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

net/publication/281991521

P-y curves on model piles : Uncertainty identification

Article · June 2010

CITATION READS

1 115

4 authors, including:

Frédéric Rosquoët L. Thorel

Université de Picardie Jules Verne Université Gustave Eiffel
21 PUBLICATIONS 508 CITATIONS 197 PUBLICATIONS 2,510 CITATIONS

SEE PROFILE SEE PROFILE

Jacques Garnier
Université Gustave Eiffel
49 PUBLICATIONS 662 CITATIONS

SEE PROFILE

All content following this page was uploaded by Frédéric Rosquoët on 01 September 2021.

The user has requested enhancement of the downloaded file.

P-y curves on model piles: uncertainty identification
F. Rosquoët
Laboratoire des Technologies Innovantes : Equipe Phénomènes de Transferts et Construction Durable, IUT
d’Amiens, France
L. Thorel, J. Garnier, M. Khemakhem
Laboratoire Central des Ponts et Chaussées, Nantes, France

ABSTRACT: The effect of horizontal cyclic loading applied to a pile has been investigated on centrifuge
models with the aim of determining the P-y reaction curves. The instrumented pile allows the bending mo-
ment profiles to be obtained. The uncertainty on the moment measurement is first quantified. The displace-
ments are determined from a polynomial function fitted on the moment data and calculated by double-
integrating the moment. An analytical integration type is used. The soil reaction is obtained through double
derivation: the moment data are interpolated using quintic Splines, which are derived twice. The Spline func-
tions are selected by studying the static equilibrium of the pile in force and moment. An analysis of the pile
response through the obtained P-y curves is discussed including uncertainty of measurements. The principle of
the analysis is to introduce an uncertainty on the experimental bending moment data, and to quantify its effect
on the P-y curve.

d2y Mz  2
1 INTRODUCTION Mz   EI 2
 yz     dz (1)
dz EI
Lateral cyclic loading on vertical piles is generally
dMz  d 2 M z 
produced by waves, wind, boat accosting and moor- and  Tz   Pz   (2)
ing to quays, variable overloads or thermal dilata- dz dz 2
tions. Centrifuge modelling method enables a para- If it is relatively easy to calculated the uncertainty on
metric study with realistic boundary limits and the pile head displacement (obtained with the uncer-
control of the testing conditions (Garnier, 1995). Se- tainty on displacement sensor) and the maximum
ries of works on pile in sand have been carried out in bending moment (Rosquoët, 2008), it is more diffi-
the L.C.P.C. centrifuge on laterally loaded piles: sin- cult to estimate the effect of this uncertainty on the
gle pile (Mezazigh et al., 1994; Mezazigh, 1995), double derivation and double integration.
group of piles (Remaud et al., 1998; Remaud, 1999), The present study is conducted to examine the ef-
single pile under impact (Bonab, 2003), single pile fects of uncertainty on the obtained P-y curves. Fi-
under cyclic loading (Rosquoët, 2004; Rosquoet et nally, it should be mentioned that the experimental
al., 2007) and seismic loading (Chenaf, 2007). All conditions, in which all tests are carried out, corre-
tests are carried out using the L.C.P.C. centrifuge spond to real load service conditions (far from fail-
(Corté & Garnier, 1986) on piles installed in sand. ure).
The soil-structure interaction model, is classically
used to design laterally loaded piles. This model,
which requires the knowledge of the P-y curves, is 2 EXPERIMENTAL DEVICE
based on the elastic beam theory to determine the
pile lateral displacements y(z), and the bending mo- A 1:40-scale model pile (Figure 1) has been instru-
ments M(z). A similar method is used for the inter- mented with 20 pairs of strain gauges facing each
pretation of the experimental results. However, in other every 15 mm (figure 2). Thanks to a prelimi-
this case the bending moments are measured where- nary calibration, the bending moment versus depth
as the pile lateral displacement, y(z), and the soil re- profile is achieved for twenty different depths.
action profile, P(z), are calculated by double integra-
tion (Eq. 1) and double derivation (Eq. 2) of the
bending moments, respectively.
 the pile and the electric signal by a deformation of
strain gauges the bending moment is:

M z i  4
EI ui
10 3 (3)
B u
With E the Young’s modulus, I the bending stiff-
ness, ui the electric tension, u the supply electric ten-
sion, Ω the gauge factor and B the pile diameter.
The pile is driven into the sand sample at 1g
(earth gravity) before rotating the centrifuge.
The force controlled load is applied with a servo jack
and transmitted to the pile through a cable (Figure 3)
4 cm above soil surface.

