Revisiting Methods for the Design of Rock Socketed Piles

Ian W. Johnston, Ph.D., CPEng 1

Abstract: The three principal methods specifically developed for the design of rock socketed piles were originally presented in the 1980s. One
of the methods is an empirical technique based on the normalized performance of many full-scale pile tests while the other two are based on
numerical and analytical techniques. However, there appears to be no detailed assessment of how well each method can predict the performance
of piles socketed into rock. This paper discusses the three methods and then compares their predictions for load-settlement with the results of
several foundation load tests particularly referencing settlements that might be acceptable for serviceability. It is demonstrated that while all
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

three methods can produce reasonable predictions of performance, one of them tends to overpredict performance while the other two appear to
underpredict performance. Only one of the methods produces load-settlement curves that bear a reasonable similarity to the expected shape of
actual load-settlement curves. DOI: 10.1061/(ASCE)GT.1943-5606.0002414. © 2020 American Society of Civil Engineers.

Introduction with a socket formed in the rock. It is often necessary to use casing
to maintain stability in the overburden. The rock socket is drilled to
Around 40 years ago, there was considerable activity involving rock its design depth below the casing before installing reinforcement
socketed piles, with a number of groups around the world devoting and then placing the concrete. Key elements in this process are
much effort to understanding their mechanisms of behavior, devel- making sure the base of the socket is clean if the pile is to develop
oping methods of analysis and design, and exploring techniques for a base resistance component, and that the side of the socket is clean
effective and economic construction. In addition to rapidly develop- and adequately rough to provide the required side resistance. Fig. 1
ing computational techniques allowing more detailed assessment of illustrates the main features of a rock socketed pile. It should be
performance (e.g., Pells and Turner 1979; Donald et al. 1980; Rowe noted that, although not considered in this paper, there will be a
and Pells 1980; Rowe and Armitage 1987b), a key element of this component of side resistance in the overburden and elastic com-
early work was the detail and extent of laboratory and field testing pression of the shaft where the pile passes through the overburden.
undertaken to support development (e.g., Johnston et al. 1980; As has been discussed extensively elsewhere (e.g., Johnston
Williams and Pells 1981; Glos and Briggs 1983; Horvath et al. 1977; Williams 1980; Lam and Johnston 1982; Johnston et al. 1987),
1983; Johnston et al. 1987). when a vertical load is applied to a socketed pile, there is an initial
Although the various elements of socket behavior (particularly re- elastic settlement of the pile with no relative movement between the
lating to side wall roughness) have been a fertile source of research pile and the rock. As the load is increased, the side resistance devel-
investigations over the years since, there has been little significant ops further under a constant normal stiffness condition, with vertical
development in methods for design. The design methods for complete sliding of the concrete pile over the roughness of the socket causing
piles (piles with both side and base resistance contributing signifi- dilation of the socket with a significant increase in resistance. With
cantly to total resistance) proposed during the 1980s by Williams et al. more load, shearing through the asperities takes over as ultimate side
(1980), Rowe and Armitage (1987a), and Carter and Kulhawy (1988) resistance is progressively approached (Fig. 2).
remain the principal means of designing rock socketed piles with re- Base resistance is essentially a bearing capacity issue wherein a
spect to both load and settlement. It is recognized that other means of number of factors can influence ultimate capacity (Williams 1980;
socket design involving general and specialist computational methods Choi 1984; Zhang and Einstein 1998; Haberfield 2013). Of particu-
have been developed since the 1980s but, as will be discussed later, lar importance is the strength of the rock and the influence that dis-
they require specialist knowledge that limits their application. continuities have on this strength. However, as has been discussed by
The author has revisited the above three methods for the design many over the years (e.g., Williams et al. 1980; Horvath et al. 1980;
of rock socketed piles with an aim of investigating their effective- Rowe and Armitage 1987a, b; Haberfield and Lochaden 2019), base
ness, plausibility, and reliability through a comparison with the re- resistance develops at a much slower rate than side resistance and, for
sults of a selection of full-scale foundation load tests. an embedment of at least two diameters, even at large displacements
it rarely displays an ultimate load that is constant but increases with
increased displacement.
Behavior of a Rock Socketed Pile

A rock socketed pile is generally a high-capacity bored pile that is

formed by drilling through any overburden down to a rock surface,
Overview of Design Methods

1
Professorial Fellow, Dept. of Infrastructure Engineering, Univ. of Williams, Johnston, and Donald Method
Melbourne, Parkville, Melbourne, VIC 3010, Australia. ORCID: https://
The Williams, Johnston, and Donald (WJD) method was developed
orcid.org/0000-0002-0301-4673. Email: [email protected]
Note. This manuscript was submitted on December 6, 2019; approved from the results of about 50 field tests conducted principally on
on July 27, 2020; published online on October 14, 2020. Discussion period piles socketed into the Silurian mudstone of Melbourne. Full details
open until March 14, 2021; separate discussions must be submitted for in- of the tests may be found in Williams (1980); those of the design
dividual papers. This paper is part of the Journal of Geotechnical and method, in Williams et al. (1980). The method makes use of the
Geoenvironmental Engineering, © ASCE, ISSN 1090-0241. principles of normalization wherein relationships are expressed

© ASCE 04020144-1 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

 Fig. 1. Rock socketed pile.
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Side stress reduction factor. (Copyright 1980, “The design of

socketed piles in weak rock,” A. F. Williams, I. W. Johnston, and I. B.
Donald, reprinted with permission from Taylor and Francis.)

