INTRODUCTION
A key element in the analysis of laterally loaded piles is the ultimate unit
resistance that can be exerted by the soil against the pile. This peak resistance
is often incorporated in p-y curves (soil springs), which are employed in
beam-column analyses of piles.
The ultimate lateral resistance for various situations has often been de-
fined by a combination of inferences from experimental measurements,
simplified wedge analyses, and judgment. Values derived in this manner
have been used very successfully in engineering practice (Broms 1964; Mat-
lock 1970; Reese et al. 1975). These approaches come into question, how-
ever, when the effects of complex conditions are sought, such as soil layering
and varying strength with depth, soil-pile adhesion, suction, and pile-head
fixity. A more accurate model of the soil-pile system is needed for exploring
these various effects.
Because the problem of finding the ultimate soil resistance for a laterally
loaded pile is three dimensional and nonlinear, finding a rigorous analytical
solution is highly unlikely, and finding a rigorous numerical solution is
expensive. However, an efficient, approximate solution that consistently
accounts for the various parameter effects including nonhomogeneous strength
conditions can be developed.
In this paper, the upper-bound method of plasticity is used to estimate
the collapse load of a laterally loaded pile. These results are then used to
infer the ultimate unit soil resistance along the pile length. Using this ap-
proach, parameter studies can be carried out to achieve a better under-
standing of the mechanics of pile collapse under lateral loading.
1Res. Advisor, Exxon Production Res. Co., P.O. Box 2189, Houston, TX 77252-
2189.
2Sr. Res. Specialist, Exxon Production Res. Co., P.O. Box 2189, Houston, TX.
Note. Discussion open until June 1, 1993. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on March 22, 1989.
This paper is part of the Journal of Geotechnical Engineering, Vol. 119, No. 1,
January, 1993. 9 ISSN 0733-9410/93/0001-0091/$1.00 + $.15 per page. Paper
No. 27010.
91
COLLAPSE MECHANISM
Qualitative sketches of a lateral collapse mechanism are presented in Fig.
1. Near the soil surface, a deforming conical wedge forms and is pushed up
by the pile, which itself may translate, rotate, or both. If the pile is relatively
short, the bottom may "kick out" or, for long piles, a plastic hinge may
form in the pile. For the initial development, a gap is assumed at the back
of the pile, but this restriction can be easily removed, as discussed in a later
section. Below the wedge, the soil is assumed to flow horizontally around
the pile.
To accommodate this mechanism, the soil wedge must conform to the
pile at the pile-soil interface, and must move tangentially to the rigid soil
boundary.
F
~ GA -•ote: Wedge Deforms
and Stips on Failure
Plane
OGEFA,.UR F
-4-
FLOW AROUND (b)
FAILURE
(,/)
FIG. 1. Pile Deformation Mechanisms: (c) Rigid Pile; (d) Deforming Pile
NO SOIL ~'~
DISPLACEMENT
1 .O I
~ UNIT VELOCITY
SCALE
(a)
/ x 9
(b)
DIsPLNOEM/ENT 1.0
(c) ~'~
SCALE
dissipation can be found by integrating over the wedge volume and over
the area at the wedge-rigid-soil interface. Further dissipation can occur at
the pile-soil interface. For generality, one can specify the ratio of strength
mobilization at the pile wall ranging from zero adhesion to full adhesion.
The terms for this computation are derived in Appendix III. Because of the
complexity of the integrals, they are evaluated numerically in this study.
This further allows the strength to be modeled as nonhomogenous (as well
as anisotropic, but this is not pursued here).
As mentioned, the soil resistance below the wedge is assumed to be
developed as a purely lateral flow around the pile. Thus, the plane strain
solutions developed by Randolph and Houlsby (1984) can be used directly.
That is, the dissipation is taken as
d[~ = Nps.lv~,rD dz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7)
where N_ = a dimensionless lateral unit resistance factor; vp = local lateral
pile velocity (a linear function of the pile-head virtual velocity); D = pile
diameter; and dz = differential length along the pile. This function is then
integrated along the moving part of the pile (either direction) below the
wedge, again allowing a varying strength, s,. It should be mentioned that
the transition from wedge to plane-strain deformation, as defined here, does
not rigorously satisfy the upper-bound definition, since the mechanism for
plane strain is not explicitly incorporated. Furthermore, the plane-strain
solution is not strictly valid for a rotating pile or for a nonhomogeneous
95
unnecessary complication.
The only remaining dissipation sources are plastic hinges in the pile (if
they occur). This dissipation is simply
E = MpCt .................................................. (8)
where M p = plastic moment; and [~ = pile rotational velocity about the
hinge location, which is also proportional to the virtual pile-head velocity.
For the immediate purposes here, dissipation due to soil shearing at the pile
tip is assumed small and neglected. However, this topic is mentioned again
in a later section.
The external work terms arise from pile-head loads translating and pile-
head moments rotating. The unknown pile-head load can be a single lateral
load or moment or a scale factor relating load and moment. Other known
loads or moments can be imposed. The remaining work term arises due to
the lifting of the soil weight in the wedge.