Figure 1. Instrumented model pile

Figure 3. One way loading device

3 HYPOTHESES
3.1 P-y curve constructions
Practically, displacements are determined from a
polynomial function fitted on the moment data. Dis-
placement is calculated by double-integrating the
moment using two integration constants achieved
from the boundary conditions at the pile cap. An an-
alytical integration type is used, with two integration
Figure 2. General view of the model pile constants (boundary conditions on head and tip of
the pile) to ”rise” to the displacement. The solution
Tested with an acceleration of 40g, the scale model for P(z) is obtained through double derivation car-
represents a full size prototype pile with a external ried out by the Slivalic-LCPC piece of software
diameter of 0.72 m, a length of 12 m and a bending (Degny, 1985). The moment data are interpolated us-
stiffness of 476 MN.m2. Confirmed with former ex- ing quintic Splines, which are derived twice. The
periments (Remaud, 1999), the assumption that the Spline functions are selected by studying the static
pile, in dense to medium dense sand, is flexible (or equilibrium of the pile in force and moment.
long) is used to set the boundary condition M = 0 at
the pile tip. 3.2 Uncertainty
The ratio between the pile diameter B and the
Fontainebleau sand d50 is 18/0.2 = 90 larger than the It is assumed that the measurements are only sub-
minimum value of about 45 needed for eliminating jected to random uncertainties and that the limiting
any grain size effects (Garnier et al., 2007). distribution generated by a great number of meas-
To consider the elastic linear behaviour of the pile ured values x go to only one maximum X, while the
material, there is a relationship between the load ap- other values are regularly distributed here and there
plied Q in the extremity of the beam and the electric of this maximum value thus following a normal law
data ui measured by each pairs of strain gauges. At (or Gauss function) (Taylor, 1982). The probability
each level zi, according to the bending stiffness of that a value x is located between an interval [a,b] is
 the equation (4) to introduce the standard deviation mm) is known with uncertainty of 0.5 mm. Uncer-
σ. tainty of calibration mass is equal to the precision of
xX 2
the digital scale: 0.1 g (  P). Uncertainty due to the
position of the strain gauge is  0.6% for the ex-
 
b 
Px  a,b  1  2
2
2
e dx (4)
a treme position (zi = 307 mm) of the strain gauge and
for a calibration mass of 257 g.
3.3 Uncertainty on bending moment Uncertainty due to the installation of the pile in
measurement the soil (rotation during the driven) is negligible.
This effect has been estimated using finite element
It is possible to characterize three types of uncertain-
calculation with CESAR-L.C.P.C. (Rosquoët, 2004).
ty (NF E11-062, 1985) during the measurement of the
Finally, uncertainty on maximum bending mo-
bending moments:
ments is equal to 3.7%.
- uncertainty related to measurements;
- uncertainty due to the positioning of the gauges;
- uncertainty due to the installation.
4 P-y CURVES
Uncertainty of measurement can be given starting
from the data obtained during the calibration. The 4.1 Constructions
assembly carried out during this operation, is canti-
The theory of beams makes it possible to obtain the
lever beam type. The bending moment is calculated
displacement y(z) and the soil reaction profile P(z)
by equation (5).
by respectively a double integration and a double
 2Y derivation of the bending moments. Displacements
M x  EI x (5) are determined by a polynomial function fitted on
z 2
the moment data. An analytic integration type is
If we consider that the behaviour of the material used, with two integration constants (boundary con-
is elastic linear, a relation exists between the load P ditions on head and tip of the pile).
applied to the extremity of the beam and the electric However the double derivation is much more diffi-
signal ui measured by each pair of gauges. cult. In mathematics, the result of the double deriva-
If uncertainties are independent and random, equa- tion is the curvature of the curve. The recorded val-
tion (6), the propagation of uncertainties allows us to ues of bending moment vs. depth must first be
write: properly fitted by a function that will then be derived
twice. King (King, 1994) has demonstrated that a
E 2 I 2 B 2
         single polynomial function is not satisfactory and
Mmzi  E   I   B  (6) that it would be better to use a quintic spline. Cubic

