Proceedings of the XVI ECSMGE

Geotechnical Engineering for Infrastructure and Development

ISBN 978-0-7277-6067-8

© The authors and ICE Publishing: All rights reserved, 2015

doi:10.1680/ecsmge.60678

Calibration procedure for embedded pile modeling

based on in situ pile load tests
Procédure de calibration pour la modélisation de pile intégrée basée
sur des tests de chargement de pieux in situ
F. Tradigo*1, R. Castellanza2, M. Partovi3 and G.Schreppers3
1
ARUP, Milano , Italy; PhD student at Politecnico di Milano, Milano, Italy
2
Dept. of Earth and Env. Sciences, Università degli Studi di Milano Bicocca, Milano, Italy
3
TNO Diana, Delft, Netherland
*
Corresponding Author
ABSTRACT A specific methodological path for the proper interface law calibration is presented. An extensive parametric study is per-
formed in order to clarify the role of both stiffness and strength parameters in the embedded pile element. The effectiveness of the adopted
frictional law for embedded piles is demonstrated in the capability of simulating a full scale single pile test in layered soil. The pile was in-
stalled in a sequence of saturated sand, silty clay and silty sand. The in situ test refers to a 50 m long pile with a diameter of 2 m loaded
with two Osterberg cell. A proper post-process procedure enabled to measure in situ lateral friction law along depth and, independently, to
evaluate the load-displacement law of the pile tip. A preliminary numerical prediction based on the soil parameters only will be compared
against results obtained from parameters calibrated on the pile test, and their discrepancy will be discussed. Finally, practical suggestions
in order to retrieve the necessary engineering parameters from in situ tests will be given, discussing their influence on numerical analyses.
RÉSUMÉ Un chemin méthodologique spécifique pour l'étalonnage de la loi de l'interface est présentée. Une étude paramétrique
approfondie est effectuée afin de clarifier le rôle des rigidité et de paramètres de résistance dans l'élément de pile intégré. L'efficacité de la
loi de frottement adopté pour les piles intégrées est démontré dans la capacité de simuler un test simple de pile à grande échelle dans le sol
en couches. La pile a été installée dans une séquence de sable saturé, argile limoneuse et de sable limoneux. L'essai in situ renvoie à un 50
m de long poils d'un diamètre de 2 m chargé avec deux Osterberg cellule. Une bonne procédure de post-traitement a permis de mesurer en
droit de frottement latéral situ le long de la profondeur et, indépendamment, pour évaluer la loi de la pointe du pieu charge-déplacement.
Une prévision numérique préliminaire sur la base des paramètres du sol ne sera comparé les résultats obtenus à partir de paramètres calibrés
sur le test de pile, et leur écart sera discuté. Enfin, des suggestions pratiques afin de récupérer les paramètres techniques nécessaires à partir
des essais in situ seront donnés, discuter de leur influence sur les analyses numériques.

cretized soil domains the stiffness heterogeneity in-

troduced by piles.
1 INTRODUCTION

The purpose of this work is to present a methodolog-

ical approach to deal with the calibration of “embed-
ded pile” (EP) elements in FEM analyses. In recent Model = Soil (solid) + Pile (beam) + Interface (line-to-solid)

times, EPs have been implemented in various numer-

ical codes, mainly in order to (i) reduce the computa-
tional cost of analyses (ii) enhance the extrapolation
of pile internal axial forces and bending moments y x
(iii) provide a more efficient tool for design in terms z

of pile location and orientation. EP elements are in

essence 1D beam elements, reproducing within dis-
Figure 1. Embedded pile model and response behaviour.

3771

Downloaded by [ University of New South Wales] on [19/09/17]. Copyright © ICE Publishing, all rights reserved.
 Geotechnical Engineering for Infrastructure and Development