Fig. 2. Typical load-settlement curves for a socketed pile.

in dimensionless terms so that design data become independent of

dimensions. For processes that follow the same basic mechanisms,
this considerably reduces the scatter of data and allows the use of
much simplified design curves. A summary of this design method
is provided in the following subsections.
Fig. 4. Principles for normalizing side resistance. (Copyright 1980,
Assessment of Side Resistance “The design of socketed piles in weak rock,” A. F. Williams, I. W.
A key relationship providing an estimate of failure or ultimate side Johnston, and I. B. Donald, reprinted with permission from Taylor
stress, f su , is and Francis.)
f su ¼ αqa ð1Þ

where α = side stress reduction factor; and qa = unconfined com-

fs f f sp
pressive strength of the rock. The variation of α with qa from a ¼ se − ð3Þ
number of pile-loading tests, as available at the time, is shown f su fsu f su
in Fig. 3.
The principles for normalizing side stress are illustrated in Fig. 4. This process was applied to all Williams’ side resistance tests to
For any settlement, ρ, the side stress on the pile is given by fs, produce the normalized design curve shown in Fig. 5.
which is
Assessment of Base Resistance
fs ¼ fse − f sp ð2Þ The principles for normalizing base resistance are similar to those
used for side resistance and are illustrated in Fig. 6. For any settle-
where f se = elastic side stress; and f sp = plastic reduction in side ment ratio, ρ=D, where D = diameter of the socket, the base stress,
stress as defined in Fig. 4. fb , is defined by
Eq. (2) can be normalized by dividing by the measured ultimate
side stress, f su , to give fb ¼ fbe − fbp ð4Þ

© ASCE 04020144-2 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

 Table 1. Variation of N s as a function of L=D
L=Da Ns
0 0.0065
1 0.0109
3 0.0147
5 0.0169
10 0.0185
15 0.0196
a
N s for intermediate values of L=D can be obtained by linear interpolation.
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Normalized design curve for side resistance. (Copyright 1980,

“The design of socketed piles in weak rock,” A. F. Williams, I. W.
Johnston, and I. B. Donald, reprinted with permission from Taylor
and Francis.)

Fig. 7. Normalized design curve for base resistance. (Copyright

1980, “The design of socketed piles in weak rock,” A. F. Williams,
I. W. Johnston, and I. B. Donald, reprinted with permission from
Taylor and Francis.)

Therefore, with the relevant L=D ratio, N s can be selected and

fb1 estimated from Eq. (5) to give the normalized base resistance as
fb f f bp
¼ be − ð6Þ
fb1 fb1 fb1

This equation was applied to all Williams’ base resistance tests

to produce the normalized design curve shown in Fig. 7.

Design Procedure for a Complete Pile

Fig. 6. Principles for normalizing base resistance. (Copyright 1980, The procedure starts by assuming that all the design load is taken
“The design of socketed piles in weak rock,” A. F. Williams, I. W. by the pile side to give an initial trial socket length of
Johnston, and I. B. Donald, reprinted with permission from Taylor
Qd
and Francis.) L¼ ð7Þ
fsu πD
where Qd = design load. As argued by Williams et al. (1980) and
where fbe = elastic base stress; and fbp = plastic reduction in base supported by others such as Rowe and Armitage (1987a), Carter
stress as defined in Fig. 6. and Kulhawy (1988), and, more recently, Haberfield and Lochaden
Unlike the side resistance tests, few of the base resistance tests (2019), for piles in reasonably competent rock utilizing side and
reached failure and an alternative to ultimate base resistance was base resistance, it is settlement that usually controls the design
required in order to normalize Eq. (4). After consideration of of rock socketed piles. Therefore, the maximum allowable settle-
the effects of embedment on bearing capacity, it was decided to use ment of the pile, ρmax , needs to be decided. Then, using a factor of
the estimated base stress at a settlement ratio (ρ=D) of 1% or fb1. safety for settlement, Fρ , the total load, Qe , required to develop this
Using the results of the many base resistance tests, it was found that factored settlement with the pile responding elastically (referred to
f b1 could be reasonably estimated as as the elastic load) is given by

f b1 ¼ N s Em ð5Þ ρmax Em D
Qe ¼ ð8Þ
Fρ I ρ
where N s is a settlement-based bearing capacity factor (Table 1);
and Em = elastic modulus of the rock mass. A full explanation where the settlement influence factor I ρ is given in Fig. 8 (Donald
of the derivation of N s can be found in Williams et al. (1980). et al. 1980). Then, using Fig. 9 (Donald et al. 1980), the distribution

© ASCE 04020144-3 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

 resistance is ignored) or piles that are base-only (where side resis-
tance is ignored).

Rowe and Armitage Method

The Rowe and Armitage (RA) method uses a series of numerical
solutions to account for nonslip and slip between the pile shaft and
the surrounding rock. The method involves selection of input data
(the same as in the WJD method), these being ρmax , D, Qd , Ec , qa ,
and Fρ . On the basis of a large number of field tests for which f su
and Em were correlated with qa (Rowe and Armitage 1987a), the
parameters fsu and Em are then estimated from the relationships as
pffiffiffiffiffi
f su ¼ 0.45 qa ð9Þ
pffiffiffiffiffi
Fig. 8. Elastic settlement influence factor. (Copyright 1980, “Theoretic Em ¼ 215 qa ð10Þ
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

analyses of rock socketed piles,” I. B. Donald, H. K. Chiu, and S. W.