M zi 2 2
  k i  Q 
2
splines have indeed been used successfully to fit the
       
    ki   Q  bending moment profiles.
In the present analysis, the double derivation is
with the following uncertainties (NF ENV 13005, carried out using the software Slivalic-LCPC
1999): (Degny, 1985). The selection of the Spline function
  E : on Young modulus of aluminum (  0.7%) ; is done by minimizing the elastic energy stored into
 I : on moment of inertia (  1.8%) ; the bent pile and by controlling the static equilibrium
 B : on diameter (  0.6%); of the pile in force and moment.
 Φ : on the power tension of the strain gauge However, before analyzing the obtained P-y reaction
(  0.01%) ; curves and studying the cyclic loading effects, these
 ki : uncertainty of strain gauge coefficient (negli- experimental P-y curves are first validated by back
gible); analyzing the response of the pile as described be-
 Q : on calibration mass (  0.04%). low.
Finally uncertainty related to moment measure-
ments is
4.2 Validation
M  z i  3.1%  M  z i (7)
The obtained P-y static curves are validated by back
Uncertainty due to the position of the strain gauge analysis, thanks to the Pilate-LCPC software (Ro-
can be calculated starting from equation (8) : magny, 1985). This software is commonly used in
practice to design horizontally loaded piles from P-y
M z i a z i P curves derived from in situ pressuremeter test
   (8)
M  z i a  z i  a  z i  P (M.E.L.T. French code of practice Fascicule 62).
The experimental P-y curves and the boundary con-
We understand the strain gauge position with un- ditions are introduced into Pilate. The obtained re-
certainty of 0.1 mm (  zi). The distance between sults (bending moments, shearing loads, pile dis-
embedding and the load application point (a = 407 placements and soil reactions versus depth) are
compared with experimental data for each load in- Displacement (mm)
crement. A good agreement is observed between re- -20 0 20 40 60 80 100 120
sults calculated by Pilate and experimental data. At 0

the maximal static applied load F (960 kN prototype 2

value), the difference is always less than 10 % as al-
so demonstrated by Remaud (Remaud, 1999). 4   M z    3.7%dz

Depth (m)
6   M z    3.7%dz
5 EFFECT OF THE UNCERTAINTY ON P-Y 8

CURVES 10 Prototype value

Test P32
It is very difficult to directly calculate the conse- 12 ID = 86 %
quences on reaction P and displacement y of uncer- Figure 5. Effect on displacement profile (y) generated by sys-
tainty on the bending moment measurement. In fact tematic +3.7 % and -3.7 % uncertainty on bending moment
the classical methods of uncertainty based on uncer-
tainty propagation are unsuitable for such a problem. Soil reaction (kN/m)
We suggest to compute, from an example, the influ- -200 -100 0 100 200 300 400 500 600
0
ence that an uncertainty on the bending moment
could have on the soil reaction and pile displace- 2 d 2 M  z    3.7% 
ment. dz 2
The uncertainty on the bending moment meas- 4

Depth (m)
urements (3.7 % generated by uncertainty related to d 2 M  z    3.7% 
6
dz 2
measurements, uncertainty due to the positioning of
the gauges and uncertainty due to the installation) is 8

in the first step put systematically on all recorded 10 Prototype value

values of moment and then in a second step included Test P32
randomly. 12 ID = 86 %
To simplify this study of sensibility, only the case of Figure 6. Effect on soil reaction (P) generated by systematic
a pile under the monotonous maximum load is con- +3.7 % and -3.7 % uncertainty on bending moment
sidered.
Table 1. Uncertainty on pile displacement (y) and soil reaction
(P) generated by +3.7 % and -3.7 % systematic uncertainty on
5.1 Systematic uncertainty values of bending moment.
We generate a 3.7 % systematic uncertainty on the Depth y values Uncertainty P value Uncertainty on
(m) (mm) on y (%) (kN/m) P (%)
bending moment measurement and we observe the (m)
effect on the calculated displacement and the soil re- 0 96.76 1.51 133.36 3.70
action after the double derivation and double integra- 1.2 59.22 3.96 472.37 3.70
tion procedure. The uncertainty on the displacement 2.4 29.83 9.83 537.54 3.70
and the soil reaction is the relationship between the 3.6 9.76 32.48 266.99 3.70
values calculated from the experimental measures of 4.8 -1.84 167.90 9.91 3.70
moment (considered as being the "true" values) and 6.0 -7.03 39.59 -76.11 3.70
the values obtained by modifying the moments of 7.2 -8.06 28.73 -138.70 3.70
+3.7 % and –3.7% (Figures 4, 5 and 6 and Table 1). 8.4 -6.84 25.83 -152.76 3.70
9.6 -4.68 25.25 -116.61 3.70
Bending Moment (kN.m) 10.8 -2.33 25.34 -56.71 3.70
-200 300 800 1300 1800 2300 2800 3300
0
It may be seen that a systematic uncertainty of 3.7 %
2 -3.7 % on bending moment will be the same magnitude as
4 the uncertainty on the soil reaction P. The double
Depth (m)

derivation is not affected by a systematic uncertainty.

