Piles

Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)

© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

A back analysis of vertical load tests on bored piles in granular soil

L. Tosini, A. Cividini & G. Gioda

Politecnico di Milano, Department of Structural Engineering, Milan, Italy

ABSTRACT: Two back analyses are discussed of load tests carried out on bentonite slurry piles bored in
granular soils. The first case concerns a test on a pile belonging to a 12 pile group. The back analysis permits
calibrating an axisymmetric finite element model that reproduces, with reasonable accuracy, the experimental
results. The calibrated model is then extended to three-dimensional conditions and applied to the analysis of
the entire group. The results suggest some comments on the different assumptions that can be adopted in the
calculations and on their effects on the global load-settlement curve of the pile group. The second case concerns
a load test in which, in addition to the load-settlement data, also the axial strains along the pile were measured
through electrical extensometers. The numerical back analyses highlight an apparent contradiction between the
two sets of experimental data. On their bases some conclusions are drawn on the possible causes of the observed
inconsistency and on the influence of the construction technology on the soil-pile interaction.

1 INTRODUCTION of a load test on a single pile. The back analysis

of the test permits an acceptable calibration of the
When dealing with the design of deep foundations in numerical model in 2D axisymmetric regime. How-
granular deposits the prediction of their settlements is ever, when extended to 3D conditions for the entire
not straightforward due to the difficulties met in defin- group, the calculations show the appreciable influence
ing the values of the mechanical parameters influenc- of the different assumptions that can be introduced in
ing them. In most cases, in fact, the design is based on the analysis of the deep foundation.
the results of penetrometer tests, which provide only The second case concerns a load test in which, in
approximated mechanical parameters of the granular addition to the load-settlement curve, also the axial
soil. In addition, the pile settlements depend on the strain within the pile was measured through exten-
mechanical characteristics of the pile/soil interface someters applied to its steel reinforcement. In this case
that, in turn, are influenced by the adopted construc- the experimental data present an apparent contradic-
tion technology. Being aware of the possible limited tion that leads to comments on the possible influence
accuracy of the computed settlements, load tests are of details of the construction technique.
customarily carried out to quantitatively assess the
behaviour of the deep foundation under loading.
In some instances, however, the load test does not 2 FIRST LOAD TEST
directly provide the sought results, e.g. when the settle-
ment of a pile group has to be evaluated on the basis of The test was carried out on a 17 m long bentonite
a load test on a single pile. In other instances the results slurry bored pile with a diameter of 120 cm. Due to
present some apparent inconsistencies that make their the geometrical characteristics of the foundation mat,
interpretation somewhat controversial. the pile head is located about 5 m below the ground sur-
Here the back analyses of two load tests face. The soil deposit consists of sand and gravel with
on bored piles are presented based on two- a marginal percentage of silt. The design was based
and three-dimensional, elastic-plastic finite element on the results of standard penetration tests shown in
calculations. Figure 1. Note that the high NSPT values reported at
Among the various approaches proposed in the lit- two depths are likely to depend on the presence of
erature for the analysis of piles under vertical loads, boulders.
see e.g. (Coyle & Reese 1966; Poulos & Davis Figures 2a and 2b report, respectively, the estimated
1968; Butterfield & Banerjee 1971; Ottaviani 1975; variation with depth of the friction angle ϕ
and of
Randolph & Wroth 1978), the finite element method the elastic modulus ratio E/E* (Mandolini & Viggiani
was adopted here since it can be easily applied to inho- 1997). In the present case the modulus E* coincides
mogeneous deposits accounting for their non linear with that of the first layer and was evaluated on the
behaviour. basis of the load test data.
The first examined case aims at evaluating the load- The numerical results and the experimental data are
settlement curve of a 12 pile group based on the results shown by solid and dashed lines in Figure 3.

589
Figure 1. Results of the standard penetration test.
Figure 3. First load test: experimental results (solid line)
and numerical simulation based on the calibrated finite
element model (dashed line).

Figure 2. Estimated variation with depth of the friction

angle ϕ
(a) and of the elastic modulus ratio E/E* (b).

The numerical analyses were carried out in axisym-

metric conditions modelling both the pile and the 9
layers of the surrounding soil through a mesh of four
node isoparametric elements. To account for the distur-
bance caused by the pile construction, thin interface
elements (Desai et al. 1984) were placed along the
pile contour and at its base. Figure 4 shows a detail of
the finite element mesh in the vicinity of the pile.
An elastic behaviour was assumed for the pile, while
an elastic perfectly plastic constitutive model, obeying
Mohr-Coulomb yield criterion with a non associated
flow rule, was adopted for the soil layers. The modulus Figure 4. A detail of the axi-symmetric finite element
E* (cf. Figure 2b) was calibrated first by matching the meshes in the vicinity of the pile.
slope of the initial part of the load settlement curve.
Note that, since E* is calibrated on the basis of the load analyses are: the angle of plastic dilation ψ, govern-
test, the evaluated variation of the soil elastic modulus ing the flow rule, that depends on the friction angle ϕ

with depth is already influenced by the disturbance due through a parameter δ, i.e. tan ψ = δ tan ϕ
; the reduc-
to construction. Consequently, it was assumed that the tion factor α of the interface friction angle ϕ*, i.e. tan
elastic modulus of each interface element coincides ϕ* = α tan ϕ
; the coefficient of horizontal pressure K0
with that of the corresponding soil layer. The remain- relating the normal stress between pile and soil to the
ing mechanical parameters entering in the numerical vertical stress.

