Vous êtes sur la page 1sur 14

Prediction of Pile Performance in Permafrost Under Lateral Load

R. K. ROWLEY'A N D G. H. W A T S O N ~
Mnckenzie V ~ l l e Pipe
y
Line Resenre11 Limited, C d g n r y , Alberta
AND

B . LADANYI
Centre d'Ingr'nierie Nordiqlre, Ecole Polytecl~nique,Montre'nl, Qlribec H3C3A7
Received April 15, 1975
Accepted August 7, 1975
In 1971 lateral pile tests were performed at Inuvik, N.W.T. to determine design values in creep
for timber and steel pipe piles. Formulation of a theoretical method for prediction of pile load
capacity from basic permafrost creep parameters was a part of this program.
In 1972 pressuremeter creep tests were performed at the same site so that test pile performance
could be predicted independently and compared with measured results. The pressuremeter tests
made and the creep parameter values obtained, the predicted pile performance curves and the
comparisons with actual test data are presented in this paper.
A Menard pressuremeter was used to determine permafrost creep parameters at an average
ground temperature of 29.2 "F (-1.6 "C). Values for the representative creep modulus and
exponents for this soil were derived from the test data. These were used for calculating
pressure-deflection curves and subgrade reaction moduli for the laterally loaded piles.
To demonstrate the applicability of the theory, the subgrade reaction moduli derived from the
pressuremeter tests were compared with those deduced directly from the pile load tests. A
reasonable agreement was found with the subgrade reaction moduli values obtained for the
timber piles. However, the steel pipe piles responded as though the modulus was much higher.
The confirmation field tests were made at a site having ice-rich permafrost whose behavior
approaches that of pure ice. To apply the basic theory generally, additional field confirmation
tests would be needed, for example, in ice-poor permafrost and covering much longer periods of
time.
En 1971 une serie d'essais de chargement lateral de pieux fut effectuee 2 Inuvik, T.N.O., ayant
comme but la determination des charges de service, affectees par le fluage du sol gele, tant pour
les pieux en bois que pour les pieux en acier de forme tubulaire. Ce programme a egalement
comport6 la formulation d'une methode theorique permettant de determiner, a partir des
parametres de fluage du pergtlisol, la capacitC portante d'un pieu sous une chargeherale.
Afin de verifier la mtthode thiorique proposee et predire le comportement des pieux d'essai
d'une maniere independante, on a effectut en 1972 sur le mkme site une serie d'essais
pressiomCtriques tant B court qu'a long terme. On prksente dans cet article les parametres de
fluage tirCs de ces essais, leur utilisation dans la prevision du comportement de pieux et les
comparaisons avec les resultats des essais de chargement actuels.
Le pressiometre Menard fut utilise pour la determination des proprietes de fluage du pergilisol,
dont la temperature moyenne etait de 29.2 "F (- 1.6 "C). Ces parametres ont CtC utilises ensuite
comme base pour le calcul des courbes charge-deflection des pieux et des modules de reactiondu
sol gele.
Pour dimontrer la validite de la thiorie, les modules de reaction calcules a partir des essais
pressiomCtriques furent compares avec ceux tires directement des essais de chargement de
pieux. On a trouve une bonne correspondance avec les modules de reaction determines partir
des pieux en bois, mais la reponse des pieux en acier fut comme si le module etait beaucoup plus
ClevC.
Ces essais de verification ont ete effectues sur un site contenant un pergClisol trts riche en
glace, dont le comportement n'etait pas t r t s different de celui d'une glace pure. Afin de pouvoir
appliquer cette theorie d'une f a ~ o nplus genirale, il sera necessaire d'effectuer d'autres essais
semblables, par exemple, dans un pergelisol avec une faible teneur en glace et en utilisant des
pkriodes de temps sous charge beaucoup plus longues.
'Presently with Standard Oil Co. of California, San Francisco, California 94105.
2Presently with Standard Oil Co. of Indiana, Napierville, Illinois 60540.
Can. Geotech. J., 12,510(1975)

ROWLEY ET AL.: PILE PERFORMANCE

Introduction
Maximizing the use of competent foundation
materials is one of the criteria used in selecting
a pipeline route. This criterion is particularly
important in permafrost terrain. As much of
the route as possible is normally located in
lower ice-content soils. In higher ice-content
permafrost, a thaw bulb could form around a
warm buried pipeline causing settlement. Similarly, a freeze bulb could form around a cooled
belowground pipeline causing heave.
Mackenzie Valley Pipe Line Research
Limited recently studied the effectiveness of a
pile-supported aboveground pipeline in avoiding disturbance of permafrost. An experimental
pipeline was installed near Inuvik, N.W.T.,
and a proposed aboveground design configuration tested (Rowley et al. 1973a). Limited data
were available on the lateral load capacity of
piles embedded in permafrost. In order to obtain design parameters, a series of lateral pile
load tests were run at the Inuvik test site.
It was recognized that design parameter
values would vary with the permafrost conditions encountered. Appropriate design values
can be determined either from a large number
of pile load tests or from theory if the loadcreep properties of the permafrost are known.
Some results of this field investigation were
shown in a recent paper by Rowley et al.
(1973b), which also contained an outline of
the proposed method for theoretical prediction
of load-deflection behavior for laterally loaded
piles in permafrost, on the basis of pressuremeter creep test data. As, however, such data
were not available at that time, no comparison
between the predicted and the observed pile

behavior could have been made. A series of


pressuremeter tests were, therefore, run at the
Inuvik test site in September 1972 to determine the load-creep properties of the permafrost. The results were used in conjunction with
the theory shown in Rowley et al. (1973b), to
calculate lateral design loads for piles.
This paper describes the test results and a
comparison between the predicted and the observed pile behavior. For more generality, the
comparison was made on the basis of lateral
subgrade reaction modulus values, i.e., between the values calculated from the pressuremeter creep data and those deduced directly
from the pile tests.

