Vous êtes sur la page 1sur 6

1293

Load tests on full-scale bored pile groups


Guoliang Dai, Rodrigo Salgado, Weiming Gong, and Yanbei Zhang

Abstract: The interactions between closely spaced piles in a pile group are complex. Very limited experimental data are
available on the loading of full-scale bored pile groups. This paper reports the results of axial static load tests of both full-
scale instrumented pile groups and single piles. The load tests aimed to ascertain the influence of number, length, and
spacing of the piles on pile group load response. Experiments varied in the number of piles in the group, pile spacing, type
of pile groups, and pile length. All piles had a diameter of 400 mm. Two-, four-, and nine-pile groups with pile lengths of
20 and 24 m were tested. As the isolated piles and some piles in the pile groups were instrumented, the load transfer and
load–settlement curves of both piles in isolation and individual instrumented piles in the groups were obtained. The
interaction coefficient for each pile in the group was back-calculated from the measured data. The interaction coefficients
are shown to be dependent on pile proximity, as usually assumed in elastic analyses, but also on settlement and on the size
of the group.
Key words: pile groups, load transfer, settlement ratio, interaction coefficient.
Résumé : Les interactions entre des pieux placés à proximité les uns des autres dans un groupe de pieux sont complexes.
Très peu de données expérimentales sont disponibles sur les chargements de groupes de pieux foncés, à l’échelle réelle.
Cet article présente les résultats d’essais de chargement statique axial sur des groupes de pieux et des pieux individuels
instrumentés à l’échelle réelle. Les essais de chargement visaient à confirmer l’influence du nombre, de la longueur et de
l’espacement des pieux sur le comportement de groupes de pieux. Les essais ont permis de varier le nombre de pieux dans
un groupe, l’espacement entre les pieux, le type de groupe de pieux et la longueur des pieux. Tous les pieux avaient un
diamètre de 400 mm. Des groupes de deux pieux, de quatre pieux et de neuf pieux, avec une longueur de 20 m et de 24 m
respectivement, ont été testés. Puisque les pieux individuels et quelques pieux à l’intérieur des groupes ont été
instrumentés, les courbes de transfert de charge et de tassement total ont été obtenues autant pour les pieux isolés que pour
les pieux individuels faisant partie d’un groupe. Le coefficient d’interaction pour chaque pieu dans le groupe a été rétro-
calculé à partir des données mesurées. On a démontré que les coefficients d’interaction dépendent de la proximité des
pieux, tel que normalement supposé dans les analyses élastiques, mais aussi du tassement et de la taille du groupe.
Mots-clés : groupe de pieux, transfert de charge, ratio de tassement, coefficient d’interaction.
[Traduit par la Rédaction]

Introduction (length to pile diameter) greater than 100) and an associated


need for more refined designs. In pursuing more refined de-
Infrastructure construction in China has proceeded at a fast signs, one of the areas where real data from instrumented
pace. Many large-span bridges have been built across the structures is most lacking is pile groups.
Yangtze River, across the Yellow River, and even across the Settlement analyses of pile groups (e.g., Poulos 1968;
ocean. Examples include the Jiangyin Yangtze River Bridge, Butterfield and Banerjee 1971; Randolph and Wroth 1979;
Runyang Yangtze River Bridge, Sutong Yangtze River Poulos and Randolph 1983; Poulos 1989; Chow and Teh 1991;
Bridge, Hangzhou Bay Sea Bridge, and Donghai Sea Bridge. Lee 1993a, 1993b; Mandolini and Viggiani 1997; Mylonakis
As an illustration of this type of foundation in engineering and Gazetas 1998; Randolph 2003; Leung et al. 2010) are
practice, the two main-span foundations of the Sutong Yangtze based on a variety of approaches, which include boundary-
River Bridge consist of 131 117 m long piles with diameters in element methods, the hybrid load transfer approach, and the
the 2.8 –2.5 m range with pile caps having plan dimensions of finite element method. Despite some theoretical advances in
50 m by 48 m and thickness varying between 5 and 13 m. the analyses and prediction of pile group behavior in the last
These bridges all have large spans and consequently require few decades, analyses are still based largely on simplifications
piles with large load capacity in pile-based foundation solu- of the problem and of the constitutive behavior of the soil.
tions. This in turn leads to the use of a large number of piles Consequently, static load tests on groups remain the most
with large diameters and long lengths (sometimes referred to reliable means of assessing pile group response under design
as “super long” piles, which might be generally thought of as loads. Some model and field pile group vertical load tests have
piles with lengths greater than 100 m or slenderness ratio been performed (Whitaker 1957; Hanna 1963; Barden and
Received 8 November 2011. Accepted 31 July 2012. Published at www.nrcresearchpress.com/cgj on 8 November 2012.
G. Dai and W. Gong. School of Civil Engineering, Southeast University, No. 2 Sipailou, Nanjing, Jiangsu, 210096, China.
R. Salgado and Y. Zhang. School of Civil Engineering, Purdue University, 550 Stadium Mall Drive, West Lafayette, IN 47907, USA.
Corresponding author: Guoliang Dai (e-mail: daigl@seu.edu.cn).

