Experimentally derived CPT-based p-y curves for soft clay

F. Guo and B.M. Lehane

University of Western Australia, Perth, Australia
J. Ju
Southeast University, Nanjing, China (visitor to University of Western Australia)

ABSTRACT: This paper presents results from centrifuge scale lateral pile tests conducted on instru-
mented piles jacked into normally consolidated and lightly overconsolidated kaolin clay. These experi-
ments allow derivation of load transfer p-y curves. The curves are then used to develop a p-y formula-
tion, which is expressed as a function of the measured CPT qnet values, the depth and the pile diameter.
The equations provide a good match to experimental data and are also shown to be consistent with re-
cent centrifuge tests and numerical analyses.

INTRODUCTION

The p-y load transfer approach is the most popular method for predicting the response of piles to lateral
load. The p-y curves recommended by the American Petroleum Institute, API (2011) are the Industry
standard and those used for soft clay are based primarily on the backanalyses from field tests performed
by Matlock (1970). Matlock’s p-y curves are related to the clay’s undrained shear strength (su) measured
in an unconfined compressive strength test. In practice, however, sample disturbance effects coupled with
factors such as the rate dependence and anisotropy of undrained strength make the selection of a su profile
difficult. Direct correlations between p-y curves and in-situ test parameters are therefore preferable.
There have been a number of proposals relating p-y curves to in-situ test parameters. For example,
Briaud et al. (1983) show that pressuremeter data correlate strongly with p-y curves. Anderson et al.
(2003) also show that both pressuremeter (and dilatometer) based correlations perform well but argue that
correlations with the CPT are less reliable.
This paper re-examines the potential of using CPT based p-y formulations for normally consolidated
and lightly overconsolidated clay. Results from lateral load tests performed on instrumented piles in the
drum centrifuge are presented and these are used to derive p-y curves for kaolin at overconsolidation ratios
(OCRs) of 1 and 2. The paper explores the relationship between these p-y curves and the measured net
cone resistances (qnet). A direct CPT-based formulation for p-y curves is proposed that fits the kaolin data
very well. This formulation, which is also shown to be generally consistent with the API (2011) recom-
mendations, can be tested against other lateral load test data to examine its general suitability.

EXPERIMENTAL SETUP AND PROCEDURES

The experiments reported here are part of a wide-ranging study into the lateral behaviour of piles at the
University of Western Australia, UWA (e.g. Truong & Lehane 2014, Suryasentana & Lehane 2014). The
particular experiments reported here relate to lateral tests conducted in the drum centrifuge on open-ended
displacement piles in kaolin. A full description of the UWA drum centrifuge and its operation is provided
in Stewart et al. (1998). The first test (T1) was conducted in a normally consolidated sample of kaolin at a
centrifuge acceleration of 80g while the second test (T2) was performed at 40g with OCR=2. The length
to diameter ratio(L/D) in each test was 11.8 and piles were designed so that their capacity was controlled
by the soil strength and not the pile bending capacity, i.e. short pile failure with near rigid rotation about a
point at depth, as described by Broms (1964).
The model pile employed was an aluminium tube with an outer diameter of 10mm and wall thickness
of 1.0mm. Electrical resistance strain gauges were spaced at intervals of about 15mm on the tension and
compression outer faces and arranged in a half bridge configuration. These gauges were protected by a
0.5mm thick epoxy coating making an effective outer pile diameter (D) of 11mm, which at prototype scale
represents 0.88m and 0.44m diameter piles at 80g and 40g respectively. The pile gauges were exercised
and then calibrated for bending moment measurement prior to centrifuge testing. Figure 1a shows the ar-
rangement of gauges and Figure1b shows the lateral loading arrangement at the pile head.

Figure 1. (a) Schematic of centrifuge pile (unit: mm); (b) View of loading arrangement at pile head