From a practical point of view, the lack of nodal 2 EMBEDDED PILE INTERFACES – SOIL/PILE
connectivity between the piles and the surrounding CONTACT PROPERTIES
soil is very useful especially in optioneering design
phases. Moreover, as illustrated by Figure 1, nodal The EP approach has been introduced extensively in
connectivity along the pile shaft and in correspond- the last decade and successfully applied to a number
ence of the pile tip is automatically restored by of relevant projects (e.g. Murrells et al., 2009). Pile-
means of beam-to-solid or node-to-solid special in- to-soil interfaces can be modelled accordingly to a
terface spring elements, respectively (e.g. Sadek and number of different constitutive models, as shown in
Shahrour, 2004). Figure 2a and 2b with respect to the pile shaft and the
A number of different constitutive models can be pile base, respectively. Generally, reversible non-
considered for pile-to-soil interfaces, ranging from linear behaviours are assumed.
non-linear elastic to elasto-plastic. Nonetheless, the
calibration of interface parameters is not an easy task,
and the role played by e.g. interface stiffness can be
critical from both a mechanical and a numerical point
of view. To this respect, model calibration is often
performed with respect to in situ full scale load pile
tests measurements.
The aim of the present work lies in testing the ap-
plicability of EP element in simulating in situ pile
load tests in layered soil. In particular, the im-
portance of employing frictional interface (FI) ele-
ments along the pile shaft, as recently implemented
in the 3D FEM code DIANA 9.5, is also clarified
with respect to alternative interface models.
In the following sections, a specific methodologi- (a)
cal path for the proper interface law calibration will
be proposed, and the role of both stiffness and
strength interface parameters will be clarified by
means of an extensive parametric analysis.
Analyses refer to a fully instrumented 50m long
and 2m diameter pile installed in a sequence of satu-
rated sand, silty clay and silty sand (Becci and Nova,
2007). Two Osterberg cells were placed at different
depths along the pile shaft and close to the pile base.
Cells were activated independently, in order to un-
couple the measurements of mobilization of the pile
base resistance from the shaft contribution. A proper
procedure enabled to measure in situ lateral friction
law along depth and, independently, to evaluate the (b)
load-displacement law of the pile tip. A preliminary
numerical prediction based on the soil parameters on- Figure 2. Interface behavior for pile shaft (a) and pile tip (b); LE:
Linear Elastic; SF: Non Linear Elastic:.
ly will be compared against results obtained from pa-
rameters calibrated on the pile test, and their discrep-
Figure 3 illustrates the results obtained from a single
ancy will be discussed. Finally, practical suggestions
vertically loaded EP: the mobilization of shaft and
in order to retrieve the necessary engineering pa-
base resistances can be easily identified by a change
rameters from in situ tests will be given, discussing
of stiffness in the global load-settlement curve, which
their influence on numerical analyses

3772

Downloaded by [ University of New South Wales] on [19/09/17]. Copyright © ICE Publishing, all rights reserved.
 Tradigo, Castellanza, Partovi and Schreppers

can be obtained by simply superimposing the and

base contributions. >20 m

datum beam +46.8 m slm

4500

4000
Curve S+T: Shaft + TIP F F F Displacement. transducer
between datum beam and
3500
10 monitored cross
sections by 3 strain
3000
Curve S: Shaft gauges each – 7
section in the upper +26
2500 part
F [kN]

L
C
2000 L
3 ECTs for pile compression A
sampling
1500 Y
+11

1000
Curve T: tip
O-Cell
500 shaft tip shaft&tip 2x3 LWVTs for
cell opening
0 sampling
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045
s [m] O-Cell

Figure 3. Embedded pile model and response behaviour.

In the present case study, frictional interface Figure 4. The Po Viaduct near Piacenza (Italy) and the Osterberg
cell layout for load pile test (adapted from Becci at al., 2007)
springs are employed (MC in Figure 2a), so that the
limit shear stress value τmax is defined as:
σ ' h = k σ 'v
(1) a)
τmax σ'h tan δi + ci ' =k σ'v tan δi + ci '
=
βi =τmax / σ'v
where k is the at rest horizontal pressure coeffi-
cient of the soil domain, while δι and ci are the inter-
face friction angle and cohesion, respectively for the
soil stratum i. Indeed, parameters can vary along the
pile length in order to take into account for the pres-
ence of different soil layers. It is also worth noting
the importance of modelling the dependence of the
maximum shear stress on the normal confining pres-
sure sh. In fact, the behaviour of a single isolated b)
pile is expected to be significantly different from the
case of a capped pile configuration: the presence of
the raft and of surrounding piles is likely to induce a
non-negligible increase of soil stresses along the pile
shaft, so that interface maximum shear stress grows
according to Equation 1.

3 IN SITU TEST

The present case study refers to a design pile load

tests performed, according to the ASTM D1143 Figure 5. a)Pile A: shear stress vs. pile displacement measure-
Quick Load Test Method, on a large diameter pile ments (from Becci at al., 2007); b) Pile A: Lower Osterberg Cell
Load-Displacement Curve (from Becci at al., 2007)
(50 m long), near the actual Viaduct pier at the north
side of Po river. As shown in Figure 4, two Oster-
berg cells were installed in each shaft so that piles Piles were fully instrumented in order to measure
could be loaded almost up to their geotechnical fail- not only the load-displacement values in correspond-
ure. ence of the Osterberg cells, but also the pile head