6 +3.7 % This result seems coherent. Indeed the curvature cal-
8
culated during the double derivation. remains the
same in spite of the translation of the curve of the
10 Prototype value uncertainty factor.
Test P32 On the other hand. the same systematic uncertain-
12 ID = 86 %
ty seems to have much more effects on the pile dis-
Figure 4. Systematic uncertainty of +3.7 % and -3.7 % on ex- placement. The uncertainties vary between 1.5 and
perimental value of bending moment 170 %. The double integration being a surface calcu-
lation. a small modification of the moment values.
especially in the zone of the largest moment. can Soil reaction (kN/m)
significantly modify the result. -200 -100 0 100 200 300 400 500 600
0
However. it is important to notice that the most
important uncertainty (170 %) is obtained at a depth 2
of 4.8 m which corresponds to the pile rotation point
and to extremely small soil reaction. The pile dis- 4

Depth (m)
d 2 M  z    3.7% 
placement errors in this zone will not have percepti- 6 dz 2
ble effects on the study of the pile response to lateral
loads. 8

10 Prototype value
5.2 Random uncertainty Test P32
12 ID = 86 %
The calculation of the effects of a systematic uncer- Figure 9. Effect on soil reaction (P) generated by random
tainty on the double derivation and double integra- 3.7 % uncertainty on bending moment
tion procedures shown in the previous paragraph
gives some information on the phenomena. Howev- Table 2. Uncertainty on pile displacement (y) and soil reaction
(P) generated by a ±3.7 % random uncertainty on experimental
er. uncertainties are generally random. To investigate values of bending moment.
this other aspect. we allocated randomly a 3.7 %
Depth Uncertainty on Uncertainty on
change to some measures of moment (between 5 and (m) y (%) P (%)
10). We present (figures 7. 8 and 9 and table 2) the 0.0 0.38 34.39
effect of this random uncertainty on pile displace- 1.2 1.05 0.97
ment and soil reaction. 2.4 2.86 - 8.92
Bending Moment (kN.m) 3.6 9.67 -8.71
-200 300 800 1300 1800 2300 2800 3300
0
4.8 - 48.95 298.05
6.0 - 11.30 10.60
2 Random 7.2 - 8.18 - 1.62
±3.7 % 8.4 - 7.39 - 0.71
4
9.6 - 7.18 - 0.70
Depth (m)

6 10.8 - 7.18 - 1.54

8 A random uncertainty modifies the curvature of

10 the bending moment versus depth and the results ob-
Prototype value
Test P32 tained after the double derivation. The smallest soil
12 ID = 86 % reaction are more affected by uncertainties on bend-
Figure 7. Random uncertainty of 3.7 % on experimental val- ing moment (layer near to the surface or at depths
ues of bending moment close to the pile rotation point).
The pile displacement seems less sensitive to ran-
Displacement (mm) dom uncertainty. Indeed. the calculation of the sur-
-20 0 20 40 60 80 100 120
0 face is less affected when some points are only per-
turbed by the uncertainty. We can nevertheless
2
notice that the systematic uncertainty will always
4 limit the random uncertainty on displacement.
Depht (m)

