Vous êtes sur la page 1sur 13

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

net/publication/237370658

Prediction of the horizontal load-displacement curves of pile groups based on


the results of single pile tests

Article  in  Canadian Geotechnical Journal · January 2011


DOI: 10.1139/cgj-37-5-951

CITATIONS READS

2 152

1 author:

Antonio Sousa Coutinho


National Laboratory for Civil Engineering
4 PUBLICATIONS   24 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Antonio Sousa Coutinho on 31 May 2014.

The user has requested enhancement of the downloaded file.


Color profile: Disabled
Composite Default screen

951

Prediction of the horizontal load–displacement


curves of pile groups based on the results of
single pile tests
António G.F. de Sousa Coutinho

Abstract: This paper presents the prediction of horizontal load–displacement curves of pile groups based on the results
of single pile tests. Although the same basic model is employed, two different approaches are taken: one assumes soil
to be linear elastic–plastic, and the other assumes it to be elastic nonlinear. The model is calibrated on the basis of the
results of a full-scale single pile test. Special emphasis is placed on model calibrations, since the success of any
prediction method depends on a careful characterization of the soil. Some new approaches for determining the soil
parameters are presented. Two methods for predicting load–displacement curves, one from each model approach, are
then proposed and discussed. Special emphasis is placed on group efficiency in the elastic–plastic method and on the
boundary conditions of the single pile and the pile group in the elastic nonlinear method. Using the soil characteristics
from the model calibrations, the load–displacement curves for a given pile group are then predicted. These predictions
are compared with the results of a full-scale pile group test carried out at the same site as that of the single pile test.
Agreement between the predictions and the test results tends to validate the methods proposed.

Key words: displacement predictions, pile groups, model calibration, pile tests.

Résumé : Cet article présente la prédiction des courbes de charge horizontale vs déplacement de groupes de pieux
basée sur les résultats d’essais sur pieux simples. Quoique le même modèle fondamental soit utilisé, deux approches
différentes sont suivies: une suppose que le sol est linéaire élasto-plastique, et l’autre suppose qu’il est élastique non-
linéaire. Le modèle est calibré sur la base des résultats d’un essai sur pieu simple à pleine échelle. Un soin particulier
est accordé aux calibrages du modèle puisque le succès de quelque méthode de prédiction que ce soit dépend d’une
caractérisation attentive du sol. L’on présente de nouvelles approches pour la détermination des paramètres du sol. L’on
propose et discute deux méthodes pour la prédiction des courbes charge/déplacement, soit une pour chaque approche de
modélisation. Une attention spéciale est accordée à l’efficience de groupe dans la méthode élasto-plastique et aux con-
ditions aux frontières du pieu simple et du groupe de pieux dans la méthode élastique non-linéaire. En partant des
caractéristiques du sol fournies par les calibrages du modèle, les courbes charge/déplacement pour un groupe de pieux
donné sont alors prédites. Ces prédictions sont comparées aux résultats d’un essai sur groupe de pieux à pleine échelle
réalisé sur le même site que celui de l’essai sur pieu simple. La concordance entre les prédictions et les résultats
d’essai tendent à valider les méthodes proposées.

Mots clés : prédictions de déplacement, groupes de pieux, calibrage de modèle, essais sur pieu.

[Traduit par la Rédaction] Sousa Coutinho 962

1. Introduction subject to a large margin of uncertainty. Therefore, the


Laboratório Nacional de Engenharia Civil (LNEC) was
A new highway plan was proposed in Portugal at the be- asked to schedule and conduct a research programme on the
ginning of the 1980s. This plan included the construction of behaviour of horizontally loaded piles. The immediate pur-
several bridges across large, deep alluvial valleys. The foun- pose of the programme was the validation of the design cri-
dations were to be reinforced concrete piles, and in some teria that had been adopted (Guedes de Melo 1987).
cases horizontal loads were critical to the designs. When the However, inasmuch as the opportunities for carrying out
geotechnical studies began in the mid-1980s, it was decided tests of this kind are few and far between, it was decided
to conduct experimental studies on the behaviour of horizon- that the scope of the research programme should be as wide-
tally loaded piles, given that analytical design methods are ranging as possible. Hence, in addition to the short-term ob-
jective, some other medium- and long-term aims were also
Received December 22, 1998. Accepted January 19, 2000. defined (Sousa Coutinho 1995). Briefly, the research
Published on the NRC Research Press website on September programme had three main goals, which were to be achieved
28, 2000. sequentially: (i) to develop reliable equipment, especially for
A.G.F. de Sousa Coutinho. Geotechnics Department, the instrumentation of cast-in-place concrete piles, with which
Laboratório Nacional de Engenharia Civil, Av. do Brasil, 101, to run horizontal load tests on piles; (ii) to develop reliable
1700-066 Lisboa, Portugal. methods for data reduction, including the load distribution

Can. Geotech. J. 37: 951–962 (2000) © 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:18 PM
Color profile: Disabled
Composite Default screen

952 Can. Geotech. J. Vol. 37, 2000

Fig. 1. Schematic cross section of the test site at Arabe bridge. cu, undrained shear strength; NSPT, SPT blows; z, depth; γ, unit weight.