Finally, the governing equations can be written as
PARAMETERSTUDIES
Several studies were carried out to investigate conditions of interest. In
the first study, the pile was constrained to translate rigidly through the soil
(no rotation). For the purposes here, the soil was taken as weightless;
however, the effect of weight can simply be added to the weightless solution.
The soil-pile adhesion was taken to be zero. The solution gives only an
estimate of the total collapse load. To estimate unit resistance as a function
of depth, a series of solutions was obtained, in which the pile embedment
was successively increased in each solution. Note that here, the word "so-
lution" refers to an optimized or least-upper-bound solution. It was as-
sumed that the increased resistance for a longer pile was due to the cor-
responding increment of pile length, i.e., the distribution of load in the
upper region of the pile was assumed unchanged by adding pile length.
Results of this study are shown plotted in Fig. 3 for two different profiles--
uniform strength and increasing strength with depth. It is interesting to note
that they have a distinctly different shape. Further, it is fair to point out
that this difference has generally not been addressed, except possibly em-
pirically, in assigning P, values t o p - y curves. The predictions using equations
proposed by Matlock (1970) and Reese et al. (1975) are also plotted. Reese
et al. (1975) equation is an empirical modification of the planar wedge model
based on laboratory and field tests in stiff clay. For the comparison here,
ultimate values of unit resistance were taken to be 9.0s,D, consistent with
Matlock's (1970) solution and with the Randolph-Houlsby (1984) solution
96
\
(J = y~) \
L/D
6 \
Notes:
(1) Upper Bound Calculations for
smooth pile (no adhesion]
10 (2) Plane strain limit set to 9 0 for
Upper Bound
PRESENT A N A L Y S I S 13] Reese Criterion set equal to 9.O at
depth
O U N I F O R M STRENGTH ]4) Yon Mises Criterion used for Upper
12 LINEARLY INCREASING STRENGTH Bounds
L/D
6
PRESENT A N A L Y S I S
o UNIFORM STRENGTH - S M O O T H
z= U N I F O R M STRENGTH - ROUGH
Notes:
10 9 Limitingvaluesatdeptharetakento
be Np = 9 Smooth
Np 12 Rough
Consistent with Randolph and
Houlsby's exact sO~tiO~S (S. 14
and 11.94)
12
A final consideration here is the yield condition. Studies were carried out
comparing the unit resistance for both a Tresca and von Mises yield con-
dition. Stated in terms of the strength parameter, s,, the results are quite
comparable, with Tresca giving about 5 % higher values. Of course, one has
to be consistent in how the strength parameter is determined when applying
these results. In the writers' view, undrained analysis is sufficiently ap-
proximate, that there is little evidence to favor one criterion over the other.
ADDITIONAL CONSIDERATIONS
20
LIJ
~ CENTRItEUGE TEST RESULTS
C/)
9 "2 ---2::'-'-2--
9 . . . . . . . . . . . . . . . .
N 4
O Note: Theoreticalcurvesand datapointsshownfor
Z diametersof 1,24mand 2.45m
0 , r , I r I ,
0 2 4 6 8 10
DEPTH IN PILE DIAMETERS (L/D)
99
EMPIRICALMODEL
Implementation of the upper-bound model requires several numerical
integrations, and, hence, a computer code is needed. Furthermore, because
of the optimization required, some automated, nonlinear optimization pro-
gram is highly desirable. This obviously is not always convenient, and, hence,
an approximate (and limited) empirical solution based on the model is
developed herein.
The ultimate unit resistance on the pile can be written as
Np = f
, 0t ..........................................
' s, I D /
For a uniform strength soil or a linearly increasing soil with zero surface
strength, N e is clearly a function of z / D alone.
Considering the variations of Np with depth as suggested in Figs. 3 and
4, a trial empirical function of the form
was selected. In this case, N1 is the limiting value at depth (e.g., 9.0 for the
smooth pile), and (N1 - Nz) is the intercept at the soil surface (e.g., N2 =
7 for the smooth pile). The parameter { is taken to be a linear function of
h, where h = Suo/s,~D.
A combination of formal least-squares curve fitting using a nonlinear
optimization program and trail and error were used to define the function
as
= 0.25 + 0.05h )t < 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13a)
= 0.55 h -> 6 ........................................ (13b)
100
The total force required to fail a pile under lateral loading (pure translation)
was determined by both methods and compared.
Fig. 6 shows these results for five different increasing strength with depth
profiles for smooth piles constrained against rotation. The agreement is
clearly excellent. Thus, for the relatively simple, but important, strength-
profile type considered here, (12) and (13) give an accurate prediction of
the limit-analysis results.
f =
f: DSulN p dz .......................................... (14)
F
e D Suo
0 5 10 15 20 25 30 35
6 STRENGTH
PROFILE
= 10.0 CONDITIONS:
(1) CONSTRAINED AGAINST
ROTATION
8 (2) SMOOTH PILE
10
SYMBOLS DENOTE UPPER BOUND RESULTS
LINES ARE EMPIRICAL EQUATION
F = DZs,e [ 19-~
~ 1 2 +. 7 2 ( e - ~ 1 7 6 - . .................... (16)
The right-hand sides of (15) and (16) are equated, and, after simplification,
we obtain
"%.