Can. Geotech. J. 49: 1293–1308 (2012) doi:10.1139/t2012-087 Published by NRC Research Press
1294 Can. Geotech. J. Vol. 49, 2012

Monckton 1970; Cooke et al. 1980; Briaud et al. 1989; Ismael followed by a number indicates a specific pile within that group. For
2001; Bai et al. 2006; Yetginer et al. 2006); however, due to instance, QZ2–1 is a reference to pile No. 1 in the QZ2 group. The
the difficulties and cost of full-scale load tests, most pile group layout of the testing area is shown in Fig. 1.
tests were scaled down regardless of whether they were per-
formed in the field or laboratory. There are few full-scale, in Site characterization
situ bored pile group load tests reported in the literature (Bai The test site was located in Jiangning, Nanjing, China, and
et al. 2006). the ground at the site was level. The subsoil of the test site is
The absence of rigorous methods of analysis and the scar- the QinHuai River floodplain. One auger boring (BH) was
city of full-scale high-quality data from instrumented load drilled at the test site to a depth of 29.50 m. This boring
tests on single piles and pile groups means that the use of showed a uniform, thick soft clay layer starting at 17 m and
conventional methods to design large-scale, heavily loaded extending all the way to the bottom of that boring (and, from
pile foundations with very long pile lengths requires extra knowledge of the site, beyond). This auger boring depth was
caution. Some research has been done on super-long single 11B (B ⫽ 400 mm, the pile diameter) deeper than the test pile
piles (Wei 1996; Fei 2000; Yu 2002; Fang 2003; Xie et al. base for piles with 24 m length. Static cone penetration tests
2003). Results from this research show three key characteris- (CPTs) were performed in the vicinity of BH to give a con-
tics of super-long, large-diameter piles that differ from con- tinuous record of the soil resistance with depth. Figure 2 shows
ventional piles: (i) pile weight is a larger percentage of pile all four CPT logs available for the site, the locations of which
bearing capacity; (ii) accounting for pile compression becomes are identified in Fig. 1. As CPTu is not commonly used in
more important in settlement estimation; (iii) mobilization of China, pore-water pressure measurement was not possible
base resistance requires excessive pile head settlements due to with the cone used. Table 2 summarizes the soil properties of
greater axial compressibility of the pile, which implies that each soil layer crossed by the test piles. The subsoil profile
ultimate loads will be associated with much greater ratios of includes multiple layers of silt and clay. The thickness of each
shaft to base resistance. Although most published work on layer was identified based on CPT data. Soil samples were
long piles has focused on single piles, long piles are most often obtained by using split-spoon samplers. The ground water
used in groups. The typical approach in a project involving level was found at a depth of 2.60 m.
long piles is to load-test single piles for verification and to Pile installation and instrumentation
estimate pile group settlement by approximate methods. The piles were installed using the slurry method. Pile in-
The present paper aims to start filling this knowledge gap by stallation started on 28 June 2008 and ended on 3 August
analyzing full-scale, in situ bored pile group tests. The aim of 2008. The piles were drilled to depths of 20 or 24 m and
the tests and analysis is to investigate the following crucial protruded 0.1 m above ground level. A 0.4, 0.8 or 1.2 m thick
issues in particular: (i) rates of the pile head and pile base load reinforced concrete cap was subsequently poured on the pile
mobilization with settlement; (ii) variation of the shaft resis- groups and single piles. The eight caps were completed all at
tance, which is responsible for the difference between these once between 5–13 August 2008. Figure 1 shows a layout plan
two rates, between single piles and piles in a group in various of the test piles and pile groups. The pile caps rested on the
arrangements; (iii) proportion of load applied on a pile cap ground and may be considered rigid for practical purposes.
shared between the piles in the group; (iv) pile group effi- The pile spacing was 2.5B in groups QZ2, QZ4, and QZ9 and
ciency variation with settlement. 3.0B in groups QZ2L, QZ4L, and QZ9L. Details of the pile
spacing in the groups and dimensions of the pile caps is
Experimental program summarized in Table 1.
Axial forces along the depth were monitored by two strain
Pile configurations
gauges installed evenly at each cross section for all test piles.
As summarized in Table 1, the field load tests were per-
There were six instrumented sections in each instrumented pile
formed on: (i) an isolated single pile with length L ⫽ 20 m;
(as shown in Fig. 2). A vibrating-wire load cell measured the pile
(ii) an isolated single pile with L ⫽ 24 m; (iii) a two-pile group
top load of each pile in the pile groups (for four- and nine-pile
with spacing sp ⫽ 2.5B (where B is the pile diameter of
groups) during the loading process. Fourteen vibrating wire cells
400 mm) and L ⫽ 20 m; (iv) a two-pile group with sp ⫽ 3.0B
were used in total, one each on the top of the following piles: QZ4-1,
and L ⫽ 24 m; (v) a four-pile group with sp ⫽ 2.5B and L ⫽
QZ4-4, QZ4L-1, QZ4L-4, QZ9-1, QZ9-4, QZ9-5, QZ9-6, QZ9-9,
20 m; (vi) a four-pile group with sp ⫽ 3.0B and L ⫽ 24 m;
QZ9L-1, QZ9L-4, QZ9L-5, QZ9L-6, and QZ9L-9.
(vii) a nine-pile group with sp ⫽ 2.5B and L ⫽ 20 m; (viii) a
nine-pile group with sp ⫽ 3.0B and L ⫽ 24 m. All piles in the Test procedure
experiments had a diameter, B, of 400 mm. The concrete The load tests were performed using the kentledge load