Kaolin slurry was placed in the drum channel at a water content of 120% (twice its liquid limit) and then
allowed to consolidate at 80g for a period of about 6 days. The slurry was progressively topped up as con-
solidation progressed and the final consolidated sample height achieved was 165mm. Piles were installed
vertically at a rate of 1mm/s to a final embedment of 130mm and then allowed to rest for a period of at
least 4 hours to enable installation induced excess pore pressures to dissipate; this period was assessed us-
ing the procedure of Randolph (2003) to lead to at least 90% consolidation. Test T1 was then conducted
using the loading arm described by Levy (2007) to apply lateral load (see Figure 1b). The load was ap-
plied via a cubic lightweight hollow cap (push-fitted into the top of the pile) at an (undrained) rate of
0.35mm/s and at a height (e) above the soil surface of 19 mm. Loading continued until the pile head lat-
eral displacement reached 1D. This displacement was inferred from the monitored relative rotation of the
drum’s tool table with respect to the drum channel. The second test (T2) was conducted using the same
procedure after reducing the centrifuge acceleration to 40g and allowing a period of 2 days for effective
stresses to equalize.
Cone Penetration Tests (CPTs) were performed just prior to and after each test (T1 and T2) using a
6mm diameter cone; the UWA centrifuge piezocone was not available at the time of testing. The profiles
of net cone penetration resistance (qnet) measured in these tests (assuming an average kaolin unit weight of
16.5 kN/m3) are plotted on Figure 2. Although some irregularity is evident (possibly reflecting the sample
top-up history), it is seen that qnet can generally be assumed to increase linearly with depth at a rate (kc) of
19.6 kPa/m and 30.3 kPa/m at OCR=1 and 2 respectively.
 Cone net resistance, qnet (kPa)
0 50 100 150 200
0

2
~ 30.3 kPa/m
OCR = 2
4
Depth, z (m)

Pile tip level

6 T2
~ 19.6 kPa/m
OCR = 1
8 Pile tip level
T1

10

12
Figure 2. CPT qnet profiles at OCR=1 and OCR=2

EXPERIMENTAL RESULTS

The lateral load (H) vs. pile head displacement (yh) responses measured in Tests T1 and T2 (and converted
to prototype units) are shown on Figure 3a. As expected, the lateral pile capacity at 80g is significantly
higher than that at 40g. The observed responses are presented in an alternative form on Figure 3b by nor-
malising H by kcD3 and yh by D. This normalized form of presentation shows that the lateral capacity is
fully mobilized at y/D = 0.5 and that this capacity varies directly with kcD3. The latter observation is con-
sistent with the work of Broms (1964) for short piles (for which capacity is a function of L/D), where
strength increases linearly with depth. Hence, the normalised capacity (H/kcD3) of 16 is only applicable to
the particular length to diameter ratio examined (L/D = 11.8).
250 18
16
200 14
12
150
H/kcD3

10
H (kN)

T1
T2 8
100 T1
6
T2
50 4
2
0 0
0 200 400 600 800 0.0 0.2 0.4 0.6 0.8
yh (mm) yh/D
Figure 3. Lateral load test results of T1 and T2 in (a) prototype units and (b) normalized form

The development of bending moment in the piles during the two tests (using equivalent prototype units) is
presented on Figure 4. It is seen that the maximum moments occur at a depth below ground level of about
6D (5.2m in T1 and 2.6m in T2). This depth is typical of piles in soft clay but significantly greater than
usually observed for piles in sand or stiff clay.

M (kN.m) M (kN.m)
-2.0 -1.0
e = 1.52 m e = 0.6 m
0 500 1000 1500 0 50 100 150
0.0 0.0

2.0 1.0
Depth, z (m)
Depth , z (m)

4.0 2.0

6.0 3.0

8.0 4.0
42kN
215kN
156kN 30kN
10.0 5.0
89kN 23kN
33kN 15kN
12.0 6.0
Figure 4. Bending moment profiles measured in tests T1 and T2

Best estimate moment variations with depth (z) below the level of the applied load were fitted to the
measured moments using the following relationship form (where a0, a1, a2…ai are the best fit coeffi-
cients):

1 )( + +… … ) (1)

These best-fit profiles are shown together with the measured M data on Figure 4. The net stress per metre
run on each pile (p) at any depth, z, was then obtained by double differentiation of Equation (1). The net
pressure, P, is obtained by dividing the net stress by the pile diameter. The displacements (y) at the corre-
sponding depth and level of applied load were determined by double integration of Equation (1). The con-
stants of integration were determined from the measured pile head movement (yh) at z = 0 and assuming
that the pile rotated as a rigid body about a point at depth where the net stress (p) equated zero.
The P-y curves derived in this manner are presented on Figure 5 for the two overconsolidation ratios
considered. It is apparent that, although piles were brought to ultimate failure, full ultimate pressures were
not developed in close proximity to the point of rotation, which occurred at a depth of about 9.0D. Net
pressures estimated at very shallow depths are more approximate in view of their very low values (e.g.
less than 3 kPa) and hence their higher sensitivity to the fitted M profile.
The data on Figure 5 are re-plotted on Figure 6 with P values normalized by the CPT qnet value at the
corresponding depth and lateral displacements normalized by the pile diameter. It is interesting to note
that in deeper soil horizons and at both OCRs, where the flow around mechanism dominates, ultimate P
values are almost identical in value to qnet. P/qnet values decrease progressively below unity as z/D reduces
below about 3 where the proximity of the soil surface has an increasing influence on the failure mecha-
nism.
 140 100
z = 1D z = 1D
z = 2D 90 z = 2D
120 z = 3D
z = 3D 80
z = 4D z = 4D
100 70 z = 5D
z = 5D
z = 6D 60 z = 6D