Ground Conditions
The test site was located on a relatively flat
area some 2 mi (3.2 km) north of Inuvik and
1 mi (1.6 km) east of the Mackenzie River.
The subsurface strata were defined by two
continuously sampled holes at either end of the
site. In general, the subsurface conditions were
well defined and uniform. Near surface strata
showed little change in elevation across the
site. The soil profile is generalized in Table 1.
A mixture of peat and moss, approximately
6 in. (15 cm) thick, overlay the site. During
construction, this layer was compressed to less
than 1 in. (2.54 cm) in thickness beneath the
gravel working surface.
The ground temperatures at the site were
monitored at two locations by strings of thermistors installed on two timber piles. A graph
of air and ground temperature readings, from
March 1971 to February 1972, was shown in
the paper by Rowley et al. (1973b). From the
Soil profile

Depth below ground surface


Description
of soil strata
Organics
Clayey silt (organic)
Peat in ice matrix
Grey clayey silt
with ice lenses and
layers to 2 in. (5 cm)
thick
Gravelly till

Frozen bulk density

(ft)

(m)

(p.c.f.)

&dm3)

0-0.5
0.5-1.5
1 .5-5

C0.15
0.15-0.46
0.46-1.52

90-1 10
69-80

1440-1760
1100-1280

5-9

1.52-2.75

80

1280

Water or ice content

(%I

Not sampled
1U.5
570-180

512

CAN. GEOTECH. J. VOL. 12, 1975

TABLE2. General information on five laterally loaded piles


B,
in.

Pile
T-2-L

Timber

T-3-L

Timber

S-4-L

Steel pipe

S-5-L

Steel pipe

S-6-L

Steel pipe

L,
in.

e,
in.

Zp,

in."

Ep,

lo6 o.s.i.

EpIm
lo6 Ib in.'

11.0
(28.0)
10.25
(26.1)
10.75
(27.4)
10.75
(27.4)
10.75
(27.4)

graph, it can be seen that, during the pile


tests, i.e., at the end of November 1971, the
ground temperature at the depth of 3 to 9 ft
(0.92 to 2.75 m), i.e., along the most important portion of the embedded length of piles,
was about 28.4 O F (-2.0 OC). The air temperature during the pile tests was about -4 O F
(-20 OC).
On the other hand, the ground temperature
recorded in September 1972 during and at the
level of pressuremeter tests, i.e., between 4 and
5 ft (1.20 and 1.50 m), was about 29.2 O F
(-1.56 OC), while the air temperature varied
between 28 and 31 O F (-2.22 and -0.56
OC).

Pile Installation and Testing Equipment


An array of 23 piles (18 timber and 5 steel
pipe) was installed in early September 1971.
All test piles were placed in 18 in. (46 cm)
diameter augered holes extending down to the
gravel till layer. Each annulus was backfilled
with medium to coarse sand with 10-15%
moisturc content and rodded to eliminate
bridging. The pipe piles were then filled with
sand and those designated for lateral-load tests
were filled with concrete from about 1 ft
(0.3 m) above ground surface to the top.
Of the total, only five piles, i.e., two timber
and three steel piles, were loaded laterally. To
provide the required reaction capacity, the
timber tcnsion piles were connected to battered
reaction piles by heavy gusset plates. Lateral
load was applied by means of a 100 ton
(0.89 MN) capacity hydraulically operated
ram. To reduce eccentricity effects, the ram
loading was applied to the piles through a
lubricated ball-and-socket joint.

General information on the five laterally


loaded piles is given in Table 2.
In Table 2, L denotes the embedded length
of the pile, e is the height above the ground of
the point of lateral load application, B is the
pile diameter, I,,is the moment of inertia of the
pile section, and E, is the Young's modulus of
the pile.

Load Test Results


The first four of the five piles in Table 2
were submitted to long-term loads applied in
several increments. Each load was allowed to
remain on the pile for at least 24 h unless the
data indicated that displacements had practically ceased. A load of 40 tons (356 kN) was
applied to test pile S-6-L and was kept constant
for 80 h.
Load test results for three of the piles
(T-2-L, T-3-L, and S-5-L) are plotted as time
versus deflection curves in Figs. 1, 2, and 3.
Lateral deflections shown in the figures were
taken from the dial gage located nearest to the
point of load application. Since the behavior
of the remaining two steel pipe piles (S-4-L
and S-6-L) was quite similar to that of pile
S-5-L, and since the results of the pile S-4-L
have already been analyzed earlier (Rowley
et al. 1973b), only the results obtained with
two timber piles (T-2-L and T-3-L) and one
steel pipe pile (S-5-L) will be analyzed in the
present paper.
Processing of Pile Load Test Data
As is usual in the field of creep testing, the
results of stage-loaded tests, involving various
load increments and time intervals, should be
processed in such a manner that they can be

513

ROWLEY ET AL.: PILE PERFORMANCE

3 0Oo

10

20

30

40

50

60

70

80

PO

103

TlME (HOURS)

FIG. 1. Lateral test results: pile No. T-2-L.


