravel backfill.” J. Geotech. Geoenviron. Eng. 128共9兲,
Canadian Geotechnical Society. 共1992兲. Canadian foundation engineer-
711–723.
ing manual, 3rd. Ed., Canadian Geotechnical Society, BiTech,
Romstad, K., Kutter, B., Maroney, B., Vanderbilt, E., Griggs, M., and
Vancouver, B.C., Canada.
Chai, Y. H. 共1996兲. “Longitudinal strength and stiffness behavior
Clough, G. W., and Duncan, J. M. 共1991兲. Foundation engineering hand-
of bridge abutments.” Rep. to California Dept. of Transportation,
book, 2nd Ed., H. Y. Fang, ed., Chapman and Hall, New York,
Univ. of California, Davis, Calif.
223–235.
Rowe, P. W., and Peaker, K. 共1965兲. “Passive earth pressure measure-
Coulomb, C. A. 共1776兲. “Essai sur une application des règles des maxi-
ments.” Geotechnique, 15共1兲, 57–78.
mis et minimis à quelques problèmes de statique relatifs à
l’architecutre.” Mèm. acad. roy. près. Divers savants, Vol. 7, Paris Sowers, G. B., and Sowers, G. F. 共1961兲. Introductory soil mechanics and
共in French兲. foundations, Macmillan, New York.
Douglas, D. J., and Davis, E. H. 共1964兲. “The movements of buried Terzaghi, K. 共1943兲. Theoretical soil mechanics, Wiley, New York.
footings due to moment and horizontal load and the movement of U.S. Naval Civil Engineering Laboratory 共U.S. Naval C.E.L.兲. 共1962兲.
anchor plates.” Geotechnique, London, 14共2兲, 115–132. “Deadman anchorages in sand.” Rep. No. Y-F015-15-001, U.S. Naval
Duncan, J. M., and Mokwa, R. M. 共2001兲. “Passive earth pressures: C.E.L., Alexandria, Va.
Theories and tests.” J. Geotech. Geoenviron. Eng., 127共3兲, 248–257. U.S. Navy. 共1982兲. Foundations and earth structures design manual 7.2,
Fang, Y. S., Chen, T. J., and Wu, B. F. 共1994兲. “Passive earth pressures Dept. of the Navy, Naval Facilities Engineering Command,
with various wall movements.” J. Geotech. Eng., 120共8兲, 1307–1323. Alexandria, Va.

1164 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / SEPTEMBER 2006

You might also like