Professional Documents
Culture Documents
Laterally Loaded Rock Socket Design - A Review of The P-Y Curve Design Approach
Laterally Loaded Rock Socket Design - A Review of The P-Y Curve Design Approach
Laterally Loaded Rock Socket Design - A Review of The P-Y Curve Design Approach
Laterally loaded rock socket design – A review of the P-y curve design approach
ABSTRACT: While laterally loaded piles socketed into bedrock are common, designing them is still an evolving area of geotechnical
engineering. Pile testing for lateral loading is relatively uncommon for rock socketed piles and even when tests are carried out, often
they are terminated once the piles fail structurally; without establishing the ultimate geotechnical capacity of the bedrock. Secondary
structures in rock mass are often not explicitly considered in back-analysing test data to validate design approaches.
This paper briefly reviews available design methods for laterally loaded rock sockets and then discusses the P-y curve approach in
more detail. Limitations of widely used P-y curve models are discussed and an interim approach to using available P-y models
adaptively is presented. Validation of the proposed approach is presented via back analyses of laterally loaded pile tests available
from published literature and parametric studies using 3-Dimensional numerical analyses.
KEYWORDS: lateral loading, piles, rock sockets, p-y curves, secondary structures
Table 2. P-y curve models available for laterally loaded rock socket design in commercial computer programs
P-y Curve Available in The guidance provided on Applicability
Weak Rock Model Pygmy, LPILE, RSPile, MPILE, UCS between 0.7 MPa and 3.5 MPa
Reese 1997 PileLAT, Oasys ALP
Massive Rock Model Pygmy, LPILE, RSPile, MPILE, No guidance is provided. However, the Hoek-Brown
Liang 2009 PileLAT, Oasys ALP approach was originally developed for UCSintact greater
than 15MPa and GSI range 30 to 75.
Vuggy Limestone Pygmy, LPILE, RSPile, MPILE, UCSintact between 6.9 MPa and 17.2 MPa
Nyman 1978 PileLAT, Oasys ALP
Abbs 1984 PileLAT Applicable for porous, carbonaceous rocks with brittle
behaviour at small strains and ductile behaviour at large
strains. UCS range between 0.5MPa to 5 MPa
Fragio et.al. Pygmy, PileLAT Brittle failure at shallow depths and ductile behaviour at
1985 deeper depths. UCS Range 9 MPa to 36 MPa. Uses rock
mass shear strength for assessments.
• A linear segment up to pile deflections y = yA resistance. Typically, ultimate lateral resistance for a
• A non-linear segment from y = yA to y = yrm x 16 given rock strength (UCS) would decrease with RQD.
• A constant resistance of Pur if y > yrm x 16 Therefore, Reese’s relationship is counter-intuitive and
may lead to inconsistent results. For instance, for a given
rock UCS, a rock with a lower RQD will yield more
lateral capacity than a rock with a higher RQD.
• Beyond the recommended strength range (i.e., Reese
originally recommended a strength range of 0.5MPa to
5MPa, and the recent LPILE Manual recommends a range
of 0.7MPa to 3.5MPa); Reese’s weak rock model
y =k b underpredicts pile deflections in many cases as noted by
rm rm
several authors (e.g., Gabr et.al. 2003, Cho et.al. 2001,
krm – constant ( 5e-4 to 5e-5) Yang 2006, Vu 2006). Back-analyses carried out for this
paper (e.g., See Figures 7, 10 & 11) also confirm that the
b - pile diameter
Weak rock model does underpredict the deflections in
several cases.
• Within the recommended range ( 0.7 MPa to 3.5MPa), the
results have been variable. Parametric studies were
carried out using the computer program Plaxis 3D, with
Figure 2. Graphical representation of P-y Curve for Weak Rock (Reese
1997) Mohr-coulomb elasto-plastic models to compare
defection profiles with the weak rock model (for UCS
Reese (1997) alluded to the limitations of this model values 1MPa, 3MPa & 5MPa). For a given rock strength
and referred to this model as an “interim model”. Given the
and stiffness, the results agree reasonably well with
continuing widespread use (or abuse) of this model, some key
limitations discussed by Reese are presented below: Reese’s weak rock model, provided appropriate krm
(Figure 2) values are selected ( See Figure 3 below).