among the piles in a given group; and (iii) to develop sim- the south of Portugal. Figure 1 shows a schematic of the soil
plified methods for designing piles and pile groups based on profile at the test site. In the first test a trial pile was loaded
experimental data and evidence. up to failure against a pair of capped trial piles. In the sec-
The first goal has been addressed elsewhere, first in the ond test, two groups of 16 capped piles were pulled towards
form of an overview (Toco Emílio et al. 1990) and more re- each other. These piles were service piles, since they formed
cently in a more comprehensive format (Sousa Coutinho et part of the bridge foundations. The schematic arrangements
al. 2000). The second goal has been submitted for publica- of the pile tests are shown in Figs. 2 and 3. All the piles
tion elsewhere (Sousa Coutinho, in preparation1), and the were bored and cast in place with an outer steel tube, which
third goal is the subject of this paper. was withdrawn as the hole was filled with concrete. They
were approximately 48 m long and had a nominal diameter
of 1.1 m.
2. Research programme
2.2. Prediction of load–displacement curves
2.1. Test layout Once the research programme had been scheduled, it was
Seven tests on the pile foundations of three bridges were necessary to propose a simple method for predicting the
carried out within the scope of this research programme. The load–displacement curves of the pile groups, in which the
piles that were to be tested were instrumented with strain characteristics of the soil could be determined from single
gauges, and the horizontal displacements and rotations were pile load tests. The method had to be based on simple yet re-
measured at surface level. Both trial and service piles were alistic models.
tested, and single trial piles were loaded up to failure. De- The need for a balance between the inclusion of the fun-
tails of the experimental aspects of this research programme damental aspects of pile–soil behaviour on the one hand and
have been published elsewhere (Sousa Coutinho 1995; Sousa simplicity on the other led to the selection of the Davies and
Coutinho et al. 2000). Budhu (1986) model. The reasons for this choice were as
The results of only two of the tests are presented in this follows: (i) the pile is considered to be a solid beam; (ii) the
paper. These tests were carried out at the site of the planned soil is assumed to be linear elastic–plastic, and local yield-
bridge over the River Arade near Portimão in the Algarve in ing, pile–soil separation, and lateral friction–adhesion are all
1
Sousa Coutinho, A.G. Data reduction of full-scale horizontal load tests on cast-in-place concrete piles and pile groups. Submitted for publi-
cation.

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:22 PM
Color profile: Disabled
Composite Default screen

Sousa Coutinho 953

Fig. 2. Schematic showing the position of trial test piles. d, pile diameter.

Fig. 3. Schematic showing the position of pile groups.

taken into consideration; (iii) the soil is characterized by one


ample of moduli determination, the following exposition in-
elastic parameter and one plastic parameter; (iv) the elastic
troduces the subject in a more detailed and methodical man-
and post-elastic solutions are presented as equations that can
ner, and also proposes some features that are new in relation
be applied simply; and (v) these solutions are applicable to
to the contents of past publications.
overconsolidated clays (Davies and Budhu 1986),
cohesionless soils (Budhu and Davies 1987), and soft clays 3. Model calibration
(Budhu and Davies 1988).
The Davies and Budhu (1986) model is used to obtain 3.1. Single pile test results
the soil parameters from a single pile test. According to Figure 4 represents the relationships between the loads
Poulos and Davis (1980), there are two possible approaches applied to, and the induced head displacements and rotations
by which it is possible to combine the pile test results with of, the single pile computed by integrating the strains mea-
those from the theoretical model. The first, which is more sured (Sousa Coutinho 1995, in preparation1). The straight
accurate from a theoretical point of view, involves the de- lines adjusted to the first three points of each graph (with no
termination of the initial tangent modulus from the initial origin intercept value) reveal that the relationships shown in
linear section of the experimental load–displacement and Fig. 4 are linear up to loads of about 100 kN. At that load
load–rotation curves taken at the surface level of the soil, the displacement is about 0.9 mm and the rotation is about
following which the soil yielding is taken into consider- 57 sexagesimal seconds.
ation. The second, which is more intuitive than the first, in-
volves the determination of secant moduli from the 3.2. Variation of soil modulus with depth
aforementioned experimental relationships, which are sub- The soil modulus can vary with depth in accordance with
sequently used in association with the elastic theory. In this any of the possible functions, but experience has shown that
approach the effects of soil plasticity are not explicitly con- almost all soils obey either a constant or a linear increase in
sidered, although the moduli values which are obtained do distribution as the depth varies. As a general rule, the former
include them. is applicable to overconsolidated clays, whereas the latter
The mechanical parameters of the soil referred to in the applies to sands and normally consolidated clays. However,
Davies and Budhu (1986) model were to be obtained from a when both the load–displacement and the load–rotation rela-
single pile test carried out at the site of the bridge over the tionships are known, it is possible to check which of the two
River Arade. Although Poulos and Davis (1980) offer an ex- is more applicable at a given site.

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:26 PM
Color profile: Disabled
Composite Default screen

954 Can. Geotech. J. Vol. 37, 2000

Fig. 4. Surface displacement and rotations of the tested single pile. H, horizontal load; us, pile head displacement in linear elastic soil;
θs, pile head rotation in linear elastic soil.