Sloan, reprinted with permission from Taylor and Francis.) Different values of fsu and Em can be selected if there is reason
to do so. Then, with the design settlement set at ρmax =2 and partial
factors (0.7 is suggested) applied to fsu and Em , the relevant nu-
merically derived curves are used to determine the required length
of the socket, L. Although the original paper only considered an
application in which a design length can be estimated, it is possible
to determine a load-settlement curve based on the RA method for a
selected geometry ðL; DÞ and material properties (Ec , Em , qa ).
The manual process adopted by the author to construct a
load-settlement curve involved selection of the above input param-
eters and, for an appropriate unfactored settlement, estimation
of a design load. Then, with this settlement and design load, the
RA method parameters I d (a settlement influence factor) and
ðL=DÞmax (maximum L=D ratio if the entire load is taken on the
side) are calculated. With these two parameters, the appropriate de-
sign chart corresponding to the ratio of pile to rock modulus was
selected (or several design charts where extrapolation was required)
and a design length to diameter, ðL=DÞd estimated. This value was
compared with the originally selected L=D and, if not within close
Fig. 9. Elastic load distribution. (Copyright 1980, “Theoretical ana- proximity, the process was repeated with a new input L selected.
lyses of rock socketed piles,” I. B. Donald, H. K. Chiu, and S. W. Sloan, This iteration was repeated until the estimated L=D was in close
reprinted with permission from Taylor and Francis.) agreement with the selected L=D. A number of other settlements
were selected and the process repeated until the required range for
the load-settlement curve was obtained. Further details of the
method are readily available in Rowe and Armitage (1987a, b).
of this total elastic load into the elastic side load, Qse , and the elas- One of the problems encountered with the RA method was that
tic base load, Qbe , can be established. In these figures, Ec is the it could not be used where certain combinations of input parameters
elastic modulus of the pile. fell outside the range of design curves.
The method now requires these elastic loads to be relaxed by the
respective plastic reduction factors using Eqs. (3) and (6), the nor-
malizing factor, fsu , from Fig. 3, N s from Table 1, and fb1 from Carter and Kulhawy Method
Eq. (5). This then leads to the relaxed loads at the side and base of The Carter and Kulhawy (CK) method is based on analytical tech-
the socket and thus to the total relaxed load for the initial trial niques and empirical data that allow for elastic and nonelastic
socket length. If this load is less than the design load (which it usu- behavior of a socketed pile. Using the input data with several equa-
ally is), the length of the socket is increased and the procedure re- tions describing elastic, side slip, and side shearing mechanisms,
peated until the total relaxed load equals the design load. The final, the load-settlement response can be developed.
and very important, step is checking that the pile design satisfies The primary variables in the general CK method are shown in
geotechnical strength requirements. Fig. 10. For a complete pile, the response shown in Fig. 10(a) com-
An example of the application of the WJD design method is prises two linear parts. The first part, with a slope S1c , describes
given in the Appendix. elastic behavior with no relative movement between the concrete
By digitizing the various design charts, a spreadsheet has been of the socket and the surrounding rock. The second part, with a
developed by the author to design a socketed pile with the end slope S2c and an intercept on the load axis of I c , describes move-
result being a socket length and a load-settlement curve for the ment with full slip along the side of the socket with the base load
selected geometry ðL; DÞ and material properties (Ec , Em , qa ). An- continuing to develop. The curved region between these two linear
other spreadsheet produces a load-settlement curve directly for se- portions is where sliding starts and gives way to shearing as shown
lected pile geometry and material properties. in Fig. 2. Also, for a complete pile the method simply gives the
The WJD method can be used for the design, or development, slope of the full slip movement section with no limitation on
of load-settlement curves for piles that are side-only (where base the amount of settlement that occurs. Clearly, the load sustainable

© ASCE 04020144-4 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

 Fig. 10. Definition of primary variables for the CK method: (a) complete pile; (b) side-only pile; and (c) base-only pile.
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

Table 2. Variables used to predict load-settlement response of rigid socketed piles by CK method
Variable Complete pile Side resistance–only pile Base resistance–only pile
 D  πL D  πEm L  D 
Slope of elastic behavior
S1c ¼ 2
Em þ E S1s ¼ S3b ¼ Em
1−ν ζD 1þν m ð1 þ νÞζ 1 − ν2

πEm D πEm L
Slope of slip behavior S2c ¼ S2s ¼ —
2R4 R1
πEm D2 πEm LD
Intercept of slope of slip behavior Ic ¼ R5 Is ¼ R2 —
slope with load axis 4R4 2R1
Ultimate side resistance — Qu ¼ f su πDL —

by the socket will not, in reality, increase indefinitely but will Although an important variable when assessing the performance
gradually decrease with increased settlement. This behavior is of side only piles by the CK method, the ultimate side stress, f su , is
not modeled by the CK method. not required for complete piles in this method. As complete piles
For a side resistance–only pile, the response is modeled by the are being considered in this paper, further consideration of fsu in
three linear legs shown in Fig. 10(b). The first leg, with slope S1s , the CK method will not be undertaken.
describes the elastic behavior as above. The intermediate leg, with
slope S2s (and intercept on the load axis of I s ), describes movement Other Methods
with full slip along the side of the socket; once the ultimate load,
Qu ð¼ fsu πDLÞ, is generated, the pile settlement will continue to As indicated earlier, there are other means of socket design using
increase but with no increase in load. general and specialist computational methods. One approach
For a base resistance–only pile, the response comprises one lin-
ear part describing the elastic development of the base resistance. Table 3. Factors for use with the variables in Table 2
The CK method does not describe how this response must be lim-
Factors Evaluation of factor
ited by progressive yielding at the base. !
There can be some difficulty with the CK method when evalu- R1 –R5 1
R1 ¼ ð1 þ νÞ ζ þ
ating input parameters for the general solution to a compressible 2 tan φ tan ψ
pile relative to the surrounding rock. However, this can be over- 1þν c
come by considering a rigid pile defined by R2 ¼
tan φ tan ψ Em
π
  R3 ¼ ð1 − ν 2 Þ
Ec D 2 2
≥1 ð11Þ R3 R1
Em 2L R4 ¼
R3 ð2L
D Þ þ R1
R2 R3 ð2L
DÞ
For values of Ec , Em , L, and D typically encountered with rock R5 ¼
R3 ð2L
D Þ þ R1
socketed piles, Eq. (11) often holds, allowing the simpler relation-  q 2
ships for rigid piles to be used. Other tan φ tan ψ ¼ 0.001 a 3
pa
A summary of the CK method’s key variables for the develop-  q 2
a 3
ment of a load-settlement curve for a rigid rock socketed pile is c ¼ 0.1pa
pa
provided in Table 2. Table 3 provides factors for use with the var- 
iables in Table 2. A summary of the CK method can be found in L
ζ ¼ ln 5ð1 − νÞ
Kulhawy and Carter (1992). D