\

2.5t
St

50

75t

100-

Y)

I
V

-Z
z
g150C

u
U
L
3

g2w-

250

...
.-

.-. ,
GSlVII

3 0Oo

10

PAD

20

10

80

50

60

70

80

PO

IW

TlME (HOURS)

FIG. 2.

Lateral test results: pile No. T-3-L.

used not only for predicting the creep response


of the tested structural element within the time
interval covered by the tests, but can also be
extended beyond it, to reasonably longer intervals of time.

With this purpose in mind, an attempt was


made to put into an analytical form the results
of stage-loaded pile load tests shown in Figs.
1 to 3. Since the creep curves at each stage
had an approximately parabolic shape, i.e.,

514

CAN. GEOTECH. J. VOL. 12. 1975

%1

PILE
TYPE

Lu
s
E rNrGDT H~ A

TIP
DlA
EMBEDMENT

DATA

STEEL

1lo2 ~ 1
1DV4
bFT

G R O U N D TEMPERATURE

3 00;

10

20

30

40

50

60

70

,
'
O

I
0 3 ' ; 28'F

@ -

TIME (HOURS1

FIG.3. Lateral test results: pile No. S-5-L.

they were essentially of a primary creep type,


the method used for expressing analytically
these curves was very similar to that used previously for processing the data obtained in
stage-loaded pressuremeter tests (Ladanyi and
Johnston 1973).
For any laterally stage-loaded pile, the total
deflection y at the point of load application is
at any given moment equal to the sum of
instantaneous and creep deflections

In order to find the parameters, the creep


curves are first plotted in a log-log plot, where
the instantaneous deflection, taken as the one
observed at t = 0.1 h (6 min) is subtracted.
Subsequently, the instantaneous deflections are
plotted against the loads in a log-log plot, from
which the values of Qi and i are determined.
The remaining creep deflections, plotted against
time in a log-log plot, give a set of nearly
parallel straight lines from which, as in the
case of the pressuremeter test data (Ladanyi
and Johnston 1973), one can subsequently
determine the values of parameters j, k, and

In the present tests it was found that both,


yills, and y,.,
can conveniently be described
by separate power law expressions of the form

Qc.

and

Pile T-2-L (Timber)

where Q is the applied load, At is the time


interval within one particular stage, yi and y,
are unit deflections (e.g., 1 in. or 1 cm), and
5, is unit time (e.g., 1 h ) . The experimental
parameters to be determined from the tests are
the two proof loads, Qi and Q,, and the three
exponents, i, j, and k.

Pile T-3-L (Timber)

In such a manner, the following analytical


expressions were obtained (with y in inches
and Q in tons).

ROWLEY ET AL.: PILE PERFORMANCE

TABLE
3. Calculated total deflections at the ground surface, yo
T-2-L
YO,

t,
h

0.1
10
100
1 000
10 000

~n.
(cm>
0.185
(0.470)
0.331
(0.841)
0.580
(1 .473)
1.255
(3.188)
3.075
(7.811)

T-3-L

K,
p.s.i.
(M Pa)

~n.
(cm)

21 800
(150.3)
8 625
(59.5)
3 600
(24.8)
1 095
(7.6)
333
(2.3)

0.434
(1.102)
0.613
(1 .557)
0.836
(2.123)
1 .340
(3 .404)
2.477
(6.292)

YO,

Pile S-5-L (Steel Pipe)

By substituting in any particular set of the


above equations the loads and the times used
in the corresponding stage-loaded tests shown
in Figs. 1 to 3, it will be found that they describe very closely the experimentally determined deflection curves. They can, therefore,
replace the experimental data in the following
calculations.
In order to be able to compare the behavior
of the three tested piles under the same loading
conditions, the above analytical expressions
were used to calculate the deflections of the
piles under a step load of Q = 10 tons (89 kN),
and for time intervals of up to 10 000 h (417
days). It is clear, however, that in practice
extrapolation far beyond the maximum testing
time is always delicate, because, for such long
time intervals, ground temperature variations
and, in some cases, frozen soil consolidation
may cause important deviations from the behavior predicted on the basis of relatively
short-term observations.
The calculated total deflections at the ground
surface, yo, obtained by subtracting a small
elastic deflection of the pile above the ground
surface from the total deflection at the point of
load application, defined by Eqs. [4] to [9],
are given in Table 3, together with the corresponding values of K moduli, determined as
described in a previous paper (Eqs. [7] to [9]
in Rowley et al. 1973b). It should be noted

S-5-L

K,
p.s.i.
(MPa)
5850
(40.3)
3440
(23.7)
2115
(14.6)
1044
(7.2)
392
(2.7)

YO,

in.
(cm)
0.0215
(0.055)
0.0345
(0.088)
0.0515
(0.131)
0.0910
(0.231)
0.1822
(0.463)

K,
p.s.i.
(MPa)

165 000
(1 138)
91 000
(628)
49 150
(339)
20 930
(144.3)
7 450
(51.4)

that, in the present paper, K is defined as the


ratio between the applied lateral force acting
on the soil per unit length of pile and the
resulting deflection.
The K values for the three piles in Table 3
are shown graphically in Fig. 4 and a corresponding line for pile S-4-L is also shown for
comparison. It will be seen that the two timber
piles behaved quite similarly. The steel pipe
piles, however, appeared much more rigidly
embedded, giving much higher K values. The
different behavior of the two types of piles is
very probably related to different adhesive and
thermal properties of the two pile materials, as
will be discussed later.
The calculation of K moduli also showed
that the two timber piles would behave as
infinitely long (flexible) up to about 1000 h,
and as rigid for longer periods of time.