strength (f cd ) was 25 MPa for both the piles and the caps. All method. The dead load applied by square precast concrete
caps were reinforced with 12 mm rebars at a two-way spacing piles was placed evenly on the reaction platform, before the
of 150 mm. The concrete reinforcement cover was 70 mm in test. The length of the reaction beams was 12 m and the
the caps and 35 mm in the piles. distances between the supporting points and the center of
In this paper, tests on isolated single piles are denoted DZ. Pile the pile caps were all larger than 5 m. It is possible that, for the
group tests are denoted QZ. The suffix L is used to indicate that nine-pile group, this distance might not be sufficient to avoid
the pile length, L, is 24 m (and so is longer than 20 m, the length creating an impact on values of measured settlement. How-
for the other set of piles). For example, QZ2 represents the two-pile ever, this impact would be negligible except for small loads
group with shorter pile length and QZ2L indicates the two-pile group and settlements. The dead load was 1.2 times the estimated
with longer piles. Finally, a dash after the pile group reference ultimate load of the test pile group. Load was applied by

Published by NRC Research Press


Dai et al. 1305

Now we write eq. [10] three times, once for each of the three pile than comparable coefficients (at the same spacing) for the
types (a, b, and c), keeping in mind that the settlement of each is nine-pile group. The presence of additional piles around in-
equal to the settlement wg of the pile group. Any three piles, one of teracting piles likely interferes with load or settlement trans-
each type, will suffice. We take piles 1, 4, and 5. For pile 1 mission between the interacting piles. Although the results
cannot be used at this time to propose any relationship for
n