© ASCE 04020144-5 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

 involves the program ROCKET, which is based on the work of it was decided to adopt the relationship proposed by Rowe and
Seidel (1993) and requires input quantifying the roughness of Armitage (1987a) for rock masses for which there were no open
the interface between the side of the socket and the rock mass. Un- discontinuities. This relationship is given by Eq. (10).
fortunately, the considerable specialist knowledge and expertise re- The ultimate side shear stress, f su , used in application of the
quired to compile this data, as well as the difficulty of obtaining WJD and RA methods is also given in Table 4. The reason for this,
data on roughness at some considerable depth in a drilled shaft, as will be discussed in more detail, is that the value of fsu can be a
limits the applicability of this approach. There are several other design choice that may influence predictions for complete pile
methods involving the quantification of roughness, all of which sockets by these two methods.
have similar limitations. Table 4 also presents the value of the relationship used to assess
In many instances, multipurpose software packages such as the rigidity of the rock socketed pile as defined in the CK method
PLAXIS 3D (version 20) and FLAC 3D (version 7.0) have been by Eq. (11). Fig. 11 presents the load-settlement results of 16
successfully applied to socketed pile design. However, these have admissible foundation load tests compared with predictions made
not been specifically designed for socketed piles and require con- by the three design methods.
siderable expertise and experience with significant calibration
against known socket behavior. Therefore, while these methods
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

have undoubted potential in this area, their application will not Discussion of the Measured and Predicted
be considered here. Load-Settlement Curves

A number of observations can be made relating to the data shown in

Comparison of Design Method Predictions with Fig. 11. With respect to the measured load-settlement curves,
Results from Tests on Complete Piles although some tests did not get far beyond the elastic phase, the
characteristic behavior described earlier and illustrated in Fig. 2
This section presents a selection of load-settlement results from was generally provided. Only one of the piles showed plunging
foundation load tests on complete piles and, using pile geometry behavior (Goeke and Hustad 1979) and the rock in this test was
ðL; DÞ, pile properties (Ec ) and properties of the rock socketed the weakest of the series, with qa ¼ 0.43 MPa. To the limits of
is considered, forming the socket ðEm ; qa Þ, assesses how well the the settlements recorded, all other piles continued to display an in-
three socketed pile design methods (WJD, RA, and CK) predict the crease in load with increased displacement.
results. In all these tests, only the settlement of the rock socket is With regard to the load-settlement curves determined using the
considered, with no allowance for compression of the pile shaft that RA method, it appears that where predictions were possible and
may exist above the socket. As noted earlier, for some of the tests, it beyond the early stages of loading, the predicted load-settlement
was not possible to apply the RA method because the RA design curves were generally above the measured curves. The predicted
charts did not extend to the full range of test parameters. curves of the WJD and CK methods generally fell below the
While the results of many complete pile socket tests are reported RA predictions and, in broad terms and with the odd exception,
in the literature, a significant proportion could not be used because seem to better capture the behavior of the piles for settlements
of inadequate or unreliable data relating to rock properties and oc- up to about 10 mm. For greater settlements, the CK method can
casionally socket dimensions, questionable data particularly as re- both underpredict and overpredict settlements while the WJD
late to the influence of materials overlying the rock socket, variable method seems to mostly underpredict. What stands out is that
materials forming the socket with consequent unknown variability the CK method, because it predicts a constant slope of the load-
of properties, significant anomalies in the reported load-settlement settlement response once slip occurs, shows no gradual reduction
curves, and the use of materials better described as soils in which the in the rate of load development as settlement increases. The WJD
socket was formed and therefore probably not displaying the socket method, on the other hand, appears to capture the curved load-
dilation characteristics typical of rocks. There were also a number of settlement characteristics quite well, with a gradual reduction in
small-scale model tests in synthetic materials (such as plaster) re- the increase in load with settlement.
ported for complete piles, but these were rejected because of the in- As discussed earlier, for piles in reasonably competent rock uti-
fluence of the generally nondilatant characteristics of these materials lizing side and base resistance, settlement usually controls design.
as is discussed by Johnston and Choi (1986). Interestingly, many of Therefore, it is critical that certain design settlement limits be set.
the tests with usable data were over 20 years old, with very few use- As socketed piles are generally of high capacity to support major
ful data sets available since 2000. The reason for this may be the construction, these limits are often quite tight. It is not uncommon
increasing use of O-Cell tests (Osterberg 1998), in which the side to find that maximum allowable settlements of the socket, ρmax ,
resistance and base resistance of a socketed pile are assessed sepa- need to be limited to about 10–20 mm, corresponding to design
rately even though the base resistance usually involves some side settlements of, respectively, 5–10 mm when applying a factor of
resistance component between the hydraulic jacks and the actual safety for a settlement, Fρ , of 2. Because these maximum allowable
base. Where reports of O-Cell tests included dimensions of the base settlements do not include compression of the pile shaft in any
section as well as where length was at least half of diameter, provided overburden above the socket, it follows that the ability of the three
other details were acceptable, these results have been included. design methods to predict the loads that would lead to these settle-
Table 4 summarizes the tests on complete pile sockets (and ments is of considerable interest. To assess this, the measured loads
acceptable O-Cell tests) used in this investigation. Usually, in ad- for each of the tests at 5 and 10 mm settlements are plotted against
dition to the criteria given above, the test data had to include the the corresponding predicted loads for each of the three methods and
socket dimensions, ðL; DÞ and reliable estimates of qa and Em . are shown in Fig. 12. Linear trend lines were fitted for each data set.
Regrettably, there were several reports of test results that met all Table 5 presents some statistical parameters related to the data pre-
criteria except for data about Em . In order to expand the available sented in the figure.
database, these tests were included with a reasonable estimate of For settlements of 5 and 10 mm, the data presented in Fig. 12
Em based on correlations between qa and Em . Several correlations and Table 5 confirm the observations made earlier. Specifically, the
were considered but, based on an extensive survey of relevant data, mean bias (the mean of the ratios of the measured load to the