The case of a constant modulus with depth is considered Fig. 5. Graphical solution of equation system [6]. Es, soil modu-
first. For flexible piles, under elastic and linear conditions, lus; Kb, pile–soil relative stiffness.
surface displacements us and rotations θs may be related to
loads H using the following equations (Davies and Budhu
1986):

H  e 
[1] us =  I uH + I uM 
Esd  d 

H  e 
[2] θs =  IθH + IθM 
Esd 2  d 

where Es is the soil modulus; d is the pile diameter; e is the


load eccentricity; and IuH, IuM, IθH, and IθM are flexibility
factors defined as

− −
2 5

[3] IuH = 1.3K b 11 ; IuM = IθH = 2.2K b 11 ;


8

IθM = 9.2K b 11

in which  −
2
−5
(1) → Es = 1.313 × 105 K b 11 + 6.061 × 104 K b 11

E 
[4] Kb = 5 −
5
5 −
8

Es [6]  (2) → Es = 6.579 × 10 K b 11 + 7.504 × 10 K b 11


 −1
 (3) → Es = 3.84 × 10 K b
7

is the relative stiffness, and E is the pile modulus. For rein- 


forced concrete piles, it is useful to compute an equivalent
modulus from the flexural stiffness of the piles, (EI)p:
The system of equations can be graphically solved by taking
( EI) p several values for Kb (usually between 10 and 106) and com-
[5] E= puting the resulting Es values, as shown in Fig. 5. It can be
πd 4 concluded that Es assumes very different values when it is
64 computed from (1) and (3) as opposed to (2) and (3).
For the case of a linearly increasing depth modulus, Es =
For the case of the pile tested, E = 38.4 GPa, d = 1.1 m, and mz (where m is the rate of the increase in soil modulus with
e = 0.3 m. depth z), the equations for surface displacements and rota-
If the values obtained in the previous section (i.e., H = tions are as follows (Budhu and Davies 1988):
100 kN, us = 0.9 mm, and θs = 57 s) are taken and substi-
tuted in the appropriate equations [3] in [1] and [2], and if
H  e 
the values of E, d, and e are used in [1], [2], and [4], the fol- [7] us =  I uH + I uM 
lowing three ratios are obtained for Es and Kb: md 2  d 

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:28 PM
Color profile: Disabled
Composite Default screen

Sousa Coutinho 955

Fig. 6. Graphical solution of equation system [11]. m, rate at Table 1. Computation of the values for determination of the
which soil modulus increases with depth. plasticity parameter.
H (kN) uc (mm) us (mm) Iuc h H/d3 (kN/m3)
20 0.2 0.2 1 14.68 15.03
40 0.3 0.4 1 14.68 30.05
80 0.8 0.8 1 14.68 60.11
115 1.2 1.1 1.14 20.84 86.4
220 3.4 2 1.69 44.62 165.29
330 6.8 3 2.26 69.02 247.93
375 8.4 3.4 2.45 77.52 281.74
427 13 3.9 3.33 115.63 320.81

Fig. 7. Determination of the plastic parameter based on displace-


ment and rotation data.

H  e 
[8] θs =  I θH + I θM 
md 3  d 

where
−3 −5
[9] IuH = 3.2K b 9 ; IuM = I θH = 5.0K b 9 ;
−7
I θM = 13.6K b 9

are flexibility factors, where


E
[10] Kb =
md [13] uc = Iucus
Substituting the appropriate values in [7], [8], and [10], we [14] θc = Iθcθs
have
where Iuc and Iθc are the yield influence factors. Under cer-
 (1) → m = 2.893 × 105 K − 39 + 1.233 × 105 K − 59
tain conditions, which are usual in these types of soils
 b b
 (Budhu and Davies 1988), these factors are given by
6 −9 6 −9
5 7
[11]  (2) → m = 1.355 × 10 K b + 1.005 × 10 K b
 h − 14k 0.32
−1 [15] I uc = 1 +
 (3) → m = 3.491 × 10 K b
7
40k 0.53

h − 14k 0.32
Figure 6 shows the solutions to the three equations. It seems [16] I θc = 1 +
54k 0.53
that the hypothesis of a linear increase in soil modulus is the
one which is more applicable to the soil tested. From the where k = Kb /1000, and h = H/cd 3.
graph we see that m ≈ 30 000 kN/m3 and Kb ≈ 1160. Parameter c is computed from the test results as follows.
Combining [13] with [15],
3.3. Initial tangent modulus and plasticity parameter
uc h − 14k 0.32
The initial tangent modulus is computed from the initial [17] = 1+
linear section of the load–displacement and load–rotation us 40k 0.53
curves.
The plastic term of the soil in the Budhu and Davies or
(1988) model for soft clays is characterized by the undrained u 
shear strength, cu, so that [18] h =  c − 1 40k 0.53 − 14k 0.32
 us 
[12] cu = cz
Considering the definition of parameter h,
where c is a constant. Once yielding begins, displacements
H
and rotations tend to display strong nonlinearity. The values [19] = ch
of these movements, uc and θ c, can be computed by d3

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:30 PM
Color profile: Disabled
Composite Default screen

956 Can. Geotech. J. Vol. 37, 2000

Table 2. Secant moduli for loadings greater than 80 kN in the Fig. 8. Relationships between m and Kb secant values and load-
single pile test. ings H. E, pile modulus.
H (kN) us (mm) θs (s) m (kN/m3) Kb
115 1.2 72 25 000 1400
220 3.4 173 13 000 2700
330 6.8 318 8 550 4100
375 8.4 393 7 590 4600