P (kPa)
P (kPa)

80 z = 7D z = 7D
50
60 40
40 30
20
20 10
0 0
0 100 200 300 400 0 50 100 150 200 250
y (mm) y (mm)
Figure 5. P-y curves measured in tests T1 and T2

1.2 1.2
z = 1D z = 1D
z = 2D z = 2D
1.0 1.0 z = 3D
z = 3D
z = 4D
z = 4D z = 5D
0.8 z = 5D 0.8 z = 6D
z = 6D z = 7D
P /qnet

P /qnet

0.6 z = 7D 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5
y/D y/D
Figure 6. P/qnet vs. y/D curves for T1 and T2

1.2

1.0

0.8
Pu/qnet

0.6

0.4 Eqn. 2
OCR = 1
0.2 OCR = 2

0.0
0 1 2 3 4 5 6
z/D
Figure 7. Experimental Npq factors compared with Equation (2)

The ratio of the net ultimate pressure Pu to qnet is referred to here as a bearing factor, Npq. The variations of
these backfigured Npq values with z/D are plotted for both OCRs investigated on Figure 7. Withstanding
some discrepancies at very low z/D values (for the reasons discussed earlier), good agreement between
both sets of Npq vs. z/D trends is evident. The following simple relationship is plotted on Figure 7 and seen
to provide a reasonable match to the data:

0.2 0.3 (2)

Figure 8 presents another set of normalized curves for tests T1 and T2. The net pressure (P) in this in-
stance is normalized by the ultimate pressure Pu. A comparison of these figures indicates that the curves
for OCR=1 are closely comparable to those for OCR=2 (especially at depths greater than about 3D where
exact agreement is evident). Such agreement is consistent with that seen on Figure 3b. The following
equation, which is similar to the hyperbolic tangent form employed for stiff clay by O'Neill et al. (1990) is
seen on Figure 8 to provide a good fit to the normalised curves:
. .
tanh 8.80 0 < z ≤ 3D (3a)

.
tanh 6.34 z ≥ 3D (3b)

Equations (2) and (3) provide a simple means of constructing p-y curves for use in a load transfer analysis
and only require input of the qnet profile at the pile location.
1.2 1.2

1.0 1.0

0.8 0.8
z/D = 0.5 z/D = 1
P/Pu
P/Pu

0.6 z/D = 1 z/D = 1.5 0.6 z/D = 1.5 z/D = 2

z/D = 2 z/D = 2.5 z/D = 2.5 z/D = 3
z/D = 3 z/D = 3.5 z/D = 3.5 z/D = 4
0.4 z/D = 4 z/D = 4.5 0.4 z/D = 4.5 z/D = 5
z/D = 5.5 z/D = 6
z/D = 5 z/D = 5.5 z/D = 6.5 z/D = 7
0.2 z/D = 6 z/D = 6.5 0.2 Eqn.3 ,z/D = 0.5 Eqn.3, z/D = 1
z/D = 7 Eqn.3, z/D = 1 Eqn.3, z/D = 1.5 Eqn.3, z/D = 2
Eqn.3, z/D > 3 Eqn.3, z > 3D
0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4
y/D y/D
Figure 8. Comparison of Equation (3) with normalized P/Pu vs. y/D data in tests T1 and T2

A check on the accuracy of the procedures used in the derivation of Equations (2) and (3) was made by
inputting the p-y curves derived using these equations into a standard load transfer program (ALP,
OASYS 2013); this program, as with many commercially available pile lateral loading programs, models
the pile as a series of beam elements and the soil as a series of non-interacting, non-linear p-y springs. The
ALP predictions were found to match the measured H-yh variations (Figure 3) perfectly.

COMPARISON WITH PREVIOUS RESEARCH AND API (2011)

The Npq bearing factor deduced from the centrifuge tests (Equation 2) can be compared with the more
conventional bearing factor, Np, which is used for lateral pile design in clay and relates the ultimate net
pressure with the undrained strength (su). The conventional bearing factor is given as the product of the
cone factor, Nkt, and inferred bearing factor Npq:
 where (4)