• The model is supported by a limited number of full-scale However, there are other instances where the Weak rock
load tests ( namely two, Speer 1992 and Nyman 1978). In model provided a much stiffer response (e.g., Figures 6
our opinion, these two tests are “non-typical” compared & 9).
to most day-to-day pile designs. For instance, Speer 1992
was done on a 2.25m diameter pile, which may have
resisted the load partly via axial shear caused by the pile
rotation instead of pure bending ( Reese refers to M.W.
O’Neill, 1996 in postulating this as a possible
mechanism). As for Nyman 1978 test, it was done on
Limestone for which estimating mass modulus and mass
strength had been a challenge due to the presence of voids
(Vugs) in the Limestone.
• The model was calibrated using Initial tangent moduli
from pressuremeter tests. Reese (1997) alludes that the
initial modulus may have been underestimated due to the
scale effects (i.e., between the pressuremeter probe and
the 2.25m diameter pile.).
• Reese alludes to the fact that the slope of the initial portion
of the p-y curve (Kir in Figure 2) is “very stiff” compared
to available theories at that time. It is plausible that any
underestimation of modulus, as discussed in the bullet
point above, may have resulted in a high Kir value as data
from this test is principally used to define Kir.
In addition to issues highlighted by Reese, the following issues, Figure 3. Load Deflection Behaviour | Reese Weak Rock Model Vs
principally related to consistency in design, are also Elasto – Plastic (Plaxis 3D)
noteworthy:
• Using this model in combination with other P-y curve
• The Weak Rock model scales down the ultimate lateral
models (for rock) in a project may lead to inconsistent
resistance (Pur) of the rock by applying a strength
design outcomes. For instance, as shown in Figure 4
reduction factor (αr) based on RQD (i.e., αr = 1, if RQD
below, the Weak Rock model for poor-quality Sandstone
is zero and αr =1/3, if RQD is 100). Reese postulates that
(Class V, as per Pells, 1998) provides a stiffer response
when RQD is zero, no further strength reduction is
possible as the rock is already fractured; and when RQD
is high, a reduction factor is warranted due to the
possibility of initial brittle fractures reducing the available
Proceedings of the 14th Australia and New Zealand Conference on Geomechanics, Cairns 2023 (ANZ2023)
than what Massive Rock model (see Section 3.3) where b is pile diameter) and a relatively low rock stiffness
provides for high-quality (Class II) Sandstone. beyond 0.0004b (see Figure 5 below).
Whilst the original Hoek-Brown criterion was actual failure mechanism may be brittle rather than inter-block
developed using test data with UCSintact > 15 MPa; Subsequent shear failure.
researchers (e.g., Carter & Marinos, 2014) suggest that The following recommendations are provided in
reasonable results may be obtained up to UCSintact > 5-10 MPa. using the massive rock model based on the back-analyses of
As all the six tests considered by Liang (2009) had UCSintact several tests (see Section 5 below for further discussion) and
greater than 15MPa; we have back-analysed two available tests parametric studies carried out.
with UCSintact values 3.26MPa and 5.75MPa to assess how this • Do not use this model if there are limited joints along the
model performs in the UCSintact <15MPa range (see Figures 6 pile length and the likely failure conditions are dependent
& 7 below). on joint orientation.
• Do not use this model in rocks with strong anisotropy.
• Do not use this model if UCSintact is <2MPa.
• Do not use if UCS intact is between 2MPa and 6MPa and
GSI is < 30. Use this model with caution if GSI is >30 in
this UCS intact range.
• Assess GSI on “pile scale” (i.e. compare graphical
descriptions provided in the GSI charts relative to the pile
dimensions). See Section 4 for further discussion.
• Where possible, use Emass as opposed to Eintact (particularly
for low GSI materials) as an input.
• Caution is advised in using this model where mi >15 and
GSI > 65, where brittle failure mechanism may govern
(i.e. the results may be conservative).
6 INTERIM RECOMMENDATIONS
Figure 9. Back Analysis of Seddon 1993 Load Test
Given that laterally loaded rock socket design is still an
evolving field and available design options are limited,
interim recommendations discussed elsewhere in this paper
are graphically summarized in Figure 12 and the
corresponding footnotes.
Proceedings of the 14th Australia and New Zealand Conference on Geomechanics, Cairns 2023 (ANZ2023)
Nyman 1978 Reinforced concrete pile 0-3.51m: Free Length • Weak rock model provides a close match as
Pile length: 15.2m Below 3.51m: UCSintact = 3.45MPa, RQD =0, expected when EI values were reduced as adopted
by Reese 1997
Erm = 7,240MPa
Diameter = 1.2m • Due to lack of information on pile
Initial EI = 3.73 x 106 kN/m2 modulus/reinforcement, non-linear EI values were
not computed independently (i.e. selected to match
A non-linear EI relationship is with the deflections, therefore the “close agreement’
adopted is not surprising.