© ASCE 04020144-6 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

 © ASCE

Table 4. Details of complete pile tests used in comparisons

Elastic moduli
Socket dimensions
(m) (GPa) (MPa) Ultimate side stress (fsu ) (MPa)
Diameter Length Pile Rock mass Rock UCS WJD RA Rigidity criteriona
pffiffiffiffiffi Ec D 2
Reference Pile ID (D) (L) (Ec ) (Em ) (qa ) (MPa) Rock type method ¼ αqa method ¼ 0.45 qa Em ð2LÞ ≥ 1
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

Williams (1980) M8 0.66 1.8 35 500 2.3 Mudstone 0.56 0.68 2.4
Williams (1980) M9 0.66 4.2 35 550 2.45 Mudstone 0.57 0.70 0.39
Horvath (1980) P2 0.71 1.37 37 419 11.1 Shale 0.94 1.50 5.9
Horvath et al. (1980) Airport Rd No 3 0.635 1.01 35 1,000 15.2 Shale 1.04 1.75 3.6
Mallard and Ballantyne (1976) “Solid toe” 1.146 8.5 35 128 0.8 Chalk 0.40 0.4 1.2
Wong and Oliveira (2012)b SC-02 0.75 0.4 43 2,000 30 Sandstone 1.3 2.46 18.9
Day et al. (2009)b TP1 1.5 2.50 43 645 18 Siltstone 1.10 1.90 6.0
Radhakrishnan and Leung (1989) TP1 0.81 12.4 30 517 6.0 Siltstone 0.77 1.10 0.06
Kwon et al. (2005)b P1 1.5 1.5 35 151 79 Breccia 1.78 4.00 57.9
Zhan and Yin (2000) VT2 1.05 2.0 41 760 30 Volcanic 1.30 2.46 3.7
Aurora and Reece (1977) MT3 0.75 1.52 30 256c 1.42 Clay shale 0.48 0.54 7.1
Akguner and Kirkit (2012) TP3 0.8 11 30 319c 2.2 Schist 0.55 0.67 0.12
Leung (1996) Siltstone 1.4 3.0 35 548c 6.5 Siltstone 0.79 1.15 3.5
Glos and Briggs (1983) East pile 0.61 1.4 35 654c 9.26 Sandstone 0.89 1.37 2.5
Walter et al. (1997)b — 0.91 2.6 35 685c 10.15 Sandstone 0.91 1.43 1.6
Goeke and Hustad (1979) Shaft No 1 0.76 5.49 35 141c 0.43 Shale 0.32 0.29 1.2
Note: UCS = unconfined compressive strength.
a
04020144-7

Sockets satisfying this criterion can be considered rigid (Carter and Kulhawy 1988).
b
O-cell test.
pffiffiffiffiffi
c
Estimated from Em ¼ 215 qa .
J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

 predicted load for each test) for all three methods is similar the mean bias (CVB) indicate that variability about the mean bias
(relative to 1), with the RA method overpredicting and the WJD or trend line is the least for the RA method and the greatest for the
and CK methods underpredicting. The standard deviation CK method. The root-mean squared error (RMSE) and mean ab-
about the mean bias (SDB) and the coefficient of variation of solute error (MAE) indicate that all methods produce predictions
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Load-settlement data for selected foundation load tests compared with predictions.

© ASCE 04020144-8 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. (Continued.)

of load relative to the measured values for the 5-mm settlement It may be worth observing that, as shown in Table 4, the condition
that are similar. However, for the 10-mm settlement, the CK for rigidity for 13 of the 16 tests was satisfied, thereby justifying the
method appears to produce predictions closer to the measured use of the simpler solutions developed with the CK method. Three
values. tests (Williams 1980, for Pile M9; Radhakrishnan and Leung 1989;

© ASCE 04020144-9 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 12. Comparison of measured loads with predicted loads: (a) 5-mm settlements; and (b) 10-mm settlements.

Table 5. Statistical parameters developed from Fig. 12

For loads at 5-mm settlement For loads at 10-mm settlement
Statistical parameter WJD RA CK WJD RA CK
Mean biasa 1.29 0.85 1.57 1.31 0.72 1.30
Standard deviation of bias (SDB)b 0.40 0.23 0.87 0.40 0.15 0.51
Coefficient of variation of bias (CVB)c 0.31 0.27 0.55 0.30 0.20 0.40
Root mean squared error (RMSE)d (MN) 2.17 1.94 2.44 3.81 4.79 2.36
Mean absolute error (MAE)e (MN) 1.62 1.46 1.95 2.87 3.74 1.87
a
Mean of individual measured loads to predicted loads: <1 is overpredicted; and >1 is underpredicted.
b
Indicates standard deviation about the mean bias or trend line of Fig. 12.
c
Indicates variability relative to the mean bias or trend line in Fig. 12.
d
Indicating the variability relative to the line of equality in Fig. 12 (units in MN).
e
Alternative to RMSE (units in MN).