590
Figure 5. Contour lines of the error function: δ dilatancy
parameter, α friction angle reduction coefficient. Figure 6. Detail of the horizontal section of the 3D finite
element mesh.

The above parameters, collected in vector p, were

evaluated through a back analysis that consists in
minimizing the discrepancy F between the n mea-
sured settlements û and the corresponding numerical
results u(p). In order to express F in non dimensional
form the difference between measured and calculated
displacements is divided by the maximum measured
displacement ûmax ,

The back analysis was carried out assigning differ-

ent values to K0 and working out, for each of them,
the values of α and δ that minimizes the function F.
Comments on the minimization algorithms suitable for
back analyses can be found e.g. in (Gioda & Sakurai,
1987). Figure 5 reports the contour line of the error
function for one of such minimizations.
The back analysis process led to the following
values of the sought parameters:
Figure 7. Calculated load-settlement curves: a) simple
E* = 750 MPa; K0 = 0.5; α = 0.7; δ = 0.07. superposition of 12 independent piles; b) pile group and inter-
action between soil and foundation mat; c) pile group without
Having calibrated its material parameters, the soil-mat interaction. (A dashed line represents the working
numerical model was extended to three dimensional load of the foundation).
conditions to investigate the behaviour of the 12 pile
group. Taking advantage of its double symmetry only the underlying soil is accounted for. Finally, curve (c)
1/4 of the problem was discretized into a mesh of 8 represents the 3D case in which the mat-soil interac-
node brick elements. A detail of the horizontal section tion is neglected. Case (c) could represent the case of
of the mesh, in the vicinity of the pile group, is shown an extremely severe erosion of the soil underneath a
in Figure 6. bridge pier.
The calculated load-settlement diagrams are sum- The numerical results show the increase of settle-
marized in Figure 7. Curve (a) represents the mere ments due to the interaction between the piles (curves
superposition of the load-settlement diagram calcu- b and c) with respect to the case in which the interac-
lated for a single pile (cf. Figure 3), i.e. the pile tion is neglected (curve a). In addition, the calculation
interaction is neglected. Curve (b) refers to the actual permits a quantitative assessment of the effects of
3D case in which the pile group is connected to a foun- the interaction between the foundation mat and the
dation mat and the interaction between the mat and underlying soil.

591
Figure 8. Second load test: experimental results (solid line) Figure 9. Axial force from the extensometer measurements
and numerical simulation based on the calibrated finite (solid line) and numerical results based on the calibrated
element model (dashed line). numerical model (dashed line).

In particular, it can be observed that depending

on the assumptions adopted in the calculations the the load-settlement data (cf. Figure 8), a large dif-
expected settlement of the pile group under working ference exists between the computed axial force and
loads (dashed line in Figure 7) varies between 6.5 and that deriving from the experimental data (cf. Figure 9).
10.5 mm. According to the experimental data in Figure 9 a van-
ishing vertical load reaches the pile tip and, hence, the
limit skin friction has not been reached yet.
On the contrary, the calibrated numerical model
3 SECOND LOAD TEST indicates that about half of the applied load is carried
by the base. This leads to conclude that the limit skin
This test was carried out on a 80 cm diameter bored friction was reached, at least for the upper portion of
pile having length of 11.5 m. Since the in situ inves- the pile.
tigation indicates that the soil profile is reasonably In order to overcome the above apparent contradic-
uniform, mechanical properties constant with depth tion and, hence, to limit the axial force that reaches the
were used in the design. The adopted finite element pile base, various attempts were made by modifying
mesh is similar to the one used for the first load case the soil parameters.
and depicted in Figure 4. The most successful one (see Figure 10) consisted
Applying the same procedure described for the in introducing a limited adhesion (50 kPa) at the soil-
previous case, the back analysis led to the follow- pile interface that could depend on the silty fraction
ing parameter characterizing the numerical model: present in the granular deposit.
E = 700 MPa; K0 = 0.6; α = 0.7; δ = 0.03. The conse- Unfortunately this provision eliminates any simi-
quent numerical results are compared in Figure 8 with larity between the experimental and calculated load-
the experimental ones. settlement diagrams, as shown by the diagrams in
In this case, in addition to the load-settlement data, Figure 11. Note, in particular, the difference existing
also the axial strains at three locations along the between points B and A in Figure 11 that corre-
pile where measured through electrical extensometers. spond, respectively, to the experimental and numerical
They were placed at 3.65 m, 7.10 m and 10.15 m from diagrams in Figure 10.
the pile top. The available strain measurements cor- The difficulties met in modelling both sets of exper-
respond to point A of the load-settlement diagram in imental data cannot merely depend on errors in the
Figure 8. values of the soil parameters. They should rather
The distribution of the axial load derived from the depend on some aspects of the field problem that lead
measured strain is compared with the corresponding to the barely relevant force at the pile base.
finite element results in Figure 9. A possible cause of this effect could be the pres-
It can be observed that, while the calibrated numer- ence of a very soft zone at the pile tip that depends,
ical model provides an acceptable approximation of for instance, on the partial cleaning of the excavation