兺α K
Qj interaction coefficients, they do show clearly (i) the depen-
[22] w1 ⫽ wg ⫽ ij dence of these factors on settlement (which means that factors
j⫽1 tj
derived from elasticity theory must be used with proper judg-
⫽ Qa /Kta ⫹ 2α1Qb /Ktb ⫹ α2Qc /Ktc ment) and (ii) the dependence of these factors on the size of
⫹ 2α3Qa /Kta ⫹ 2α4Qb /Ktb ⫹ α5Qa /Kta the group (because of interference of additional piles on how
piles interact). Research on clarifying and quantifying these
where Kta, Ktb, and Ktc are the pile head stiffness of the single effects is needed.
pile of type a, b, and c with the same length as the piles in the
group. Likewise, for piles 4 and 5 Summary and conclusions
n A field pile load testing program was carried out on isolated

Qj
[23] w4 ⫽ wg ⫽ αij bored piles and bored pile groups installed in a soil profile with
j⫽1
Ktj mixed layers of clay and silt in Nanjing, China. The program
⫽ Qb/Ktb ⫹ 2α1Qa/Kta ⫹ α1Qc/Ktc included two single instrumented piles and six types of pile
⫹ 2α2Qb/Ktb ⫹ α3Qb/Ktb ⫹ 2α4Qa/Kta groups with two, four, and nine piles with different pile
lengths and pile spacing. Based on the analysis of the field test
n results, the following conclusions can be reached:
兺α K
Qj
[24] w5 ⫽ wg ⫽ ij ⫽ Qc /Ktc ⫹ 4α1Qb /Ktb
j⫽1 tj 1. Using the conventional definition of ultimate load as the
⫹ 4α2Qa /Kta load causing a settlement of 10% of the pile diameter, the
two single piles DZ1 (L/B ⫽ 50) and DZ1L (L/B ⫽ 60)
mobilized essentially only shaft resistance, with loads
In this case, we have five unknowns and three equations. measured at the strain gauge level closest to the pile base
Using eq. [11], we can write four additional equations that can accounting for only 2.2% of the total load for the 20 m long
be used for the ratios of interaction coefficients. pile and 4% of the total load for the 24 m long pile.
2. The general response of an individual pile in the two-pile
α1 α2 α3 α4 groups was observed to be very close to that of the corre-
[25] ⫽ 1.10, ⫽ 1.12, ⫽ 1.04, ⫽ 1.09 sponding single pile. This is evidenced, for example, by
α2 α3 α4 α5 the values of limit unit shaft resistance of piles in the
two-pile group being approximately the same as those for
In Appendix A, we discuss how to obtain the interaction the corresponding single pile. This means that interaction
coefficients using these equations by linear optimization. between piles in the two-pile groups was relatively small.
This was not observed in the 4-and 9-pile groups, in which
Resulting influence coefficients substantial interaction and group effect was observed.
Each load–settlement data point from the load test can be
3. In general, values of Rs of both the four- and nine-pile
used to calculate one influence coefficient. The influence co-
efficient can then be expressed in terms of settlement for the groups tended to increase with settlement. The single pile
different pile groups, as shown in Fig. 13. In general, with the settlement was observed to be generally smaller than
increase of group settlement, the interaction coefficient in- the corresponding pile group settlement at the same aver-
creases, with an inflection point for small settlements (marking age load per pile when the load was relatively large.
the transition from minimal interaction for small settlements to 4. For the four- and nine-pile groups, group effect was more
a higher level of interaction) and later a tendency of stabili- pronounced for QZ4 than for QZ4L and about the same for
zation at large settlements, which is consistent with more QZ9L and QZ9, showing that the impact of the pile spac-
intense localization of shear strain around the piles at large ing is greater than that of the pile length on group load.
settlements, leading to a reduction in the interaction for incre- 5. Based on the analysis of the load distribution between the
mental settlement. The results for the two-pile groups QZ2 are group piles in the nine-pile groups, the load at the top of
inconsistent with the other results, with the interaction coef- the corner piles was observed to be the largest, followed by
ficient being practically zero. This may be because of spatial side piles and then center piles. However, the load differences
variability of the soil or other variability in the pile installation were not large, particularly for side versus corner piles.
or pile cap. For the four-pile groups, the pile spacing has a 6. Based on the results of the load tests, the individual values
larger effect on interaction than pile length, which is to be of the interaction coefficients of each pile in the group were
expected. The interaction coefficient in group QZ4 with sp ⫽ back-calculated. The interaction coefficient was seen to be
2.5B and L ⫽ 20 m is on average larger than that of group a function of settlement and the size of the group. With the
QZ4L with sp ⫽ 3.0B and L ⫽ 24 m. For the nine-pile group, increase of group settlement, the interaction coefficient
the interaction coefficients are distributed proportionally to was observed to increase.
pile center-to-center spacing. As shown in Fig. 13, the inter- 7. A method of determining interaction coefficients from pile
action coefficients for the piles in the four-pile group are larger group load test results was proposed.