Akguner and Kirkit 2012), however, returned rigidity results below relevant geotechnical parameters and their variability so that
the limit of unity. Despite this, the results shown in Fig. 12 for the the maximum allowable settlements are not exceeded.
latter two tests are in the same range as those for the other tests, with
the Williams M9 tests a little more underpredicted. Note that the tests
reported by Williams (1980) for Pile M9 and Radhakrishnan and Influence of Side Resistance
Leung (1989) did not experience enough settlement to be included
for comparison of the 10-mm test settlement. The ultimate side stress, f su , for the WJD and RA methods is
Overall, and especially when considering the variability of pile determined by Eq. (1) (with α given in Fig. 3) and Eq. (9) respec-
load-settlement curves, the predictions for all three methods for set- tively. The ultimate side stress for the CK method is required
tlements up to about 10 mm seem to be reasonable, with perhaps not for complete piles but rather for side resistance–only piles;
the slight conservatism of the WJD and the CK methods being pre- this method derives a lower-bound estimate for fsu from the
relationship
ferred. Further, the more familiar reducing curve for increased loads
produced by the WJD method rather than the apparently limitless rffiffiffiffiffiffi
fsu qa
constant rise predicted by the CK method may provide a more suit- ¼C ð12Þ
p pa
able alternative for larger permitted settlements.
It may be worth commenting at this point that the original WJD where C ¼ 0.63; and pa = atmospheric pressure (0.1013 MPa).
and RA design methods were based on the selection of a factored There have been many studies involving the prediction of
settlement at the start of the design process and then establishing a fsu (e.g., Horvath and Kenney 1979; Williams et al. 1980;
pile length before checking that the allowable strength criteria were Williams and Pells 1981; Rowe and Armitage 1987a; Seidel and
not exceeded. The CK method simply permits a load-settlement Collingwood 2001; Kulhawy et al. 2005; Asem and Gardoni 2019),
curve to be generated. However, this paper shows how load- which can be influenced by many factors, including qa , Em , L, and
settlement curves using all three methods can be generated. There- D, all of which can be assessed (Asem and Gardoni 2019). There
fore, the processes defined in many national codes and specifications are, however, a number of other factors relating to the roughness of
(e.g., Standards Australia 2009; AASHTO 2017) may be followed the socket side that can have a significant influence on fsu . Mainly
so that settlements from serviceability loads can be assessed di- for reasons of access, not only are these difficult to assess for
rectly from the unfactored load-settlement curves generated by any particular socket but, despite the significant efforts of many
each method. Clearly this requires careful consideration of the researchers, their individual influences on f su are far from well

© ASCE 04020144-10 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Variation of fsu with qa used in the three design methods.
Fig. 14. Comparison of measured loads with predicted loads for a
10-mm settlement with f su in the WJD and RA methods set to values
determined by Kulhawy et al. (2005) with C ¼ 1.

understood (Haberfield and Lochaden 2019; Asem and Gardoni

2019). It follows that when considering the complex factors influ-
encing the development of side resistance, it is perhaps not surpris-
ing that f su is difficult to accurately assess and that the scatter of Table 6. Statistical parameters developed from Fig. 14
test data is significant. For loads at 10-mm settlement
Overall, there seems to be a reasonable case for suggesting that
Statistical parameter WJD RA
the simple relationship proposed by Kulhawy et al. (2005) provides
an indication of what might be an acceptable “average.” The Mean bias 1.33 0.77
Kulhawy et al. (2005) relationship is Eq. (12) with C ¼ 1 and is SDB 0.39 0.16
a dimensionless version of the relationship proposed in the RA CVB 0.29 0.21
RMSE 3.47 3.82
method. The RA method and the Kulhawy et al. (2005) relationship
MAE 2.49 3.00
are identical for C ¼ 1.42, indicating that the former is less
conservative. The relationships used in the WJD and RA methods
and proposed by Kulhawy et al. (2005) for fsu against qa are shown
in Fig. 13. The RA method produces the highest values of fsu ; the
settlement. This suggests that the f su values used in the WJD
Kulhawy et al. (2005) method (with C ¼ 1) produces intermediate
and the RA methods may have only a relatively small effect
values for qa greater than about 6 MPa but the lowest values for qa
on predictions, indicating that the internal structures of the respec-
less than about 6 MPa; the WJD method produces the lowest values
tive methods may have more significant influence. It should be
for qa more than about 6 MPa but intermediate values for qa less
noted that fsu has no effect on the CK predictions for complete
than about 6 MPa.
piles.
Although not possible for a complete pile with the CK method,
Table 6 presents the statistical parameters related to the data pre-
depending on the degree of conservatism desired, fsu for the WJD
sented in Fig. 14. This table confirms the above observations with
and RA methods can be selected to suit. If some additional inves-
relatively small changes in mean bias, SDB, and CVB. The changes
tigations were undertaken to better define f su —for example, instru-
in RMSE and MAE indicate that there has been a small movement
mented pile loading tests (including O-Cell tests) or laboratory of the WJD and RA predicted loads toward the measured loads.
testing involving constant normal stiffness (CNS) direct shear tests
(Johnston et al. 1987) with representative interface roughness—
then using the WJD and RA methods with a more representative
Conclusions
f su would seem justified.
To investigate the effect of changing the values of fsu on the Following a need to assess the performance of some piles socketed
predictions of the WJD and RA methods for the 16 test results re- into weak rock, and finding little in recently published works to
ported earlier, the Kulhawy et al. (2005) relationship was applied to help with this task, the author took the opportunity of revisiting
the WJD and RA design methods and compared with the measured three specific design methods developed in the 1980s. The WJD
load-settlement curves. As this relationship produced lower fsu method (Williams et al. 1980) is an empirical method based on
values than in the original RA method, it was expected that the over- normalized performance characteristics of a large number of piles
prediction would be reduced. Conversely, as the Kulhawy et al. tested mainly in Melbourne, Australia. The RA method (Rowe and
(2005) relationship produced higher fsu values than the original Armitage 1987a) is based on finite-element modelling. The CK
WJD method for qa > 6 MPa, it was expected that the underpredic- method (Carter and Kulhawy 1988) is based on analytical princi-
tion would be reduced for this range but increased for qa < 6 MPa. ples. The relative merits of the three methods were assessed and
Fig. 14 presents the new version of Fig. 12(b) (comparing pre- their ability to predict the performance of socketed piles was com-
dicted and measured loads to cause a 10-mm settlement) to show pared with several sets of test data available in the literature.
the effect of changing f su values. This figure shows that the effect The results indicate that, as would be expected, none of the
is quite small, with only minor shifts in the positions of the vari- methods is perfect. However, all three can produce reasonable pre-
ous points. A very similar result was obtained with the 5-mm dictions of performance. In broad terms, the RA method predicted a