6
Su~ I IT Ie" "~..
13- ' ~ .
_\
D--'a\
",
0"-"~\
\"~
0
\
Zx
..... ~ ~ = 0n, ~2 5
~ul/Su2 \?.v
" .\, \ ~, \
oNo o s
9 CONSTRAINED AGAINST ROTATION
.,. ,.. \\, \
9 S M O O T H PILE
102
SUMMARY
An approximate, but general, three-dimensional model has been devel-
oped using the upper-bound method of the theory of plasticity. The model
can incorporate m a n y of the complex features of the physical problem,
including n o n h o m o g e n e o u s soil strength, pile-deformation mechanism
(translation to rotation, including plastic yielding of the pile), soil-pile adhe-
sion and suction, and tip rotational resistance. The predicted effects of these
features are examined in p a r a m e t e r studies. The model predictions compare
favorably with recent centrifuge test results. The method is applicable to
other related problems, such as laterally loaded caissons. An empirical
equation is fit to analysis results to allow quick estimates for commonly
occurring soil profiles.
APPENDIX l, DERIVATIONS
First, consider the collapse mechanism as shown in Fig. 1. The radial
velocity in the wedge is assumed to be
vr = Vo 1 - cz cos 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18)
where
(~ - 1)R ~
g(r, 0) = vo rl+~ cos 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (22)
103
z
= (r~
z0 \r0_--(7-X!
r'~ ........................................... (24)
ro - r 1 + c
h ( r , O) = VoZo cos 0 \ro - R] -ro---7-- r
+ - - ............................... (25)
r
E 1 = Su'V/~ijEi] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (26a)
where repeated indices indicate summation, and
E1 = 2Sul~lmax 9. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (26b)
where I~1 = the largest absolute principle strain rate. The dissipation rate
within the wedge is then
ff=ro ~z=zo[(ro--r)/(ro-R,]fj:=~r/2)
D, = 2 =n az=o E,rdOdzdr .............. (27)
Iv,I = x/577-~
/
--- v r ~ / 1 -}- ~
( / (28)
be = I E2 d A ............................................ (31)
JA
/32 = 2 v 0 1 + \ ~ ] 1 + \ Zo /
R~ 1-c
t ~zo
9J s. ,~-1 dz .............................. (32)
r~ -- z~
Third, if the soil is totally or partially adhered to the pile at the pile-soil
interface, then dissipation occurs at a rate
E~ = ns.[v, I ................................................ (33)
where ~q = 0 for no adhesion; and ~q = 1 for full adhesion. The velocity,
v,, is the resultant of vertical and circumferential slip, i.e.
v, = X/v~ + v 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (34)
The vertical slip is given by (21), (22), and (25). The circumferential slip
is due strictly to pile m o v e m e n t and is
vc = Vo sin O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (35)
The total dissipation is then
D3 = 2 f[=~~
=0 (~
J0=0 " q s , ~ R dO dz . . . . . . . . . . . . . . . . . . . . . (36)
Fourth, for soil below the wedge, the velocity is assumed to be in a horizontal
plane. The dissipation rate along a unit pile length is then
E4 = D ' s u ' N p ' v ........................................... (37)
where N e is determined from the Randolph-Houlsby plane-strain solution.
The dissipation above the center of rotation of the mechanism is then
\c/
D6 - v~ f+=2~ (o~=sin-l(R/RX/~)Su.~V/COS20)
-}- sin2to sin2tb sin r dr dtb
Z0 J+=O J~=0
.......................................................... (43)
w h ~ = distance from the center of rotation to the pile tip, R2 =
~/R~ + R 2 (radius of the spherical surface); + = angular coordinate about
the pile centedine (in horizontal plane); and o~ = angular measure from
the pile centerline (in vertical plane).
Reese, L. C., Cox, W. R., and Koop, F. D. (1975). "Field testing and analysis of
laterally loaded piles in stiff clay." Proc., Offshore Technology Conference, Hous-
ton, Tex.
c= optimization parameter;
D = pile diameter;
/~,/~1 . . . . . /~4= local energy dissipation rates;
F = force;
h(r, 0) = function of r, 0;
le = equivalent layer thickness;
l1 = top layer thickness in two layer system;
Mp = plastic moment;
Np = nondimensional bearing capacity factor;
N1, N2 = constants;
p -- lateral soil pressure on pile;
Pmax = maximum lateral soil pressure;
R = pile radius;
Rt, R2 = radii used in derivation of end restraint;
r, z, 0 = cylindrical coordinates;
ro = radius of deforming wedge, optimization parameter;
su = undrained shear strength;
v = velocity;
vc, Vr, Vt, Vz = velocity components;
vp = pile velocity;
v0 = virtual velocity;
W, Wa = external work rates;
y = lateral pile displacement;
z0 = depth of wedge, optimization parameter;
= exponent, optimization parameter;
[3 = pile rotation rate;
= soil effective unit weight;
g, ~ij = strain rates;
-q = soil-pile adhesion factor;
h = empirical exponent;
= empirical constant; and
dO, m = spherical coordinates.
107