Equation (4) is compared, on Figure 9, with Finite Element predictions for Np presented by Jeanjean
(2009), and with Np deduced by Jeanjean (2009) from similar centrifuge lateral pile tests conducted in
normally consolidated kaolin. Good general agreement is observed assuming a typical Nkt value of 12.
There are, however, some notable differences at z/D <2 and it is not clear at present whether such differ-
ences are real or arise due to errors associated with the estimation of (very low) net pressures at shallow
depths. It is also noteworthy that the maximum Np value recommended by API (2011) depth is only 9 and
significantly lower than indicated on Figure 9.
18
16
14
12
10
Np 8
6 Centrifuge test (Jeanjean 2009)
FE Kaolin (Jeanjean 2009)
4
FE Gulf of Mexico Clay (Jeanjean 2009)
2 Equation 4 with Nkt = 12
0
0 5 10 15
z/D
Figure 9. Measured and predicted variations of Np with z/D

0.8

0.6
P/Pu

0.4 Equation 3 for z/D > 3

API 2011 with ɛc = 0.0075

0.2
API 2011 with ɛc = 0.02

0
0 0.1 0.2 0.3 0.4 0.5
y/D

Figure 10. Comparison of Equation (3) with API (2011)

The P/Pu vs. y/D variation recommended by API (2011) is compared on Figure 10 with the curve given by
Equation (3b) for z/D ≥ 3. Reasonably good agreement is observed provided that the API (2011) εc pa-
rameter is specified as 0.0075. The εc parameter is defined in API (2011) as the strain at one-half the max-
imum stress for an undrained triaxial compressive strength test and the value of 0.0075 is equivalent to a
rigidity index (Ir) of about 135. The εc value of 0.02 recommended by Sullivan et al. (1980) for soft clay
clearly leads to a much softer response than observed.

CONCLUSIONS

The instrumented lateral load tests conducted in the centrifuge in kaolin with OCR=1 and OCR=2 have al-
lowed derivation of a CPT based p-y formulation which is in good agreement with the centrifuge observa-
tions and also with previous centrifuge tests and numerical analysis. Testing of the applicability of the
formulation against other case histories would help to refine the proposed approach and move the ge-
otechnical profession towards the use of a direct CPT-based design method for laterally loaded piles in
clay.

ACKNOWLEDGEMENTS

The authors acknowledge the support provided by the Australian Research Council.

REFERENCES

Anderson, J. B., Townsend, F. C. & Grajales, B. 2003. Case history evaluation of laterally loaded piles. Journal of geotechnical
and geoenvironmental engineering, 129(3): 187-196.
API 2011. Part 4: Geotechnical and Foundation Design Considerations. ANSI/API Recommended Practice 2GEO. 1st ed. USA:
Thomson Scientific.
Briaud, J. L., Smith, T. O. & Meyer, B. J. 1983. Using the pressuremeter curve to design laterally loaded piles. Offshore
Technology Conference, 2-5 May. Houston, Texs. OTC 4501: 495-503.
Broms, B. B. 1964. Lateral resistance of piles in cohesive soils. Journal of the Soil Mechanics and Foundations Division,
ASCE, 90(2): 123-156.
Jeanjean, P. 2009. Re-assessment of p-y curves for soft clays from centrifuge testing and finite element modeling. Offshore
Technology Conference, 4-7 May. Houston, Texas. OTC 20158:
Levy, N. H. 2007. Modelling Multi-directional Behaviour of Piles Using Energy Principles. PhD, The University of Western
Australia.
Matlock, H. 1970. Correlations for design of laterally loaded piles in soft clay. The 2nd Annual Offshore Technology
Conference, 22-24 April. Houston, Texas. OTC 1204: 77-94.
O'Neill, M. W., Reese, L. C. & Cox, W. R. 1990. Soil behavior for piles under lateral loading. The 22nd Offshore Technology
Conference. Houston, Texas. OTC 6377: 279-287.
Oasys. 2013. Alp user manual (Version 19.2). Oysys Ltd. UK.
Randolph, M. 2003. Science and empiricism in pile foundation design. Géotechnique, 53(10): 847-875.
Stewart, D. P., Boyle, R. S. & Randolph, M. F. 1998. Experience with a new drum centrifuge. Proceedings of the International
Conference Centrifuge 98. Bakkerma, Rotterdam, Nethlands. 35-40.
Sullivan, W. R., Reese, L. C. & Fenske, C. W. Unified method for analysis of laterally loaded piles in clay. Numerical Methods
in Offshore Piling, 22-23 May. London, England. 135-146.
Suryasentana, S. & Lehane, B. M. 2014. Verification of a numerically derived CPT-based p-y curves for piles in sand. 3rd
International Symposium on Penetration Testing. Las Vegas. Nevada.
Truong, P. & Lehane, B. M. 2014. Numerical derivation of CPT-based p-y curves in linear elastic perfectly plastic clay. 3rd
International Symposium on Penetration Testing. Las Vegas. Nevada.