© ASCE 04020144-11 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

 larger load for any given settlement than the other two methods and Table 7. Input data
generally overpredicted measured performance, including in the ρmax (mm) Fρ qa (MPa) Em (MPa) Ec (GPa) D (m) Qd (MN)
design settlement range up to about 10 mm. While this was partly
10 2 2.3 330 35 1 5
caused by the use of higher estimates of ultimate side stress, it does
not fully explain the outcome. The WJD method tended to under-
predict performance with loads for any settlement often less than
Table 8. Derived parameters for socket design using WJD method
the measured data. For the broad range of settlements considered,
the CK method both underpredicted and overpredicted. For design Parameter Source First iteration Last iteration
settlements typically up to 10 mm, both the WJD and CK methods Ec =Em — 106 106
generally underpredicted the measured loads. The WJD method α Fig. 3 0.24 0.24
captured the shape of the load-settlement curves quite well, fsu (MPa) Eq. (1) 0.56 0.56
whereas the other two methods were unable to do this. L (m) Eq. (7) 2.83 4.33
The CK method used is based on the assumption that the sock- L=D — 2.83 4.33
eted pile is rigid relative to the surrounding rock. The results Iρ Fig. 8 0.296 0.249
Qe (MN) Eq. (8) 5.58 6.64
obtained for settlements of 5 and 10 mm appeared to justify this
Qbe =Qe Fig. 9 0.18 0.14
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

assumption, although it is possible that pile compressibility may Q

have had a minor effect on the three tests, which appeared to be Qse =Qe ¼ 1 − be 0.82 0.86
Qe
short of the CK rigidity criterion. Q
Although almost all of the data used to derive the WJD method Qse (MN) ¼ se × Qe 4.55 5.70
Qe
were based on tests conducted in one type of rock encountered in Qse
one city with unconfined compressive strengths less than about fse (MPa) ¼ 0.51 0.42
π·D·L
10 MPa, it appears that its application has been validated with test fse =f su — 0.91 0.75
results involving many rock types from many locations around the fsp =fsu Fig. 5 0.32 0.20
world. This suggests that the mechanisms of behavior for all rock fs =f su Eq. (3) 0.59 0.54
socketed piles are essentially the same, thereby justifying the ap- f
fs (MPa) ¼ s × fsu 0.33 0.31
plication of normalized design techniques. f su
There are a number of factors not taken into account in the above Qs (MN) ¼ fs · π · D · L 2.94 4.16
discussions but that could be important in the overall design pro- Ns Table 1 0.0144 0.0163
cess. These include the influence of overburden resistance on the fb1 (MPa) Eq. (5) 4.75 5.37
Q
shaft above the socket; the compression of the shaft above the Qbe (MN) ¼ be × Qe 1.03 0.94
socket; any variability in the characteristics of the materials form- Qe
ing the socket; any debris that may be present at the base of the 4Qbe
fbe (MPa) ¼ 1.31 1.19
socket that reduces full contact between the pile and the rock at π·D·L
fbe =f b1 — 0.28 0.22
the base; other than clean contact between the concrete of the pile fbp =fb1 Fig. 7 0.04 0.02
shaft and the surrounding rock; exposed rock in the socket that may fb =f b1 Eq. (6) 0.24 0.20
have weakened after drilling has been completed; and adequate f
roughness of the side of the socket. fb (MPa) = b × fb1 1.16 1.08
f b1
π
Qb (MN) ¼ f b · · D2 0.91 0.85
4
Qb þ Qs (MN) — 3.85 5.01
Appendix. Example of Application of the WJD
Method

This appendix provides an example of how the WJD method is

used to design a rock socketed pile and how to produce load-
settlement curves for the pile. The input data for the design is pro-
vided in Table 7. Table 8 shows the steps involved in the first and
last iterations.
At the start of the first iteration, the socket length is assessed as
2.83 m based on the entire design load taken by the side of the
socket. After the first relaxation iteration, the side load, Qs , of
2.94 MN and the base load, Qb , of 0.91 MN provide a total load
of 3.85 MN for a 5-mm (ρmax =Fρ ) settlement. As the design load is
greater than this, the length can be increased progressively until the
design load is reached. Based on the above, a socket length of
4.33 m is required. The WJD method (Williams et al. 1980) pro-
vides guidance about how to check if geotechnical strength require-
ments, also applicable to limit state requirements, are satisfied.
Fig. 15 shows the load-settlement curve produced by the spread- Fig. 15. Load-settlement curves for socketed pile example.
sheet noted earlier for the example given above, along with the
curves of complementary side resistance and base resistance. As
would be expected, the curve for the complete pile shows that, ultimate load at a settlement of about 25 mm. Further, this ultimate
for the design load of 5 MN, the settlement is 5 mm. It may also side load (7.66 MN) is as would be expected and equal to the f su
be of interest that the base resistance is still developing even after a value given in Table 8 multiplied by the area of the side in contact
settlement of over 50 mm whereas the side resistance reaches its with the rock.