it is found that
H
d
[20] d3 = c
dh
from which it may concluded that parameter c is given by
the slope of the graph of H/d3 versus h. The term H/d3 and
the corresponding h value were computed for each applied
load H, since all values of [18] are known: uc is the test (ex-
perimental) value and us is computed from [7]. Table 1
shows a compilation of the values obtained. A similar rela- β
 H 
tion exists for rotations. [22] Kb = α  2 
Figure 7 shows that after the linear limit has been reached  Ed 
and before the failure of the pile itself (which occurred fol-
lowing the 375 kN load step), there is actually a very well where a, b, α, and β are parameters to be determined. Ap-
defined proportionality between H/d3 and h. However, the plying logarithms, we have
value of c is not the same when computed using the dis-  H 
placement figures (c = 3.4 kN/m3) as it is when the rotation [23] ln m = ln a + b ln  2 
 Ed 
figures are used (c = 4.6 kN/m3). The following reasons may
provide an explanation for this: (i) the variation in depth of  H 
the undrained shear strength may not be absolutely linear, [24] ln K b = ln α + β ln  2 
 Ed 
and (or) (ii) the equations which defined the yield influence
factors assumed a number of conditions (Budhu and Davies Therefore, pairs (a, b) and (α, β) could be established by a
1988) that may not have been completely verified in all the logarithmic graph of the load divided by Ed2, m, and Kb, as
experimental situations. However, for the purposes of further shown in Fig. 8, using values given in Table 2. It should be
analysis it is probably sufficient to take c = 4 kN/m3. noted that these pairs are not independent (see the Appen-
The elastic–plastic soil is then defined by the parameters dix). The lower limit for [21] and [22] in terms of H is deter-
Es = 30 000z kPa and cu = 4z kPa (because m = 30 000 mined by computing its interception with straight lines m =
kN/m3 and c = 4 kN/m3), where z is the depth in metres. 30 000 kN/m3 and Kb = 1160, respectively. The ratios be-
tween applied forces H and parameters m or Kb are then ob-
3.4. Nonlinear elastic approach tained from the very beginning right up to pile failure:
When the soil moduli increase linearly with depth, the
process for determining secant moduli is exactly the same as  30 000 H ≤ 95 kN
 −1.012
that described in section 3.2. Table 2 shows the four load [25] m =  −2  H 
5.276 × 10  2 H > 95 kN
steps above 80 kN (the last loading is not given inasmuch as   Ed 
the pile failed following the 375 kN loading). 
To compute secant moduli for loadings different from
those in the test, representative curves of the experimental 1160 H ≤ 95 kN
 1.011
relationships H–u and H–θ need to be plotted. From these [26] Kb =  8 H 
6.560 × 10   H > 95 kN
plots the displacement and rotation values at selected loads   Ed 2 

can be obtained. Using the pairs (H, u) and (H, θ) as a basis,
equations similar to [6] or [11] (depending on the law gov- This procedure is very simple and makes it possible to im-
erning variation with depth) can be obtained using the tech- mediately obtain either m or Kb for the given load H in ques-
nique described earlier. However, this process is time tion.
consuming and is also somewhat arbitrary. As an alternative
it is proposed that the relationship between applied loads H 3.5. Checking the model parameters
and parameters m and Kb should be established once the soil The accuracy of the parameters obtained in this way can
has yielded: easily be checked by computing the load–displacement and
b load–rotation curves based on estimated values. Figure 9
 H 
[21] m = a 2 shows the consistency of the parameters obtained in model
 Ed  calibrations. The differences between the elastic–plastic

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:31 PM
Color profile: Disabled
Composite Default screen

Sousa Coutinho 957

Fig. 9. Comparison between measured and computed displacements and rotations. c, rate at which undrained shear strength increases
with depth.

Table 3. Deviations of the values computed by the elastic–plastic model with c = 4 kN/m3 relative to the ex-
perimental values.
us (mm) θs (s)
H (kN) Test Calculated Deviation (%) Test Calculated Deviation (%)
20 0.2 0.2 –1 12 12 3
40 0.3 0.4 –15 23 24 –5
80 0.8 0.8 3 46 48 –4
115 1.2 1.2 –2 72 76 –6
220 3.4 3.2 5 173 190 –9
330 6.8 6.3 7 318 353 –10
375 8.4 7.8 7 394 434 –9

model computed using c = 4 kN/m3 and the experimental being physical quantities of precise significance, they are
values are very small, as shown in Table 3. envelop parameters involving all the phenomena that occur
Although the shear strength computed from the pile load simultaneously in the deformation process between the piles
test, cu = 4z kPa, is comparable with the data from the previ- and the soil. However, there is no denying the fact that these
ous site investigation, cu = 20 kPa, the value for the soil values can be said to be experimental when applied to a pile
modulus, E = 30 000z kPa, is clearly greater than that which design problem.
would be assumed from the site investigation results, Lastly, the influence of pile yielding in the softening of
N SPT = 0 – 2 and cu = 20 kPa. This is reflected in the index the load–displacement relationship should be noted. Al-
E/cu = 7500, which is more than 10 times greater than the though it is known that piles also yield, whereupon their ri-
range usually obtained from laboratory and in situ tests. This gidity decreases, it is difficult to include this effect in the
is why pile load tests have to be carried out to obtain reliable model because it is not known how much softening is due to
“soil” parameters which are relevant to the problem of pile yielding and how much is due to soil yielding. The pile
loaded piles. In fact, laboratory and in situ tests do not accu- modulus is thus assumed to be constant; nonetheless, the pa-
rately reproduce the mechanical and geometrical conditions rameters obtained in the model calibration definitely include
of the pile–soil interaction phenomenology. Thus, the pre- this aspect and, as discussed above, they can be said to be
ferred approach to the problem requires the use of simple envelop parameters rather than parameters with a precise
models that are possibly farther from the physical reality of physical meaning.
the problem but are more compatible with the way in which
geotechnical characteristics are determined. This character-
4. Prediction of the load–displacement
ization is carried out by combining the selected analytical
model and the test results and computing the values that the curves of pile groups
mechanical parameters of the analytical model have to as-
4.1. Elastic–plastic method
sume to minimize the deviation between the model predic-
tion and the test results. The values of the mechanical 4.1.1. Relationship between loads and displacements
parameters determined in the manner described above are The relationship between the group surface displacement
characteristic of that model, whereas the same mechanical uG and the applied load HG under conditions of elasticity is
parameters may take on other values if the model to be cali- given by
brated is different. This dependency of the values of the me-
chanical parameters on the models shows that rather than [27] KGuG = HG