Theoretical Prediction of Load-Deflection


Curves for a Pile in Frozen Soil
In the range of allowable working loads, the
lateral load-deflection curve for a segment of a
pile is analogous to thc load-settlement curve
of a strip load of the same width pushed laterally into the soil. This similarity is regularly
used as a basis for predicting the expected lateral soil reaction from other known soil properties (e.g., Broms 1964; Matlock 1970).
The method for prediction of time-dependent pressure-deflection curves used in this
paper is based on the same similarity principle,
the only difference with the previous methods
being that the deflections were calculated on

516

CAN. GEOTECH. J . VOL. 12, 1975

---

\ECANT

K FROM

Iq-y)

CURVES

\ -

K CALCULATED F R O M

PRESSUREMETER TESTS

l0dl

10

100

IWO

10 000

TIME (HOURS)

FIG.4.

Measured and calculated time-dependent K moduli at Q = 10 tons (89 k N ) .

The symbols A and n are parameters in the


the basis of the cylindrical cavity expansion
theory instead of the usual ~oussines~-theory. frozen soil creep rate equation
The reason for this choice, as described in a
paper by Ladanyi and Johnston (1974), is
that, at small deflections, the two theories give where u,,and t , are the equivalent stress and
approximately the same results, but the former strain rate, respectively, defined as in Ladanyi
is much more convenient for handling the
( 1 972), n is the creep exponent for stress, and
creep problems.
A is defined by
As shown by Rowley et al. (1973b), the
pressure-deflection relationship needed for the
pile design can be determined by finding the where u,.is the creep modulus and e,, an arbipressure versus creep expansion relationship of trary strain rate introduced for normalization
a cylindrical cavity and by transforming it into a purposes. It should be noted that n, b, and U,
pressure versus creep settlement relationship are three basic parameters defining the creep
for a strip load. The final relationship, or the behavior of frozen soil which can be deterrequired &y curve for the pile of width B, has mined from the results of stage-loaded presthe form (Rowley et al. 1973b, Eq. [18]) :
suremeter tests, as shown by Ladanyi and
Johnston (1973).
)
(fl/2)"+
[lo] y/B = ( ~ / 8 {[1
The pressure-deflection, or the q-y curve
defined by Eq. [lo], is not only a nonlinear
function of pressure, but also a function of
where y is the displacement, q the applied time (through Eq. [ I l l ) and temperature
pressure, and po the average original ground (through the creep modulus, u,, as shown in
pressure at the level considered. The symbol T Ladanyi ( 1972) ) .
denotes the transformed time, related to the
A distinct q-y curve can, therefore, be dereal time t by
termined for each given set of time and temperature values. If such q-y curves are determined at different levels along the whole emwhere b < 1 is the creep exponent for time. bedded length of the pile for the required time

517

ROWLEY ET AL.: PILE PERFORMANCE

period and the highest probable permafrost


temperature along the pile, the design problem
of laterally loaded pile can be solved either
directly on the basis of the predicted q-y
curves (Reese 1971), or by the conventional
subgrade modulus method.
In the latter case, since the q-y plots are
nonlinear, an average (secant) lateral subgrade modulus can be determined from the
expression
where K is the subgrade modulus in the units
of stress.
Substituting Eq. [lo] into Eq. [14] shows
that the secant K modulus is a nonlinear function of the applied net pressure (q - p , ) . It is,
however, interesting to note that it becomes
independent of pressure for the combined conditions of small strains (Eq. [lo] simplifies),
and a steady state ( b = 1 ), linear-viscoelastic
( n = 1 ) creep. Equation [14] reduces then to

Since, by definition, U, is the uniaxial stress


producing uniaxial strain equal to i,t, the ratio
of the two is equal to the Young's modulus E
at the time t, so that
The value 0.85 of the ratio K / E furnished by
this analysis based on the cavity expansion
model is close to the value K / E = 0.82 obtained by Poulos (1971) by equating the
elastic continuum and subgrade reaction solutions for the displacement of a stiff fixed-head
pile with L / B = 25.
According to Vesid ( 1961) , the ratio should
also depend on the flexibility of the pile. For
an incompressible soil ( V = OS), his formula
may be written as
where K , is the modulus of subgrade reaction
for an infinite beam of width B , and P is a
flexibility number defined by

Some typical values of P can be deduced from


Broms' paper (1964, Eq. [7] and Tables 1
and 2 ) . For steel piles, in a clay with the un-

confined compressive strength greater than 2


t.s.f. (0.19 MPa), one finds P = 0.77, while
for timber piles in the same soil, P = 1.00.
Consequently, the value of Eq. [17] may be
expected to vary from 0.67 to 0.87 under such
conditions, which brackets the value deduced
from the proposed theory.
It should be noted that the foregoing analysis is valid only in the region of elastic and
creep deformations of frozen soil. As, however, at higher loads, lateral bearing capacity
failure may also occur, particularly near the
ground surface, this should be taken into
account in the determination of q-y curves. In
fact, each q-y curve should be cut off when
q attains its ultimate value, qlIlt.According to
Reesc (1 958), in saturated clays q,,,, increases
from about q,, (undrained compression strength)
at the ground surface, to 4.5q1, at a depth of
approximately three pile diameters below the
ground surface. In frozen ice-rich soils, the
same assumption can be adopted, but the eompression strength, q,,, should be made time and
temperature dependent, as will be shown later.
Another simpler way of calculating the K
value from other soil data is to take into account the well known fact that the K value has
a positive correlation with thc undrained
strength q,, of cohesive soil materials. The
relationship most often used between K and
q,, (Terzaghi 1955; Broms 1964; Davisson
1971) is