Published by NRC Research Press


1306 Can. Geotech. J. Vol. 49, 2012

Fig. 13. Interaction coefficients versus settlement for each pile group: (a) two-pile group with L ⫽ 20 m; (b) two-pile group with L ⫽ 24 m;
(c) four-pile group with L ⫽ 20 m; (d) four-pile group with L ⫽ 24 m; (e) nine-pile group with L ⫽ 20 m; (f) nine-pile group with L ⫽ 24 m.

(a) (b)
1.0 1.0
0.9 0.9
0.8

Interaction coefficient
Interaction coefficient

0.8
0.7 0.7 α12
0.6 0.6
α12
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
                         
Pile group settlement (mm) Pile group settlement (mm)

(c) (d)
1.2 1.2
1.1 1.1
1.0 1.0 α12
Interaction coefficient
Interaction coefficient

0.9 0.9 α14


0.8 0.8
0.7 α12 0.7
0.6 α14 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
                         
Pile group settlement (mm) Pile group settlement (mm)

(e) (f)
1.0
1.0
0.9
0.9 α1
α1 0.8
0.8
Interaction coefficient
Interaction coefficient

α2
0.7 α2 0.7
α3
0.6 α3 0.6
α4
0.5 α4 0.5
α5
0.4 α5 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
                         
Pile group settlement (mm) Pile group settlement (mm)

Acknowledgements References
This research was supported by the National Natural Sci- Bai, X., He, W., Jia, J., and Han, Y. 2006. Experimental study on the
ence Foundation of China (Grant No. 50908048) and a project interaction mechanism of cap – pile group – soil. Marine Geore-
funded by the Priority Academic Program Development of sources and Geotechnology, 24(3): 173–182. doi:10.1080/
Jiangsu Higher Education Institutions (PAPD). 10641190600788254.