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:37 PM
Color profile: Disabled
Composite Default screen

958 Can. Geotech. J. Vol. 37, 2000

Table 4. Constants for the calculation of interaction factors [33] u G = FR u H G


(Novak and El Sharnouby 1985).
where FR and u are computed according to the restriction of
Pile head β (°) A B C the pile head movements. For the case of fixed-head piles
Fixed 0 0.48 0.075 0.6 (where no head rotation is allowed), the lateral displacement
90 0.27 0.11 0.7 is given by (Budhu and Davies 1988)
Pinned 0 0.41 0.095 0.725 I uF
90 0.203 0.134 0.838 [34] uF =
md 2
where IuF = 1.4Kb–3/9 is a flexibility factor. Equation [33]
where KG is the pile group stiffness, which may be com- then takes the form
puted from interaction factors (Poulos 1975; Novak and El
−3
Sharnouby 1985). To date, the Davies and Budhu (1986) 1.4Kb 9
theory has only been applied to single piles and so it is nec- [35] uG = FR H G
md 2
essary to use solutions proposed by other authors to define
the interaction factors. The equations proposed by Novak where the subscript F specifies the factor FR for fixed-head
and El Sharnouby (1985) have been chosen instead of the piles.
Poulos (1971) solutions for the following reasons: (i) the in- In a way which is similar to the methodology used to
teraction factors are defined regardless of the length of the compute single pile movements, soil yielding may be taken
piles; (ii) the piles are assumed to be solid and elastic beams into consideration by applying a yield influence factor to the
(these two aspects mean that this approach to the problem is elastic linear group displacement. In the case of a fixed-head
closer to the theory of Davies and Budhu than to the theory single pile, that factor is given by (Budhu and Davies 1988)
of Poulos); and (iii) the interaction factors can be computed
h − 32 k 0.43
analytically. [36] F
I uc =1+
Novak and El Sharnouby (1985) proposed computing the 105 k 0.54
interaction factors using the following equation: F
Direct application of I uc in the computation of the pile
B
 E  d C
group displacement is probably inaccurate because it is
[28] αu = A     equivalent to assuming that the soil surrounding each pile
 Es   s 
yields as if the pile existed on its own or, in other words, re-
where αu is the interaction factor for displacements; s is the gardless of the influence of the yielding undergone by the
pile spacing; and A, B, and C are constants given in Table 4. soil surrounding the remaining piles in the group. In the
For angles other than 0° and 90°, Novak and El Sharnouby Poulos (1975) method the yield influence factor is a function
(1985) suggested the use of the interpolation equation of of the ratio between the current load level of the pile group
Randolph and Poulos (1982): HG and the group bearing resistance HGu, as given by

[29] αu(β°) = α u(0°)cos2 β + α u(90°)sin2 β [37] HGu = n ηFHu

where β is the departure angle as defined by Poulos (1971). where n is the number of piles in the group, ηF is a lateral
According to Novak and El Sharnouby (1985), pile group efficiency factor, and Hu is the bearing resistance of a single
stiffness can be computed by pile. In the theory of Budhu and Davies (1988), however, the
yield influence factor of a single pile is a function of the ra-
[30] KG = k p ∑ ∑ λ ij tio between the current load H and the plasticity term cd 3,
i j
where kp is the stiffness of a single pile, λ ij are the elements which is expressed in [36] by the factor h. For pile groups,
of the matrix [aij]–1, and αij are the interaction factors be- this factor should be given by
tween any two piles. Substituting the value of KG from [30] HG
in [27] gives [38] h′ =
n η F cd 3
HG
[31] uG =
k p ∑ ∑ λ ij
whereupon
i j h ′ − 32 k 0.43
[39] F
(I uc )G = 1+
By definition, the inverse of stiffness is flexibility, that is, 105 k 0.54
the inverse of kp is the displacement for a unit applied force,
Displacement of fixed-head pile group which takes soil
which may be called a unit reference displacement, u
yielding into consideration is then given by
(Poulos 1971). Further, defining the quantity
−3
1 1.4 K b 9
[32] FR = [40] uG = F
(I uc )G FRF HG
∑ ∑ λ ij md 2

called group interaction factor, and assuming a linear–elastic 4.1.2. Evaluation of the lateral efficiency factor
soil, the equation with which to compute the surface dis- The evaluation of the lateral efficiency factors is a very
placements of a pile group is difficult problem. Poulos (1975) refers to the scarcity of

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:37 PM
Color profile: Disabled
Composite Default screen