3773

Downloaded by [ University of New South Wales] on [19/09/17]. Copyright © ICE Publishing, all rights reserved.
 Geotechnical Engineering for Infrastructure and Development

displacement and the pile axial force variations dur- While the calibration of soil and the pile mechanical
ing the test, by means of a displacement transducer behaviour mainly depends on the complexity in-
and strain gauges, respectively. volved in the chosen soil constitutive model, the def-
On the basis of the aforementioned measurements, inition of soil-to-pile interface parameters involves
authors presented the mobilization of shear stresses some tricky steps.
and base resistance during the loading process, as In the following, in order to model the load dis-
shown in Figure 5. In Figure 8 the overall load- placement curve illustrated in Figure 5, four soil stra-
displacement curve of the in situ test is also reported ta with different geomechanical properties are con-
(curve “Pile A”). sidered. The assumed interface properties are
The most relevant soil properties employed during sketched in Figure 6a.
the calibration process are shown in Figure 6. In par-
ticular, β coefficients inferred from in situ measure- 4.1 Calibration #1
ments and Nspt value with depth are illustrated.
0 50
NSPT
100 150
A preliminary analysis is carried out by employing
0
non-linear elastic interfaces. In Figure 7a, shear vs.
10 pile-soil relative displacement curves are inferred on
the basis of in situ measurements presented in Figure
5a. Similarly, a non-linear elastic curve has been em-
20

ployed for the node-to-solid pile tip interface on the

z [m]

30

basis of pile base load test data (Figure 7b).

\

40

In this case, soil compliance is neglected (a very

50
high soil Young modulus is considered), so that all
60
the displacements occur only within the pile-soil in-
terfaces.
Sand-Interpolation Clay-Interpolation
CS008 CS009
CS111 CS113
CS131
0 100 200 300 400 500 400 1 2 3 4
0
Figure 6. Soil parameters: a) coefficient βi (see Eq. 1); b) Nspt tests 5 PILE A -
sampled
350

with depth (data from Becci at al., 2007) 10

DESIGN
VALUES 300 1
DIANA
15 taumax
250
20
τ [kPa]

2
25 200
Z [m]

4 EMBEDDED PILE CALIBRATION ON IN 30

35
150 3
SITU TESTS 40
100
4
45 50
50
In order to perform the calibration of EP elements, a 55
0
0 0.02 0.04 0.06 0.08 dis [m]
a)
number of different aspects need to be attentively
τ max [kPa]

considered. First of all, depending on the soil proper-

ties and on the loading conditions, an appropriate
constitutive model has to be chosen for the soil lay- b)
ers. Secondly, the geometry and the mechanical pa-
rameter of the pile need to be inputted in order to
capture a realistic pile shortening during the loading,
especially when high axial forces are expected to
take place.
Finally, pile-to-soil interface parameters have to be
chosen. In particular, the development of a thin shear
band in the soil surrounding the pile shaft has to be tip
correctly reproduced. It is also worth noting that,
since 1D EP are sizeless, the concept of interface di-
latancy becomes ill-defined/meaningless (this is at Figure 7. Calibration #1: a) Non-linear elastic interface curves for
the shaft; b) Non-linear elastic interface curves for the tip
variance with solid-to-solid interface elements).

3774

Downloaded by [ University of New South Wales] on [19/09/17]. Copyright © ICE Publishing, all rights reserved.
 Tradigo, Castellanza, Partovi and Schreppers

In Figure 8 (curve Einf) a very good agreement be- of magnitude higher than the soil stiffness, any fur-
tween the pile test and the load-settlement curve ob- ther stiffness increment has no effect in the pile load
tained from the model is obtained when an infinitive test curve. To this respect, a scalar index Ferr can be
soil stiffness is considered, testifying the reliability of conveniently defined as the difference between the
in situ measurements. force obtained in the various parametric analysis and
It is apparent that, even if the real in situ load test the force obtained in the case with the highest stiff-
curve is well simulated, the infinitive stiffness of the ness, normalized with respect to the latter. Results
soil strata is unrealistic in order to employ the ob- are illustrated in Figure 9b. Results show that, even
tained parameters to more complex boundary prob- with very high stiffness values, the curve cannot be
lems, e.g. the bridge piled raft foundation presented back analysed.
in Figure 4. In order to overpass this limitation the following
Calibration #2 is adopted: higher soil stiffness values
are considered in combination with frictional inter-
faces.

E=Espt
70000

60000

50000

40000
Q [kN]

30000
Figure 8. Calibration #1: simulation of pile load test curve.
20000

In the following, Soil Young modulus profile es- 10000

timated on the basis of in situ Nspt test (Figure 10, 0
ENspt curve) will be considered for the soil layers. 0 0.05 0.1 0.15 0.2 0.25 0.3

As shown by Figure 8 (curve Espt), previous interface uz [m]

parameters and the new Young modulus profile lead k=1e+3 k=1e+4 Pile A k=1e+5 k=1e+6 k=1e+7 k=1e+

to eccesively overestimate the pile settlements.