Published by NRC Research Press


Dai et al. 1307

Barden, L., and Monckton, M.F. 1970. Tests on model pile groups in Géotechnique, 53(10): 847– 875. doi:10.1680/geot.2003.53.
soft and stiff clay. Géotechnique, 20(1): 94 –96. doi:10.1680/ 10.847.
geot.1970.20.1.94. Randolph, M.F., and Wroth, C.P. 1978. Analysis of vertical defor-
Briaud, J.L., Tucker, L.M., and Ng, E. 1989. Axial loaded 5 pile mation of vertically loaded piles. Journal of Geotechnical Engi-
group and single pile in sand. In Proceedings of the 12th Interna- neering, 104(12): 1465–1488.
tional Conference on SMFE, Tokyo. Vol. 1, pp. 381–386. Randolph, M.F., and Wroth, C.P. 1979. An analysis of the vertical
Butterfield, R., and Banerjee, P.K. 1971. The elastic analysis of deformation of pile groups. Géotechnique, 29(4): 423– 439. doi:
compressible piles and pile groups. Géotechnique, 21(1): 43– 60. 10.1680/geot.1979.29.4.423.
doi:10.1680/geot.1971.21.1.43. Salgado, R. 2008. The engineering of foundations. McGraw-Hill.
Chow, Y.K., and Teh, C.I. 1991. Pile – cap – pile – group interaction in Salgado, R., and Prezzi, M. 2007. Computation of cavity expansion
nonhomogeneous soil. Journal of Geotechnical Engineering, 117(11): pressure and penetration resistance in sands. International Journal
1655–1668. doi:10.1061/(ASCE)0733-9410(1991)117:11(1655). of Geomechanics, 7(4): 251–265. doi:10.1061/(ASCE)1532-
Cooke, R.W., Price, G., and Tarr, K. 1980. Jacked piles in London 3641(2007)7:4(251).
clay: interaction and group behavior under working conditions. Salgado, R., Prezzi, M., and Tehrani, F.S. 2011. Soil property-based
Géotechnique, 30(2): 97–136. doi:10.1680/geot.1980.30.2.97. methods for design of nondisplacement piles. 2011 Pan-Am CGS
Fang, P. 2003. Study on the engineering properties of super-long Geotechnical Conference. [Accepted.]
piles. Ph.D. thesis, Zhe Jiang University. [In Chinese.] Wei, H. 1996. Study on the mechanical properties of super-long pile.
Fei, H. 2000. Research of super-long hole bored pile in loess Journal of Electric Power Survey, 2: 1– 6. [In Chinese.]
subsoil. Journal of Geotechnical Engineering, 22(5): 576 –580. Whitaker, T. 1957. Experiments with model piles in groups. Géo-
[In Chinese.] technique, 7(4): 147–167. doi:10.1680/geot.1957.7.4.147.
Fleming, W.G.K., Weltman, A.J., Randolph, M.F., et al. 1992. Piling Xie, T., Yuan, W., and Yao, Y. 2003. Model test study on effect of
Engineering. Surrey University Press, Surrey. vertical bearing capacity for large-scale pile group. Journal of
Hanna, T.H. 1963. Model studies of foundation groups in sand. Highway and Transportation Research and Development, 20(5):
Géotechnique, 13(4): 334 –351. doi:10.1680/geot.1963.13.4.334. 61– 64. [In Chinese.]
Ismael, N.F. 2001. Axial load tests on bored piles and pile groups Yetginer, A.G., White, D.J., and Bolton, M.D. 2006. Field measure-
in cemented sands. Journal of Geotechnical and Geoenviron- ments of the stiffness of jacked piles and pile groups. Géotech-
mental Engineering, 127(9): 766 –773. doi:10.1061/(ASCE)1090- nique, 56(5): 349 –354. doi:10.1680/geot.2006.56.5.349.
0241(2001)127:9(766). Yu, Y. 2002. test study on the super-long bored pile in soft ground.
Lee, C.Y. 1993a. Settlement of pile groups—practical approach. Industrial Construction, 32(11): 33–35. [In Chinese.]
Journal of Geotechnical Engineering, 119(9): 1449 –1461. doi:
10.1061/(ASCE)0733-9410(1993)119:9(1449).
Appendix A. Determination of interaction
Lee, C.Y. 1993b. Pile group settlement analysis by hybrid layer
approach. Journal of Geotechnical Engineering, 119(6): 984 –997. coefficients for a nine-pile group
doi:10.1061/(ASCE)0733-9410(1993)119:6(984). As seen earlier from eqs. [22] to [25], repeated here as eqs.
Lee, J., Kyung, D., Kim, B., and Prezzi, M. 2009. Estimation of the [A1] to [A4], respectively, the available equations for a nine-pile
small-strain stiffness of clean and silty sands using stress-strain curves and group are
CPT resistances. Soils and Foundations, 49(4), 545–556.
Leung, Y.F., Soga, K., Lehane, B.M., and Klar, A. 2010. Role of n

兺α K
Qj
linear elasticity in pile group analysis and load test interpretation. [A1] w1 ⫽ wg ⫽ ij
Journal of Geotechnical and Geoenvironmental Engineering, 136(12): j⫽1 tj
1686 –1694. doi:10.1061/(ASCE)GT.1943-5606.0000392. ⫽ Qa /Kta ⫹ 2α1Qb /Ktb ⫹ α2Qc /Ktc
Loukidis, D., and Salgado, R. 2008. Analysis of shaft resistance of ⫹ 2α3Qa /Kta ⫹ 2α4Qb /Ktb ⫹ α5Qa /Kta
non-displacement piles in sand. Géotechnique, 58(4): 283–296.
doi:10.1680/geot.2008.58.4.283. n