Sousa Coutinho 959

information on the subject and recalls the experimental re- Fig. 10. Lateral efficiency factors: 䊉, Znamensky and Konnov
sults obtained from reduced models by Prakash and Saran 1985 (n = 9); 1, Oteo 1972 (3 × 3 group); 2, Prakash and Saran
(1967) and Oteo (1972), which he condensed in a graph sim- 1967 (2 × 2 group); 3, Prakash and Saran 1967 (3 × 3 group). s,
ilar to that in Fig. 10. The author knows of no further contri- pile spacing; ηF, lateral efficiency factor at failure.
butions to the solution of this problem, with the exception of
a work by Znamensky and Konnov (1985) that is based on
real (prototype) test results.
According to Znamensky and Konnov (1985), the bearing
resistance of a fixed-head pile group may be computed by
the equation
[41] HGu = n ηFKrHul
where Hul is the experimentally determined bearing resis-
tance of a free-head single pile, and the factor Kr is a re-
straint coefficient given by the ratio between the force
required to impose a unit displacement at the pile head of a
fixed-head pile and that required to impose a unit displace-
ment at the pile head of a free-head pile. Given that
Znamensky and Konnov consider that their theory is applica-
ble to long and flexible piles, the Budhu and Davies (1988)
equations for computing the displacements of piles subjected
to horizontal forces may be applied to compute that coeffi-
cient:
I uH
[42] Kr = the efficiency level of a pile group progressively decreases
I uF as the overlapping of soil yielding zones increases from the
moment the yielding begins until failure occurs. This means
whence Kr ≈ 2.3. This value is comparable to those pre- that the lateral efficiency factor η should vary with the load
sented by Znamensky and Konnov. Finally, the efficiency H. Given the importance of the evaluation of lateral effi-
factor ηF (which Znamensky and Konnov called an interac- ciency factors in predicting reliable load–displacement
tion coefficient) can be computed as a function of just pile curves for pile groups, discussion on this issue, based both
spacing and pile number: on the theory of Znamensky and Konnov (1985) and on ex-
0.016n + 0.39 perimental data, is presented as follows.
 s
  First, the group bearing resistance is computed using [41]:
d taking the load failure of the single pile Hul ≈ 425 kN, the
[43] ηF =
0.14n + 2.08 ratio of the diameter of the pile and to the pile spacing s/d =
2.73, and ηF = 0.44 computed using [43], a value of HGu ≈
The proposal of Znamensky and Konnov (1985) for an effi- 7000 kN is obtained. Second, it is assumed that the load at
ciency factor may be criticized because it does not take the which the yielding of the pile group test begins, Hc, is about
pile layout into account. For example, a two by six pile 1600 kN (this value is the load at which the single pile be-
group presumably does not possess the same efficiency if gins to yield times the number of piles in the group, i.e., -16
load H is perpendicular to the two-pile row as it does if it is × 100 kN). Thus, between the 1600 kN load, at which the ef-
perpendicular to the six-pile row. However, the authors ficiency is 1, and the 7000 kN load, at which point the effi-
stressed that the experimental results seemed to show that ef- ciency becomes 0.44, there is a large set of intermediate
ficiency depended mainly on the number of piles in the cases in which it is assumed that efficiency varies with load
group and on their spacing. Indeed, Ooi and Duncan (1994) according to the logarithmic law
have published a work in which they consider that the rela-
tionship between the behaviour of a single pile and the be- b
 H 
haviour of a pile group depends first and foremost on the [44] η = a 2
number of piles and their spacing. According to Fig. 10, the  Ed 
ηF values obtained by the Znamensky and Konnov equation
are comparable to those obtained by Oteo (1972), and some- in which coefficients a and b can be computed from the limit
what less comparable to those obtained by Prakash and cases shown earlier:
Saran (1967). However, the factors computed by the
ln η F
Znamensky and Konnov equation refer to fixed-head piles, [45] b=
which is not the case for the results of the tests conducted by H 
ln  Gu 
Prakash and Saran and Oteo. In these tests, significant head  Hc 
rotations occurred, and thus the tested piles can be consid-
ered free-head piles (Poulos and Davis 1980).
There is another problem that should be considered within  H 
− b ln  c2 
 Ed 
the scope of the evaluation of the efficiency of pile groups: [46] a =e

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:38 PM
Color profile: Disabled
Composite Default screen

960 Can. Geotech. J. Vol. 37, 2000

Fig. 11. Relationship between efficiency and loadings (with data Fig. 12. Comparison between measured and predicted displace-
from performed tests). HG, horizontal load applied to a pile ments for several hypotheses of efficiency.
group; uG, surface displacement of a pile group; η, lateral effi-
ciency factor.