• Not analysed with massive rock model as GSI is not
meaningful for a Vuggy Limestone. Plot not
included due to space limitations.
Seddon et.al. Reinforced Concrete Pile 0-0.05m: Free Length • Weak rock model provides a much stiffer response
Pile length: 0.64m Below 0.05m: UCS = 2MPa, RQD =0, compared to field test data
1993 • Stiff clay model provides a softer response;
Pile diameter: 0.3 Erm = 260 MPa
however, a better match could be obtained using
EI: 1.17 x 104 kN/m2 appropriate e50 values.
• Pile lateral capacity is underestimated with a stiff
Constant EI used as the
clay model
geotechnical capacity is much
• Not analysed with massive rock model as the UCS
smaller than the structural
value was too low
capacity
Frantzen Shale Steel hollow section with 0-0.3: Free Length • Weak rock model provides a stiffer response
1987
concrete infill 0.3-0.9: Stiff Clay (Cu=350kPa) • Massive rock model provides a softer response
Pile length: 4.07m compared to field data; however, results are within
0.9-2.1: UCSintact =3.26MPa, GSI=54, mi=6, an acceptable range for serviceability loadings.
Diameter = 0.22m Erm=123MPa, RQD=69, Krm=0.0005
Wall thickness: 9mm Below 2.1m: UCSintact =6.5 MPa, GSI=54, mi=6,
EConcrete: 25 GPa / Esteel: 200GPa Erm=123MPa, RQD=65, Krm=0.0005
Frantzen Steel hollow section with 0-0.3: Free Length • Weak rock model provides a stiffer response
Sandstone 1987
concrete infill 0.3-0.6: Stiff Clay (Cu=350kPa) • Massive rock model provides a softer response
Pile length: 4.57m compared to field data; however, results are within
Below 0.6m: UCSintact =5.75 MPa, GSI=50, an acceptable range for serviceability loadings.
Diameter = 0.22m mi=17, Erm=170MPa, RQD=45, Krm=0.0005
Wall thickness: 9mm
EConcrete: 25 GPa
Esteel: 200 GPa
Gabr et.al. 2002 Reinforced Concrete Pile with 0-0.3: Free Length • Weak rock model provides a stiffer response
I40-Short steel casing 0.3-2.1: UCSintact=11.3MPa, GSI=87, mi=9, • Massive rock model provides a reasonable match.
Pile embedded length: 3.356m Erm=161MPa, RQD=100
Diameter = 0.762m 2.1-3.1: UCSintact=12.2MPa, GSI=74, mi=9,
Wall thickness: 12.7mm Erm=145MPa, RQD=89
Note 1 - Rare in Practice. The massive rock model (with caution) is preferred to the Vuggy limestone model or the weak rock model. Rock failure may
be brittle. Pile capacity may be underestimated, and deflection overestimated with the massive rock model.
Note 2 – Select an appropriate e50 value consistent with material strength. Weak Rock model could also be used in UCS 0.6 MPa to 2MPa range.
However, caution is advised if other models are used for better quality rocks as there could be consistency issues.
Note 3 - Use an elasto-plastic method ( Oasys ALP, DEFPIG, Plaxis 3D) that uses UCSmass and Emass as inputs to verify that results are not overly
conservative. Where appropriate, a granular soil model (e.g. Reese Sand model) could be used with an appropriately high friction angle.
Note 4 – Rock strength descriptions as per Australian Standard AS 1726: 2017
7 CONCLUSIONS
While P-y curves make the assessment of laterally loaded rock associated risks.
sockets much easier, these models inevitably have limitations.
P-y curves generated with a given pile type, rock strength and
secondary structures, may not be representative at other 8 REFERENCES
locations where these conditions are different. Unfortunately, Brown, E.T. SHIRMS 2008. Estimating the mechanical properties of
the pile test database is not large enough to select appropriate rock mass. Proceedings of the First Southern Hemisphere
p-y curves for a variety of ground conditions. International Rock Mechanics Symposium, Australian Centre for
This paper provides interim recommendations on adapting/ Geomechanics, Perth, pp. 3-22.
improvising existing P-y models over a variety of ground Carter, T.G. and Marinos, V. 2014 Use of GSI for rock engineering
conditions. design. Proceeding’s 1st International Conference on Applied
Empirical Design Methods in Mining, Lima, pp. 1–19.