K 33q,,
Although the validity of this relationship has
not yet been checked for frozen soils, it is considered, nevertheless, that it may be applicable
to frozen ice-rich fine-grained soils with a
negligible internal friction.
In fact, the ratio K/q,, can be written as
~ 9 1

Noting that K / E
0.85, while, for a linearelastic-plastic material, E/q,, = 1/ef, where
cf is the failure strain, one can also write Eq.
[19] as
The latter form is more appropriate for creep
problems, because 6, usually varies very little
with time and temperature. It is interesting to
note that in the present case where cf deduced

518

CAN. GEOTECH. J. VOL. 12, 1975

from pressuremeter tests was about 0.025, Eq.


[21] gives K = 34q,,, which is close to the
originally proposed relationship.
When calculating K for a soil profile in
which the soil strength varies with depth, one
should take care to select a value of K that
reflects properly the average properties of the
soil within the zone most affecting the pile
behavior. The subgrade reaction theory shows
that this most important zone, for a freeheaded pile, extends from the ground surface
to the depth I,,, called also significant length of
pile, which is given by (Broms 1964)
where EJ,, is the stiffness of the pile section
and K the lateral subgrade modulus.
Since in frozen soils the value of K decreases
with time, lo will show a corresponding increase with time. This will have two effects:
first, with time, the pile will change from a
flexible to a more and more rigid behavior,
and, second, it will be influenced by increasingly deeper soil strata. Eventually, for a very
long time interval, the pile will end by acting
as a rigid pole and will be affected by the soil
along its full embedded length L.
For piles embedded in permafrost, there is
an additional complication in the determination
of a relevant K value, which is due to the
seasonal variation of soil temperature in the
upper soil strata, especially in the active zone.

Measurement of Creep Properties


In order to use the theoretical design approach for frozen soils, creep properties had
to be measured, and the results obtained from
the theory compared with full-scale field tests.
The field testing was carried out using standard
pressuremeter equipment. The testing technique and interpretation methods followed a
recently developed procedure (Ladanyi and
Johnston 1973).
In total six pressuremeter tests were carried
out at the Inuvik test site in September 1972.
Of these, two were standard short-term tests
with 2 min per stage, three were stage-loaded
creep tests with 15 min per stage, and one was
a long-term creep test with 10 h under load in
the last stage. All the tests were performed at
a depth of about 5 ft (1.50 m ) below the
ground surface.

Based on existing procedures (Ladanyi and


Johnston 1973), the following frozen soil
parameters were determined from the tests.
Short-term Parameters
Pressuremeter Young's modulus: 7.8 < E
< 12.8 k.s.i. (54 < E < 88 MPa)
Tensile strength: 92 < T , < 128 p.s.i. (0.64
< T , < 0.88 MPa)
Cohesion (for @ = 0 ) : 99 < c < 170 p.s.i.
(0.69 < c < 1.18 MPa)
Creep Parameters
Creep modulus (for i,. = 10-"in-l):
43.3 <a,. < 45.7 p.s.i. (0.30 to 0.32 MPa),
average a,. = 44.9 p.s.i. (0.31 MPa)
Exponent n: 3.00 < n < 3.89 (average:
3.38)
Exponent b: 0.85 < b < 0.87 (average:
0.86)
All these parameters are valid for a soil temperature of 29.2 O F (-1.56 OC), as recorded
during the tests.
Since the average soil temperature along the
piles during the pile tests was actually lower,
i.e., 28.4 OF (-2.0 OC) , the value of the creep
modulus should be temperature corrected before being used in connection with the pile
tests.
As shown in Ladanyi (1972), for small
temperature variations, the temperature correction of the creep modulus v,. can be made
by using the ratio
where B1 and 8:, are two different temperatures
expressed as the number of degrees Celsius
below the frcezing point, and acOl and ocoa are
the corrcsponding creep moduli.
For an ice-rich soil, the following form of
temperature function can be used (Ladanyi
I972),
where 0 is defined as before, n is the creep exponent for stress, U is the apparent activation
energy of the frozen soil, and R is the universal
gas constant.
The ratio U/R, which has the units of temperature, varies with the type of frozen soil.
10 000 OC, which
For polycrystalline ice U/R
is most likely also valid for the very ice-rich
soil investigated at Inuvik.

ROWLEY ET AL.: PILE PERFORMANCE

ULTIMATE LATERAL

rRES'ST

0 01

01

10

DIMENSIONLESS DEFLECTION, y l B

FIG.5. Isochrono~~s
pressure - deflection curves, calculated from pressuremeter test results.