F RF as a function of the average load level of each pile


thus giving a = 3.292 × 10–3 and b = –0.556. Equation [44] (H = HG/n). However, contrary to the initial tangent modu-
for the given conditions is illustrated in Fig. 11. lus and the plasticity parameter in the elastic–plastic model,
Figure 12 enables us to compare the experimental results the secant moduli are not intrinsic parameters of the medium
of the pile group test with predictions of the linear–elastic surrounding the piles. They are particularly dependent on the
curve computed by [35] and elastic–plastic curves computed geometrical and mechanical boundary conditions of the sys-
by [40] while using different values of η: (i) hF = 1 (inde- tem; in other words, they depend on the manner in which the
pendent yielding of each pile), (ii) η according to [44], (iii) pile deforms. In fact, a secant modulus attempts to simulta-
η = 0.54 (computed from [44] for HG = 4800 kN, the highest neously account for the elastic and plastic soil deformation
loading in the pile group test), (iv) ηF = 0.44 (the around the pile. By comparing [15] and [36] it is seen that
Znamensky and Konnov (1985) value computed from [43]), the plastic deformations takes on very different values de-
and (v) ηF = 0.37 (value extrapolated from Fig. 10). Obser- pending on whether the pile head is fixed or free. Thus [35]
vation of Fig. 12 confirms that (i) the soil yielding around should not be applied directly, since the secant moduli were
each pile is not independent of the soil yielding around the obtained from a test on a free-head pile, whereas [35] ap-
remaining piles in the group (if it were, η F would equal 1), plies to fixed-head pile groups. Nevertheless, the problem is
and thus the yield influence factor of a single pile must be still resolvable as follows.
influenced by an efficiency factor that reflects the overlap- From [31] and [33],
ping of the yielded zones; (ii) yielding is progressive,
1
increasing with the load level, since curve 3 is a better ap- [47] uG = HG = FRu HG
proximation than curves 5 and 6, which were computed for k ∑ ∑ η ij
failure values; (iii) if the group bearing resistance (computed
using the theory of Znamensky and Konnov, for example) is The previous equation can be written in the form
not going to be reached in practise, the lateral efficiency fac-
tor can assume the value for the greater load in the problem [48] u G = u G HG
to be analyzed as computed from [44] (as curve 4 shows); where
and (iv) the efficiency factor of Znamensky and Konnov
gives better results than extrapolation from the curves in [49] u G = FRu
Fig. 10, so it may be better to use the former instead of the
latter. The physical meaning of [48] is clear: under linear and
Whichever method is employed, it is clear that the evalua- elastic conditions the displacement of a pile group is com-
tion of lateral efficiency factors is a critical element in ensur- puted assuming that the total horizontal load applied to the
ing reliable predictions of load–displacement curves, and group is applied to just one pile, with a flexibility equivalent
also that this is the most subjective aspect in the computa- to that of a single pile (u), which is subjected to the influ-
tion of the effects of yielding. ence of all the piles in the group, the group interaction factor
FR. Thus, for a certain load level HG, the following relation-
4.2. Elastic nonlinear method ship is valid:
At first sight it would seem that [35] should be directly u GT u GF
employed to determine the load–displacement curve using [50] =
secant parameters, computing the values of K b, m, and uT uF

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:42 PM
Color profile: Disabled
Composite Default screen

Sousa Coutinho 961

Fig. 13. Comparison between measured and predicted displace- stiffness (this correction is not very significant and may be
ments for both linear elastic–plastic and nonlinear elastic meth- neglected in reaching a first approximation).
ods.

5. Conclusions
This paper has presented two methods for predicting the
load–displacement curves of pile groups based on the results
of single pile tests. Both methods are derived from the same
model but are based on two different approaches: linear elas-
tic–plastic and elastic nonlinear. The model was judiciously
selected to be simple, realistic, and easily associated with
the results of pile tests.
In the linear elastic–plastic method the importance of the
estimation of the group efficiency in the prediction of pile
group behaviour is highlighted. In the elastic nonlinear
method the importance of taking the boundary conditions of
both the single tested pile and the pile group to be analyzed
is highlighted.
Finally, a comparison of the predictions and the actual re-
sults of a full-scale pile group test suggests that if the chosen
model and proposed methods are accurately calibrated and
where u T is the flexibility of the tested pile; uGT is the dis- properly used, they are capable of producing realistic predic-
placement of the pile which is equivalent to that of the tions.
group, with the same boundary conditions as those of the
tested pile; uGF is the displacement of the equivalent pile in
the group for which uGT is computed, but where the piles
Acknowledgements
have fixed heads; and u F is the flexibility of a fixed-head Permission to publish this paper was given by the Director
pile. of LNEC. Thanks are also due to the Portuguese Highway
By combining [48] and [50] we find the equation for com- Authority (JAE) for authorizing the publication of the results
puting the displacement of a fixed-head pile group using se- of the tests.
cant moduli derived from the results of a single pile test,
whichever head condition is applicable to the test pile:
uF References
[51] u GF = u GT H G
uT Budhu, M., and Davies, T.G. 1987. Nonlinear analysis of laterally
loaded piles in cohesionless soils. Canadian Geotechnical Jour-
In particular, when the tested pile is also a fixed-head pile nal, 24: 289–296.
(this is more an academic hypothesis than a real one), the Budhu, M., and Davies, T.G. 1988. Analysis of laterally loaded
previous equation assumes the same form as that of [35]. In piles in soft clays. Journal of Geotechnical Engineering, ASCE,
the case of the tests under analysis, the tested pile was a 114(1): 21–39.
Davies, T.G., and Budhu, M. 1986. Non-linear analysis of laterally
free-head pile, and thus [51] takes the following form:
loaded piles in heavily overconsolidated clays. Géotechnique,
I uF I uH 36(4): 527–538.
[52] u GF = FRH HG Guedes de Melo, F. 1987. Comportamento de estacas e de grupos
e md 2
I uH + I uM de estacas sob a acção de solicitações horizontais estáticas (Be-
d
haviour of single piles and pile groups under lateral static
where the subscript H specifies the factor FR for a free-head loads). Research program, Laboratório Nacional de Engenharia
pile. It should be noted that the loads applied to the tested Civil, Lisbon.
pile possessed an eccentricity, and the secant moduli were Novak, M., and El Sharnouby, B. 1985. Pile groups under static
obtained subject to that condition. Hence, the flexibility of and dynamic loading. In Proceedings of the 11th International
the tested pile which must be considered in [52] must also Conference on Soil Mechanics and Foundation Engineering, San
take this feature into account. If the eccentricity is zero, the Francisco, Vol. 3, pp. 1449–1454.
equation can be written as follows: Ooi, P.S.K., and Duncan, J.M. 1994. Lateral load analysis of
groups of piles and drilled shafts. Journal of Geotechnical Engi-
I uF neering, ASCE, 120(6): 1034–1050.
[53] u GH = FRH HG
md 2 Oteo, C.S. 1972. Deformacion de un grupo de pilotes verticales,
solicitado por esfuerzos laterales. In Proceedings of the 5th Eu-
Figure 13 shows the results of [52], bearing in mind the ropean Conference on Soil Mechanics and Foundation Engi-
appropriate values for Kb, m, and FRH. Equations [25] and neering, Madrid, Vol. 1, pp. 397–405.
[26] were used for m and Kb as functions of the average load Poulos, H.G. 1971. Behaviour of laterally loaded piles: II — Pile
level (H = HG/n), and the value of FRH was recomputed at groups. Journal of the Soil Mechanics and Foundations Divi-
each load level, given that it varies with the relative pile–soil sion, ASCE, 97(SM5): 733–751.