592
Figure 10. Axial force from the extensometer measure- Figure 12. Axial force from the extensometer measure-
ments (solid line) and numerical results with soil-pile adhe- ments (solid line) and numerical results with a soft zone at
sion of 50 kPa (dashed line). the pile tip (dashed line).

Figure 11. Comparison between experimental results (solid Figure 13. Comparison between experimental results (solid
line) and numerical ones with soil-pile adhesion of 50 kPa line) and numerical ones with a soft zone at the pile tip
(dashed line). (dashed line).

bottom or on the presence of a zone where the concrete results, even though some discrepancy can still be
was mixed with the bentonite slurry. observed (cf. Figure 13).
The finite element model was then modified intro- This could depend on the assumed homogeneity of
ducing a layer of soft material below the pile tip and the soil deposit or on the limits of the relatively simple
reducing the soil-pile adhesion to 20 kPa. The corre- constitutive model adopted in the calculations.
sponding numerical results are shown in Figures 12 However, it seems reasonable to conclude that
and 13. the presence of the mentioned soft zone is likely to
Apparently the introduction of a soft zone improves be a possible cause of the measured marginal load
the agreement between experimental and numerical transferred to the pile tip.

593
4 CONCLUSIONS It can be finally observed that numerical modelling
is not only a useful design tool for analysing various
The two discussed case histories show that the back aspects of the interaction between deep foundations
analysis of load tests represents a practical proce- and surrounding soil. Its use in the interpretation of
dure for calibrating the numerical models of deep in situ measurements could also lead to a deeper
foundations. understanding of the effectiveness of a construction
When dealing with pile groups, the parameters technology and, perhaps, to some suggestions for
obtained from the axisymmetric interpretation of the improving its application.
load test on a single pile can be adopted for the three
dimensional analysis of the entire group. In this case,
in addition to the interaction between the piles, the ACKNOWLEDGEMENTS
finite element model can also account for the elastic-
plastic interaction between the foundation mat and the The authors wish to thank Michela Chiorboli and
underlying soil. Mario De Miranda for providing the experimental data
In the second test the numerical analysis highlighted and for their technical comments.
an apparent contradiction of the in situ measurements.
In fact, the strains measured along the pile show that a
limited load is transferred to its base. On the contrary, REFERENCES
the back analysis of the load-settlement diagram indi-
cates that at least half of the applied load reaches the Butterfield, R. & Banerjee, P.K. 1971. The problem of
pile group-pile cap interaction. Geotechnique 21(2):
base.
135–142.
This apparent contradiction is likely to depend on Coyle, H.M. & Reese, L.C. 1966. Load transfer for axi-
some field condition that is not properly accounted for ally loaded piles in clay. Journal of Soil Mechanics and
in the numerical model. Foundation Engineering ASCE, 92(2): 1–26.
In the present context it appears that the men- Desai, C.S., Zaman, M.M., Lightner, J.G. and Siriwar-
tioned discrepancy between experimental and numer- dane, H.J. 1984. Thin-layer elements for interfaces and
ical results can be, at least partially, reduced by joints. International Journal for Numerical and Analytical
introducing in the calculations a soft zone underneath Methods in Geomechanics, 8(1): 19–43.
the pile tip. This zone could be due, for instance, to a Gioda, G. & Sakurai, S. 1987. Back analysis procedures for
the interpretation of field measurements in geomechan-
poor cleaning of the excavation bottom or to the for-
ics. International Journal for Numerical and Analytical
mation of a soft mixture of concrete and bentonite at Methods in Geomechanics, 11(6): 555–583,
the pile tip. Mandolini, A. & Viggiani, C. 1997. Settlement of piled
If this explanation can be accepted, the back anal- foundations.Geotechnique, 47(4): 791–816.
ysis provided an insight into an apparent weakness of Ottaviani, M. 1975. Three-dimensional finite element analy-
the application of the construction technique at that sis of vertically loaded pile groups. Geotechnique 25(2):
specific site. 159–174.
It is worthwhile observing that a more effective Poulos, H.G. & Davis, E.H. 1968. The settlement behaviour
interpretation of the second load test could have been of single, axially loaded, incompressible piles and piers.
Geotechnique 18: 351–371.
reached if the vertical load at the pile tip were directly
Randolph, M.F. & Wroth, C.P. 1978. Analysis of the defor-
measured through a load cell or a flat jack. This mation of vertically loaded piles. Journal of Geotechnical
additional information would have in fact validated Engineering ASCE, 104(12): 1465–1488.
and completed the measurements of the electrical
extensometers.

594