The comparison between the two results clarify
the fact that in situ displacement measurements refer
not only to interfaces but to the sum of the settle-
ments occurring at the pile-soil interface and within
the surrounding soil. This aspect has to be taken into
the highest consideration during FE calibration.
Unfortunately, no information are available from
in situ measurements in order to uncouple the two
aforementioned settlements contributions. One may
choose to input very stiff interface parameters in or-
der to model a quasi elastic-perfectly-plastic re-
Figure 9. Calibration #1: a) simulated load curve by increasing the
sponse. In this particular case, interface parameters shear stiffness of shaft interfaces; b) werr vs. interface stiffness.
are used in order to fix a given shaft resistance, while
the single pile stiffness is a function of the soil young 4.2 Calibration #2
modulus. In order to clarify the minimum stiffness
necessary in order obtain negligible pile-to-soil rela- The evidences reported in the previous Section
tive displacements, parametric analyses are carried clarify that, if a deformable soil stiffness layer is con-
out and presented in Figure 9. Results show that, as sidered, in situ pile load test curves cannot be fitted
long as interface shear stiffness is at least one order for any value of the shaft interface shear stiffness.

3775

Downloaded by [ University of New South Wales] on [19/09/17]. Copyright © ICE Publishing, all rights reserved.
 Geotechnical Engineering for Infrastructure and Development

Even though it could be possible to increase the tip 5 CONCLUSIONS

normal stiffness, in the presence of very long piles no
significant variations are expected to take place. In A specific methodological path for the proper inter-
fact, load is transferred to the tip only when relevant face law calibration is presented. A case study related
pile displacement are obtained. Consequently, the to an in situ pile load test (50 m long pile with a di-
initial stiffness of the simulated load-displacement ameter of 2 m loaded with two Osterberg cell) is con-
curve would not change as a function of the tip stiff- sidered in order to validate the calibration of a single
ness parameter. embedded pile. Two different calibration procedures
The soil stiffness can be increased as shown in are presented and an extensive parametric study is
Figure 10, according to G-γ curves presented by performed in order to clarify the role of both stiffness
Seed and Idriss (1970). The pile load test curve can and strength parameters in the embedded pile ele-
therefore be easily back analyzed, as shown in ment interface with respect to the soil interfaces. The
Fig.11. effectiveness of the adopted frictional law for em-
During Calibration #2 the frictional interfaces de- bedded piles is demonstrated in the capability of
fined in Equation 1 and reported in Fig. 2a have been simulating a full scale single pile test in layered soil.
adopted, assuming the values of βi shown in Fig. 7a The soil stiffness of the strata is alighted to be the
(design β). In this way, the effect of the confining critical point to get a satisfactory simulation of the in
pressure is properly taken into account and the em- situ pile load test. As a consequence a slightly higher
bedded pile approach is suitable to better reproduce values are considered as it results from a back-
the raft and group pile influence. analyses procedure. The presented calibration proce-
E(kN/m2) dure should be considered as a sort of academic exer-
0 50000 100000 150000 200000 cise to get practical suggestions in order to retrieve
0
the necessary engineering parameters from in situ
10 pile load tests when embedded piles are considered.
20

z(m)
30
ACKNOWLEDGEMENT
40 The financial and technical support provided by Arup
to the first author's PhD research is gratefully
50
acknowledged; notes by by prof. Claudio di Prisco
60 and eng. Bruno Becci are also gratefully
70
acknowledged.
ENspt E=Ecalibr

Figure 10. Calibration #2: a) calibrated values for the soil strata
stiffness (E calib);
E=Ecalibr REFERENCES
70000

60000
Becci, B., Nova, R., Baù, A. and Haykal, R. (2007), Prove di
50000 carico su pali di grande diametro mediante l’impiego di celle
Osterberg, Rivista Italiana di Geotecnica (RIG), 41(4), 9-28, in
40000
Italian.
Q [kN]

30000 Murrells, C., Ibrahim, K. & Bunce, G. (2009), Foundation design

20000
for the Pentominium tower in Dubai, UAE, Proceedings of the
ICE - Civil Engineering, 162 (6), 25–33.
10000
Sadek, M. & Shahrour, I. (2004), A three dimensional embedded
0 beam element for reinforced geomaterials, International Journal
0 0.05 0.1 0.15 0.2 0.25 0.3
for Numerical and Analytical Methods in Geomechanics, 28 (9),
uz [m]
931–946.
Pile A E=Ecalibr Seed, H. and Idriss, I. (1970), Soil moduli and damping factors for
Figure 11. Calibration #1: simulations of pile load test curve dynamic response analyses, Rap. tecn., Earthquake Engineering
Research Center, Berkeley, California.

3776

Downloaded by [ University of New South Wales] on [19/09/17]. Copyright © ICE Publishing, all rights reserved.