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Tuesday, September 26, 2000 3:06:45 PM
Color profile: Disabled
Composite Default screen

962 Can. Geotech. J. Vol. 37, 2000

Poulos, H.G. 1975. Lateral load-deflection prediction for pile IuF, IuH, flexibility factors
groups. Journal of the Geotechnical Engineering Division, IuM, IθH,
ASCE, 101(GT1): 19–34. IθM
Poulos, H.G., and Davis, E.H. 1980. Pile foundation analysis and F
Iuc, I uc , yield influence factors for single piles
design. John Wiley and Sons, New York. Iθc
Prakash, S., and Saran, D. 1967. Behaviour of laterally loaded F
(I uc )G yield influence factor for fixed-head pile groups
piles in cohesive soil. In Proceedings of the 3rd Asian Regional k parameter defined by Kb/1000
Conference on Soil Mechanics and Foundation Engineering, kp single pile stiffness
Haifa, Vol. 1, pp. 235–239. Kb pile–soil relative stiffness
Randolph, M.F., and Poulos, H.G. 1982. Estimating the flexibility KG pile group stiffness
of offshore pile groups. In Proceedings of the 2nd International Kr restraint coefficient
Conference on Numerical Methods in Offshore Piling, Univer- m rate at which soil modulus increases with depth
sity of Texas, Austin, pp. 313–328. n number of piles in a group
Sousa Coutinho, A.G. 1995. Comportamento de estacas verticais NSPT SPT blows
carregadas lateralmente à superfície do terreno (The behaviour s pile spacing
of vertical piles loaded laterally at ground surface). Ph.D. thesis, u unit reference displacement (displacement for a unit
Civil Engineering, Universidade Técnica de Lisboa, Lisbon. force)
Sousa Coutinho, A.G., Toco Emílio, F., Almeida Garrett, J., and uF unit reference displacement for a fixed-head pile
Tavares Cardoso, E. 2000. Instrumentation of bored concrete uG surface displacement of a pile group
piles for horizontal load tests. Geotechnical Testing Journal. uGF surface displacement of a fixed-head pile group
23(3): 327–337.
uGT displacement of the pile which is equivalent to that of
Toco Emílio, F., Tavares Cardoso, E., and Almeida Garrett, J. the group
1990. Concrete pile instrumentation for lateral loading testing.
uc pile head displacement in yielding soil
In Proceedings of the 9th International Conference on Experi-
us pile head displacement in linear elastic soil
mental Mechanics, Technical University of Denmark, Lingby,
uT flexibility of tested pile
pp. 1385–1392.
z depth
Znamensky, V.V., and Konnov, A.V. 1985. Calculation of bearing
α ij displacement interaction factor between piles i and j
capacity of laterally loaded pile groups. In Proceedings of the
11th International Conference on Soil Mechanics and Founda- αu displacement interaction factor
tion Engineering, San Francisco, Vol. 3, pp. 1511–1514. β departure angle (as defined by Poulos 1971)
γ unit weight
η lateral efficiency factor
ηF lateral efficiency factor at failure
λ ij elements of the matrix [aij]–1
List of symbols θc pile head rotation in yielding soil
θs pile head rotation in linear elastic soil
a, b constants in logarithmic laws
A, B, C constants
c rate at which undrained shear strength increases with Appendix
depth If we look at the relationship between m and K b , as
cu undrained shear strength defined by [10], it is possible to show that pairs (a, b) and
d pile diameter (α, β) are not independent. By combining [22] and [10], we
e load eccentricity have
E pile modulus
β
Es soil modulus E  H 
[A1] = α  2
(EI )p pile flexural stiffness md  Ed 
FR group interaction factor
FRF FR for a fixed-head pile group or
FRH FR for a free-head pile group −β b
h parameter defined by H/cd3 E  H   H 
[A2] m =   = a  2
h ′ parameter h for pile groups α d  Ed 2   Ed 
H horizontal load
HG horizontal load applied to a pile group Formally speaking, it is more correct, though not particu-
HGu bearing resistance of a pile group larly significant, to establish the relationship between Kb and
Hc yielding load of a pile group H because all the intervening factors are dimensionless,
Hu bearing resistance of a single pile which makes the constants independent of the unit system
Hul bearing resistance of a free-head single pile determined (in fact, Kb can be considered the dimensionless parameter
experimentally of m).

© 2000 NRC Canada

I:\cgj\Cgj37\Cgj05\T00-022.vp
Wednesday, September 27, 2000 1:00:52 PM
View publication stats

Vous aimerez peut-être aussi