兺α K
Qj
Mandolini, A., and Viggiani, C. 1997. Settlement of piled foundations. [A2] w4 ⫽ wg ⫽ ij
Géotechnique, 47(4): 791–816. doi:10.1680/geot.1997.47.4.791. j⫽1 tj
Mayne, P.W., and Rix, G.J. 1993. Gmax-qc relation for clays. ⫽ Qb /Ktb ⫹ 2α1Qa /Kta ⫹ α1Qc /Ktc
Geotechnical Testing Journal, 16(1) 54 – 60. ⫹ 2α2Qb /Ktb ⫹ 2α3Qb /Ktb ⫹ α4Qa /Kta
Mylonakis, G., and Gazetas, G. 1998. Settlement and additional
internal forces of grouped piles in layered soil. Géotechnique, n

兺α K
Qj
48(1): 55–72. doi:10.1680/geot.1998.48.1.55. [A3] w5 ⫽ wg ⫽ ij
Poulos, H.G. 1968. Analysis of the settlement of pile groups. Géo- j⫽1 tj
technique, 18(4): 449 – 471. ⫽ Qc /Ktc ⫹ 4α1Qb /Ktb ⫹ 4α2Qa /Kta
Poulos, H.G. 1989. Pile behave—theory and application. Géotech-
nique, 39(3): 365– 415. doi:10.1680/geot.1989.39.3.365. α1 α2 α3 α4
Poulos, H.G., and Davis, E.H. 1980. Pile foundation analysis and [A4] ⫽ 1.10, ⫽ 1.12, ⫽ 1.04, ⫽ 1.09
α2 α3 α4 α5
design. Wiley, New York.
Poulos, H.G., and Randolph, M.F. 1983. Pile group analysis: a study
of two methods. Journal of Geotechnical Engineering, 109(3): where these ratios of interaction coefficients are specific to
355–372. doi:10.1061/(ASCE)0733-9410(1983)109:3(355). the pile group geometry and conditions considered in this
Randolph, M.F. 2003. Science and empiricism in pile foundation design. paper.

Published by NRC Research Press


1308 Can. Geotech. J. Vol. 49, 2012

The residual form of eqs. [A1] to [A3] can be expressed as optimal values of the interaction coefficients using the Sim-
plex method. The algorithm for this is summarized as follows;
n 1. START with an assumed very small initial value of ␣5, and then
兺α K
Qj
[A5] r(αij)i ⫽ wg ⫺ ij calculate ␣1 to ␣4 through the constraint eq. [A4].
j⫽1 tj
2. LOOP with the constraint that ⱍαiⱍ ⬍ 1 for i ⫽ 1, 2,. . .,5.
3. INCREMENT ␣5 by a small amount to obtain its current
This residual r(␣ij)i, which is a function of the interaction value ␣5,current and calculate new values ␣1,current to
coefficients, is the difference between the measured and cal- ␣4,current of the other interaction coefficients through the
culated pile top settlement for pile type i. The coefficient of constraint eq. [A4].
interaction vector is defined as ␣ ⫽ {␣1, ␣2, ␣3, ␣4, ␣5}. A 4. COMPUTE the Frobenius norm of residual 㛳r㛳2,current for
possible objective function is the sum of the settlement differ- current values ␣current of the interaction coefficients.
ences, represented by the Frobenius norm 㛳r㛳2 of the residual 5. IF 㛳r㛳2,current ⬍ 㛳r㛳2,min. this means the current interaction
vector r ⫽ {r1, r2, r3, r4, r5}. The optimization problem can coefficient vector is a better solution, so update the solu-
now be expressed as finding the minimum of 㛳r㛳2. tion by setting 㛳r㛳2,min ⫽ 㛳r㛳2,current , and ␣optimum ⫽ ␣current.
The solution to this optimization problem is obtained by 6. ENDLOOP.
finding the optimal interaction coefficients ␣optimum at which 7. END with optimal interaction coefficient vector ␣optimum
the minimum residual 㛳r㛳2,min is obtained. We can obtain the and 㛳r㛳2,min.

Published by NRC Research Press

Vous aimerez peut-être aussi