Whilst, factors such as pile ultimate lateral capacity, pile
Carter, TG, Diederichs, MS & Carvalho, JL 2008, ‘Application of
sectional forces, etc. also matter, the recommendations are modified Hoek–Brown transition relationships for assessing
primarily based on getting the serviceability deflections right strength and post-yield behaviour at both ends of the rock
due to the following reasons: competence scale’, Journal of the Southern African Institute of
Mining and Metallurgy, vol. 108, no. 6, pp. 325–338.
• For laterally loaded sockets, pile serviceability deflections
and pile structural strength often govern the design rather Cho, K.H., Gabr, M.A., Clark,S. and Borden, R.H. 2007. Field P-y
than ultimate geotechnical capacity. curves in weathered rock. Canadian Geotechnical Journal 44:
753-764(2007).
• Variations in bending moment / shear force predictions
are relatively small (i.e. in terms of percentage change) Chong, W. L., Haque, A., Ranjith, P.G. and Shahinuzamman, A. 2011.
A parametric study of lateral load behavior of single piles socketed
compared to predicted deflections. into joint rock mass. Australian Geomechanics Journal, Volume
It is acknowledged that when it comes to improvising limited 46.
data, no approach will be universally acceptable and that Chong, W. L., Haque, A., Ranjith, P.G. and Shahinuzamman, A. 2011.
opinions may differ. However, the authors have seen Effect of joints on P-y behaviour of laterally loaded piles socketed
widespread inappropriate use of the P-y rock models and into mudstone. International Journal of Rock Mechanics &
consider that some guidance is useful in alleviating the Mining Sciences, 48(2011) 372-379.
Proceedings of the 14th Australia and New Zealand Conference on Geomechanics, Cairns 2023 (ANZ2023)
Frantzen, J., and Stratton, F.W. 1987. P-y curve data for laterally
loaded piles in shale and sandstone. Kansas Department of
Transport, Report no. FHWA-KS-87-2.
Gabr, M.A., R.H. Borden, K.H. Cho, S. Clark, and J.B. Nixon. 2002.
P-y Curves for Laterally Loaded Drilled Shafts Embedded in
Weathered Rock. North Carolina Department of Transportation.
FHWA/NC/2002-008. 289p.
Hoek, E., and Brown, E.T. 2018. The Hoek-Brown failure criterion and
GSI -2018 edition. Journal of Rock Mechanics and Geotechnical
Engineering. (2018) 1-19.
Hoek E. and Diederichs, M.S. 2006. Empirical estimation of rock mass
modulus. International Journal of Rock Mechanics and Mining
Sciences 2006;43(2):203-15.
Hoek, E., Carter, T.G. and Diederichs, M.S. 2013. Quantification of
the geological strength index chart. 47th US rock
mechanics/geomechanics symposium. San Francisco, ARMA 13-
672.
Liang, R., Yang, K. and Nusairat, J. 2009. P-y criterion for rock mass.
Journal of geotechnical and geoenvironmental engineering,
ASCE, January 2009.
Isenhower, W.M., Wang, S.T., and Vasquez, L.G. LPILE Technical
Manual, 2016, Ensoft Inc.
Marinos, V., Marinos, P. and Hoek E. 2005. The geological strength
index: applications and limitations. Bulletin of Engineering
Geology and the Environment 2005;64(1):55-65.
Parsons, R.L., I. Willems, M.C. Pierson, and J. Han. 2010. Lateral
Capacity of Rock Sockets in Limestone under Cyclic and
Repeated Loading. Kansas Department of Transportation. 86p.
Reese, L.C. (1997). “Analysis of Laterally Loaded Piles in Weak
Rock.” Journal of Geotechnical and geoenvironmental
Engineering. ASCE. Reston, Virginia. v123 n11. 1010-1017.
Seddon, K.D. and McDonald, P. 1993. Prediction of performance
(laterally loaded socketed piles). Australian Geomechanics
Journal. 1993:115-120
Turner, J. 2006. Rock socketed shafts for highway structure
foundations. NCHRP synthesis 360.
Zhang, L. and Einstein, H. 2000. Nonlinear analysis of laterally loaded
rock-socketed shafts. Journal of Geotechnical and
geoenvironmental Engineering. ASCE. Vol 126. No. 11.
November 2000.