With this U/R value, and taking into account that 6 << 273 OC, Eq. [24] may be
written as

0.86, and i,.= lo-: min-I , one gets (with


in p.s.i. and t in min) :
n,,p
= 2761-0.25-1
[281

[251

The plane strain compression strength, q,,,,,, is


then obtained by multiplying u,.~by 2 / G ,
giving

f(6)

exp(8/7.4n)

and Eq. [23] as


[26]

l n ( ~ , ~ ~=
/ ~(82, ~-~01)/7.4n
)

For the observed temperature difference of


2.0 - 1.56 = 0.44 OC and n = 3.38, one gets
from Eq. [26] a temperature correction factor
of 1.018, i.e., the creep modulus during the
pile tests was about 2 % 'higher than during the
pressuremeter tests.
In addition to the creep parameters which
can be used in Eq. [ l o ] for calculating the
pressure-deflection curves, the same parameters also allow an estimate to be made of the
time-dependent equivalent strength of the
frozen soil, v,.~, according to the equation
(Ladanyi and Johnston 1973)

if the concept of a constant strain at failure,


is adopted. Substituting in Eq. [27], U, =
45.7 p.s.i. (0.32 MPa), temperature corrected
value), cpf = 0.025 (from pressuremeter-determined stress-strain curves), n = 3.38, b =
E,.~,

~291

u,.~

q ,,,,,, = 318.7t-0.".1

where q
is in p.s.i. and t in min. On the
other hand, substituting the same creep parameters in Eq. [lo] yields
,301

y,B = 0.393 ([I + I .27


X 10-11 t o . s ~(q
; - P,,):3.:3~~2
- 1}

where q and p,, are in p s i , and t is in min.


Combining Eqs. [29] and [30], the isochronous pressure-deflection curves shown in Fig. 5
were obtained. The three curves correspond to
10, 100, and 1000 h, respectively. Each curve
is seen to reach its respective ultimate strength
of 4.5q11at the deflection of about 0.6B, which
means that, for normal deflections, it would not
be necessary to consider the ultimate strength
cutoff.
Close to the ground surface, where the lateral resistance is close to q,,, the cutoff should
be taken into account. Any of the isochronous

520

CAN. GEOTECH. J. VOL. 12, 1975

q-y curves can be used as a basis for solving


the pile problem by the finite difference method
(Reese 1971 ). The calculation will furnish
time-dependent deflections and bending moments in the pile from which the design load
can be determined.

Calculation of K Moduli and Comparison with


the Pile Test Data
The creep information can, however, also
be used for .determining the values of K modulus to be used in the conventional design
method. In order to see whether the moduli
so determined correspond to those deduced
directly from pile tests, the values of K were
calculated from Eq. [14] for a lateral load Q =
10 tons (89 kN), which was assumed to be
distributed uniformly over the significant length,
lo, of the pile, i.e., ( q - yo) Q/Blo. For the
two timber piles, for which average lo = 3, the
value of ( q - yo) amounts to about 50 p.s.i.
(0.34 MPa).
The secant K values, calculated from the
pressuremeter data in such a manner, have
been found to range from 36 950 p.s.i. (255
MPa) (at 10 h) to 674 p.s.i. (4.6 MPa) (at
1000 h ) . They are seen in Fig. 4 to be of the
same order as those deduced from the pile
tests. As, however, such secant K values are
very sensitive to the selected average pressure
(q - yo), this method is not recommended,
and a direct use of q-y curves should be preferred.
A simpler and more appropriate method for
estimating the K values from other soil properties is by using Eq. [19]. Substituting into the
latter the q,, value from Eq. [29] yields

where K is in p.s.i. and t in min. The resulting


K versus t line plotted in Fig. 4 is seen to be
located at the lower limit of K values deduced
from the two timber pile tests, and is more
than one order of magnitude lower than K
values deduced from the two steel pipe pile
tests.
There may be several possible reasons for
this discrepancy, such as:
( 1 ) The formula K = 33q,, takes into account
only the soil reaction in front of the pile and
neglects all eventual contributions from adfreeze forces acting on lateral and back sur-

faces of the pile. A limited amount of this


effect has been observed in lateral pile tests in
ordinary soils (Baguelin and JCzCquel 1972),
but it may be much more important in a frozen
soil due to its high adfreeze and tensile strength.
For small displacements, where the tensile
strength is not yet overcome, it may not be
exaggerated to assume a double value of K,
i.e., K = 66q,,. The corresponding line shown
in Fig. 4 is seen to fall in the middle of the
K lines from the timber pile tests. It is interesting to note that the short-term K values,
obtained by substituting the range of shortterm pressuremeter moduli into Eq. [16], are
seen in Fig. 4 to fall well between the two K
lines deduced from q,, values. Clearly, for the
same reasons as mentioned before, these
E-based K values should probably also be
doubled.
( 2 ) All q,, values used as a basis for K determination up to now have been calculated for
an average ground temperature of -2 "C, as
recorded during the pile tests. While this may
be right for the timber piles, it is not necessarily so for the steel pipe piles. For the latter,
due to their high thermal conductivity, it is
more likely that the temperature at the contact
with the soil was closer to the air temperature
than to the average temperature of the surrounding ground.
Since, as mentioned before, the air temperature during the pile tests was about -20 "C,
this may justify a different temperature correction in the q,, formula. Taking, as an upper
limit, Q2 = 20 "C, Eq. [26] gives, for 63 = 1.56
OC, a temperature correction factor of 2.09.
Consequently, at that temperature,
q,, = 666t-0.254
[321
and the corresponding (double) K value is
then
[33] K = 66 x fjfj6t-0.25.4 = 43 956t-o.=4
where t is in min and q,, and K in p.s.i. The
corresponding K line, assumed to be valid for
the steel pile, is seen in Fig. 4 to be still nearly
one order of magnitude too low.
( 3 ) While some additional explanations may
be found for the relatively low K values
deduced from the pressuremeter test data,
such as an eventual temperature disturbance
during the tests, or the fact that the piles were