8   M z    3.7%dz 6 CONCLUSION

10 Prototype value It is rather difficult to directly calculate the impact

Test P32 on pile displacement and on soil reaction of the un-
ID = 86 %
12
certainties on recorded bending moments during ei-
Figure 8. Effect on displacement profile (y) generated by ran- ther true scale or centrifuge model tests. Neverthe-
dom 3.7 % uncertainty on bending moment less. it is essential to quantify the errors that may
exist on the P-y reaction curves before any analysis
and use of these experimental P-y curves.
The present study is carried out on an example of
centrifuge pile lateral loading test. The prototype pile
installed into a medium dense sand is 12m long and
0.72m in diameter.
The uncertainty on bending moment measured Norme française NF E11-062 1985. Comparateurs électro-
through 20 pairs of strain gages data placed on the niques. Capteurs de « déplacement linéaire ». Méthodes
d’essai. AFNOR/CSTB. 11 p. (in French).
pile model is small and about 3.7%. Norme française NF ENV 13005 1999. Guide pour
The study shows how this uncertainty propagates l’expression des incertitudes de mesure. 118 p. (in French).
towards the pile displacement (y) and the soil reac- Remaud. D.. Garnier. J. & Frank. R. 1998. Laterally loaded
tion (P). Only one random distribution of error has piles in dense sand. Group effects. Centrifuge 98. Tokyo.
been considered so to say that systematic error is al- Balkema. Kimura et al. (Ed.) : 533-538.
ways worst has only been proved for the specific Remaud. D. 1999. Pieux sous charges latérales : étude expéri-
mentale de l'effet de groupe. Thèse de Doctorat. Université
case considered de Nantes. S.P.I.. Génie Civil. France. 328 p. (in French).
In shallow or deep layers. the resulting uncertain- Romagny. J.C. 1985. Programme de calcul d’un pieu isolé
ties are variable but always smaller than 35 %. At in- soumis à des efforts de flexion en tête et à des poussées la-
termediate depths. close to the pile rotation point térales de sol (Notice d’utilisation). L.C.P.C Paris. 1ère
(depth of 4.8m). errors on the pile displacement and version J.L. Bangratz ; R. Frank ; M. Kutniak. 69 p. (in
soil reaction may be much larger. However at such French).
Rosquoët. F. 2004. Pieu sous charge latérale cyclique. Thèse de
depths pile displacement and soil reaction are very Doctorat. Ecole Centrale de Nantes. Université de Nantes.
small and do not carry weight on the pile equilibrium France. 305 p. (in French).
and pile response to lateral loads. Rosquoët. F.. Thorel.L.. Garnier.J.. Canepa. Y. 2007. Lateral
It should be mentioned that the results of the pre- cyclic loading of sand-installed piles. Soils & Foundations.
sent study are only valid when the uncertainty (or er- Paper N°.3284. 47(5):821-832.
ror) on the measured bending moment are small (a Rosquoët. F. 2008. (in French). Pieu sous charge lateral cy-
clique : Prise en compte des incertitudes de mesures. Jour-
few per cent). nées Nationales de Géotechnique et de Géologie de
l’Ingénieur JNGG’08 - Nantes. 18-20 juin 2008. 8 p. (in
French).
REFERENCES Taylor. J. 1982. Incertitudes et analyse des erreurs dans les me-
sures physiques. Dunod. Paris. 2000. pour la traduction
Bonab. M.A. 2003. Modélisation physique et numérique d'un française. (Edition original: An introduction to error anal-
pieu isolé dans un sable soumis à un impact latéral en tête. ysis. 1882. University Science Books). 315 p.
Thèse de Doctorat. Université de Caen. 227 p. (in French).
Chenaf. N. 2007. Interaction inertielle et interaction cinéma-
tique Sol-Pieu. Thèse de doctorat. Université de Nantes.
329 p. (in French)
Corté. J.F. & Garnier. J. 1986. Une centrifugeuse pour la re-
cherche en géotechnique. Bulletin de liaison Laboratoire
des Ponts et Chaussées. n°146. Déc. 1986 : 5-128. (in
French).
Degny. E. 1985. Slivalic 5. Programme de lissage par spline
quintique (Notice d’utilisation). F.A.E.R. 1.05.10.4.
L.C.P.C.. Ministère de l’Urbanisme du Logement et des
Transports. 24 p. (in French).
Garnier J. 1995. Modèles réduits en mécanique des sols. Les-
modèles réduits en Génie Civil. Colloque A.U.G.C. : 21-44.
(in French).
Garnier.J.. Gaudin. C.. Springman. S.. Culligan. S.M.. Goo-
dings. D.. Konig. D.. Kutter. B.. Phillips. R.. Randolph.
M.F.. & Thorel. L. 2007. Catalogue of scaling laws and si-
militude questions in geotechnical centrifuge modelling.
IJPMG. ISSN 1346-213x. 3: 01-24.
King. G.J.W. 1994. The interpretation of data from tests on lat-
erally loaded piles. Centrifuge 94. Leung C.F.. Lee F.H.
and Tan T.S. editors. A. A. Balkema publishers. Rotterdam.
Netherlands :515-520.
M.E.L.T. Fascicule 62. Titre V 1993. Règles techniques de
conception et de calcul des fondations des ouvrages de Gé-
nie Civil. Cahier des clauses techniques générales appli-
cables aux marchés publics de travaux. Ministère de l'équi-
pement. du Logement et des Transports. 182 p. (in French).
Mezazig. S.. Garnier. J.. Favraud. C. & Levacher. D. 1994. Ef-
fect of slope and soil density on p-y reaction curves for piles
in sand. Centrifuge 94. Leung et al. (Ed). Balkema : 101-
108.
Mezazigh. S. 1995. Etude expérimentale de pieux chargés laté-
ralement : proximité d'un talus et effet de groupe. Thèse de
Doctorat. Université de Nantes. 272 p. (in French).

View publication stats