521

ROWLEY ET AL.: PILE PERFORMANCE

actually surrounded by a frozen sand annulus


which was stronger than the true natural soil
tested by the pressuremeter, one should, nevertheless, be able to explain in the first place
large differences in K values deduced from the
timber and the steel piles, respectively.
Since the laterally loaded pile theory used
for calculating the K values shown in Fig. 4 is
supposed to take into account properly all the
effects of different pile stiffnesses, the K values
deduced from all the pile tests should have
been of the same order of magnitude. This not
being the case, one is then forced to conclude
either that certain assumptions in the theory are
not quite correct (e.g., about a constant value
of K with depth), or that actual steel pile stiffnesses were much higher than the assumed
values. This latter possibility was checked by
calculating the increase in the steel pile stiffness
due to their filling with sand and, their upper
part, with concrete. The calculated increase in
stiffness was, however, much too small to explain their observed behavior.
These questions cannot be answered on the
basis of present, still very incomplete, experience with the laterally loaded piles in permafrost. For finding proper answers, other similar studies should be carried out in future
under well controlled conditions involving various frozen soils, various methods of pile
installation, and, in particular, a number of
true long-term pile tests covering at least one
whole year.

Time to Failure of Piles


After a long time interval, which from these
pile tests was found to be about 1000 h ( 6
weeks), the piles start behaving practically as
rigid poles (i.e., L/lo < 1S ) . For such a rigid
pile behavior, the ultimate load at which the
soil will start failing everywhere at the contact
with the pile can be estimated by assuming,
according to Broms (1964), that the soil
reaction is zero down to 1.5 times the pile
diameter B, and that it can attain 4.5qlIB
below that depth. Assuming a uniform distribution of this ultimate soil reaction along the
embedded portion of the pile, as shown in
Fig. 6a, the ultimate lateral load, QlIl,, can be
calculated from the moment equilibrium to give

while the depth of the center of rotation of the


pile below the ground surface is found to be
equal to (1.5B
a ) , where

in which

L and e denoting as before the total length of


the pile and the height of the point of load
application above the ground surface, respectively.
For example, for pile T-2-L, one has B =
11 in. (28 cm), L = 105 in. (265 cm), e =
24 in. (61 cm), from which L' = 88.5 in.
(225 cm), e' = 40.5 in. (103 cm), a = 55 in.
(140 cm) and, from Eq. [34],
A similar calculation for pile T-3-L gives

Taking the average for the two piles and substituting q,, from Eq. [29] yields
with Qlllt in kips and t in minutes (1 kip =
1000 Ibs = 44.5 k N ) Eq. [40] is represented
by the lower line in Fig. 6. The time to failure
is then
valid for t > 60 000 min (6 weeks).
For example, for tf = 1000 h, Eq. [40]
would predict for the two timber piles a failure
load of
Quit = 347.2(60 000) -K*"

= 21.2 kips =
10.6 tons (94.3 kN)

which seems correct in comparison with the


values of 15 and 17.5 tons (134 and 156 kN)
found for these piles after a total loading time
of less than 100 h.
If a similar calculation is made for the pipe
pile S-5-L (upper line in Fig. 6 ) , one finds
that, for tf = 1000 h, and a soil temperature at
the contact with the pile of -20 O C , the ultimate
load would be about 21 tons ( 1 87 kN). The
pile actually sustained 40 tons (356 kN) after

522

CAN. GEOTECH. J. VOL. 12, 1975

(a)
01

lo1

-DEFiECliON, SOIL REACTION AND BENDING


MOMENT DISIRIBUTION FOR A SHORT FREE
HEADED PltE lAF1ER BROMS IVbdI
I

lo2

I o3

I o4

1o5

TIME TO FAILURE (HOURS)

FIG.6. Time-dependent ultimate lateral resistance for Inuvik piles, predicted from pressuremeter tests.

modulus K to be used in the design by the


conventional method. Finally, one can also use
time-dependent strengths of frozen soil to
determine the ultimate load or the time to
failure of the piles.
2. The calculated K moduli were compared
with those deduced directly from the pile tests.
A fair agreement was found with K values
obtained for the timber piles. The steel pipe
piles, however, responded as if the soil modulus were much higher, for which some tentative
explanations are given. A good agreement was
obtained between the calculated and the
observed ultimate loads for all the piles.
3. This study shows, as many studies of the
same problem in unfrozen soils have shown
before, that the prediction of the behavior of
laterally loaded piles directly from the soil
Conclusions
data is not a simple problem. In unfrozen soils,
1. This analysis has shown how, in ice-rich many years of systematic research and field
permafrost, the basic data necessary for the observations have produced a large amount of
design of laterally loaded piles can be deduced valuable data which enabled new and imfrom the results of pressuremeter creep tests. proved design methods for laterally loaded
The data may be given either in the form of piles to be developed.
time- and temperature-dependent pressureIn frozen soils, the behavior of which differs
deflection curves, to be used as a basis of a in many aspects from that of unfrozen soils,
computer design method, or, alternatively, very few complete studies of this problem have
they can furnish the time- and temperature- been made up to date, and proper experimendependent values of the lateral subgrade tal background for a final assessment of any

80 h. Figure 6 shows that a good agreement


between predicted and observed ultimate
lateral resistances can be obtained by this
method.
This method enables the design load Q to
be selected in such a manner that it includes a
required safety with respect to the time of
failure. For example, according to Eq. [41], at
a load of 6 tons (53 kN), such a pile would
last for about a year, while at 3.4 tons (30
kN), it would not fail before 10 years. Nevertheless, for long-term extrapolation, discrepancies with the real pile behavior may be large
because of still little known effects of the
frozen soil consolidation and the cyclic ground
temperature variation.

ROWLEY ET AL.: PILE PERFORMANCE

design method for laterally loaded piles is still


lacking. This paper has shown one possible
method for designing such piles and for obtaining necessary field data. fi should be noted,
however, that the check of the theory against
the true pile behavior was limited only to an
ice-rich permafrost, and to one particular
method of vile installation: frozen backfill. If
the method is to be made applicable more
generally, it should be checked also in the case
of an ice-poor permafrost, especially of a frictional type, and for some other methods of pile
installation. In addition, to check the longterm prediction of pile behavior, long-term
tests covering periods of at least one whole
year should be performed in the future.

523

2nd Int. Conf. Permafrost, Yakutsk, North Am.


Contr. pp. 712-721.
TERZAGHI,
K. 1955. Evaluation ofcoefficients of subgrade
reaction. Geotechnique, 5, pp. 297-326.
V E S I ~A.
, S. 1961. Bending of beams resting on isotropic
elastic solid. Proc. Am. Soc. Civ. Eng., 87(EM2), pp.
35-53.

Appendix - Symbols
a
length defined by Eq. [35]
A
magnitude defined by Eq. [I31
b
creep exponent for time
3
width or diameter of the pile
c
short-term cohesion of frozen soil
e
height above the ground surface of the
point of lateral load application
E
short-term Young's modulus of frozen
soil
Young's
modulus of the pile
E,
BAGUELIN,F., and JEZEQUEL, J. F. 1972. Etude
i, j, k exponents in Eqs. [2] and [3]
expirimentale d u comportement d e pieux sollicitis
moment of inertia of the pile section
Horizontalement. Ann. Inst. Tech. BLtim. Trav. PubI,
lics, 297, pp. 154-204.
K
modulus of subgrade reaction for a strip
BROMS,B. B. 1964. Lateral resistance of piles in cohesive
load or pile of width B
soils. Proc. Am. Soc. Civ. Eng.,90(SM2), pp. 27-63.
K,
modulus of subgrade reaction for an
DAVISSON,M. T. 1971. Lateral load capacity of piles.
infinite strip load
Highw. Res. Rec. 333, pp. 104-1 12.
LADANYI,
B. 1972. An engineering theory of creep of fro1,
significant length, defined by Eq. [22]
zen soils. Can. Geotech. J. 9, pp. 63-80.
L
embedded length of the pile
LADANYI,
B., and JOHNSTON,G. H . 1973. Evaluation of
n
creep exponent for stress
in-situ properties of frozen soils with the pressuremep,
original lateral ground pressure
ter. Proc. 2nd Int. Conf. Permafrost, Yakutsk, North
Am. Contr. pp. 310-318.
q
lateral pressure
1974. Behavior of circular footings and plate anundrained compression strength of frozen
q,,
chors embedded in permafrost. Can. Geotech. J. 11,
soil
pp. 531-553.
Q
lateral load
MATLOCK,H. 1970. Correlations for design of laterally
Q,,, ultimate lateral load
loaded piles in soft clay. 2nd Ann. Offshore Technol.
Conf., Houston, Tex., Pap. No. OTC 1204, pp.
R
universal gas constant
1-577-1-588.
t
time
MENARD,L., BOURDON,G., and GAMBIN,M. 1969.
short-term tensile strength of frozen soil
T,
MCthode generale de calcul d'un rideau ou d'un pieu
U
apparent activation energy
solliciti horizontalement en fonction des resultats
pressiometriques. Sols-Soils, No. 22-23, pp. 16-29.
lateral deflection of the pile
y
P o u ~ o s ,H. G. 1971. Behavior of laterally loaded piles:
flexibility number, defined by Eq. [18]
I-single piles. Proc. Am. Soc. Civ. Eng., 97(SM5), 3! ,
2
,
arbitrary strain rate
pp. 711-731.
P,
equivalent strain rate
REESE,L. C. 1958. Disc. Trans. Am. Soc. Civ. Eng. 123,
0
number of degrees Celsius below 0 "C
pp. 1071-1077.
1971. The analysis of piles under lateral loading.
a,
creep modulus
Proc. Symp. Interaction Struct. Found., Univ. of
creep modulus at temperature 0
a,,
Birmingham, Birmingham, Engl.
a
,
equivalent stress
ROWLEY,R. K., WATSON,T. M., WILSON,T. M., and
T
transformed time, defined by Eq. [I 11
AULD,R. G. 1973n. Performance of a 48-inch warmoil pipeline supported on permafrost. Can. Geotech. J.
v
Poisson's coefficient
10(2), pp. 282-303.
NOTE: Dot over symbols denotes time rate.
ROWLEY,R. K., WATSON,G. H., and LADANYI,
B. 1973b.
Subscript f denotes failure.
Vertical and lateral pile load tests in permafrost. Proc.

Vous aimerez peut-être aussi