© ASCE 04020144-12 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144

 Data Availability Statement Foundations on Rock, edited by P. J. N. Pells, 55–64. Rotterdam,
Netherlands: A.A. Balkema.
Some or all data, models, or code that support the findings of this Kulhawy, F. H., and J. P. Carter. 1992. “Socketed foundations in rock
study are available from the corresponding author upon reasonable masses.” Chap. 25 of Engineering in rock masses, edited by F. G.
request (load-settlement data, spreadsheets). Bell, 509–529. Oxford, UK: Butterworth Heinemann.
Kulhawy, F. H., W. A. Prakoso, and S. O. Akbas. 2005. “Evaluation of
capacity of rock foundation sockets.” In Proc., 40th US Symp. on Rock
References Mechanics. Anchorage, AK: American Rock Mechanics Association.
Kwon, O. S., Y. Choi, O. Kwon, and M. M. Kim. 2005. “Comparison of the
AASHTO. 2017. AASHTO LRFD bridge design specifications. 8th ed. bidirectional load test with the top-down load test.” Transp. Res. Rec.
Washington, DC: AASHTO. 1936 (1): 108–116. https://doi.org/10.1177/0361198105193600113.
Akguner, C., and M. Kirkit. 2012. “Axial bearing capacity of socketed sin- Lam, T. S. K., and I. W. Johnston. 1982. “A constant normal stiffness
gle cast-in-place piles.” Soils Found. 52 (1): 59–68. https://doi.org/10 direct shear machine.” In Proc., 7th Southeast Asian Geotech. Conf.,
.1016/j.sandf.2012.01.012. 805–820. Hong Kong: Southeast Asian Geotechnical Society, Hong
Asem, P., and P. Gardoni. 2019. “Evaluation of peak side resistance for rock Kong Institution of Engineers.
socketed shafts in weak sedimentary rock from an extensive database Leung, C. F. 1996. “Case studies of rock-socketed piles.” Geotech. Eng.
of published field load tests: A limit state approach.” Can. Geotech. J. 27 (1): 51–67.
Downloaded from ascelibrary.org by Kevin Scott on 10/15/20. Copyright ASCE. For personal use only; all rights reserved.

56 (1–2): 1816–1831. https://doi.org/10.1139/cgj-2018-0590. Mallard, D. J., and J. L. Ballantyne. 1976. “The behaviour of piles in Upper
Aurora, R. P., and L. C. Reece. 1977. “Field tests of drilled shafts in clay- Chalk at Littlebrook D power station.” Géotechnique 26 (1): 115–132.
shales.” In Vol. 1 of Proc., 9th Int. Conf. on Soil Mechanics and https://doi.org/10.1680/geot.1976.26.1.115.
Foundation Engineering, 371–376. Tokyo: Japanese Society of Soil Osterberg, J. O. 1998. “The Osterberg load test method for drilled shafts
Mechanics and Foundation Engineering. and driven piles: The first ten years.” In Proc., 7th Int. Conf. on Piling
Carter, J. P., and F. H. Kulhawy. 1988. Analysis and design of drilled shaft and Deep Foundations, 1.28.1–1.28.11. Hawthorne, NJ: Deep Founda-
foundations socketed into rock. Rep. No. EL-5918, Research Project tions Institute.
1493-4. Palo Alto, CA: Electric Power Research Institute. Pells, P. J. N., and R. M. Turner. 1979. “Elastic solutions for the design and
Choi, S. K. 1984. “The bearing capacity of foundations in weak rock.” analysis of rock-socketed piles.” Can. Geotech. J. 16 (3): 481–487.
Ph.D. thesis, Dept. of Civil Engineering, Monash Univ. https://doi.org/10.1139/t79-054.
Day, R. A., I. W. Johnston, and D. Yang. 2009. “Design of foundations to Radhakrishnan, R., and C. F. Leung. 1989. “Load transfer behaviour of
Second Gateway Bridge—Brisbane.” In Proc., 7th Austroads Bridge rock-socketed piles.” J. Geotech. Eng. 115 (6): 755–768. https://doi
Conf. Auckland, New Zealand: Convention Management. http://www .org/10.1061/(ASCE)0733-9410(1989)115:6(755).
.cmsl.co.nz/assets/sm/3561/61/0046-E169Day.pdf. Rowe, R. K., and H. H. Armitage. 1987a. “A design method for drilled piers
Donald, I. B., H. K. Chiu, and S. W. Sloan. 1980. “Theoretical analyses in soft rock.” Can. Geotech. J. 24 (1): 126–142. https://doi.org/10.1139
of rock socketed piles.” In Proc., Int. Conf. on Structural Foundations /t87-011.
on Rock, edited by P. J. N. Pells, 303–316. Rotterdam, Netherlands: Rowe, R. K., and H. H. Armitage. 1987b. “Theoretical solutions for axial
A.A. Balkema. deformation of drilled shafts in rock.” Can. Geotech. J. 24 (1): 114–125.
Glos, G. H., III, and O. H. Briggs Jr. 1983. “Rock sockets in soft rock.” https://doi.org/10.1139/t87-010.
J. Geotech. Eng. 109 (4): 525–535. https://doi.org/10.1061/(ASCE) Rowe, R. K., and P. J. N. Pells. 1980. “A theoretical study of pile-rock
0733-9410(1983)109:4(525). socket behaviour.” In Proc., Int. Conf. on Structural Foundations
Goeke, P. M., and P. A. Hustad. 1979. “Instrumented drilled shafts in clay- on Rock, edited by P. J. N. Pells, 253–264. Rotterdam, Netherlands:
shale.” In Proc., Symp. on Deep Foundations, 149–165. New York: ASCE. A.A. Balkema.
Haberfield, C. M. 2013. “Performance of footings in rock based on Seidel, J. P. 1993. “The analysis and design of pile shafts in weak rock.”
serviceability.” Austr. Geomech. 48 (1): 1–50. Ph.D. thesis, Dept. of Civil Engineering, Monash Univ.
Haberfield, C. M., and A. L. E. Lochaden. 2019. “Analysis and design Seidel, J. P., and B. Collingwood. 2001. “A new socket roughness factor for
of axially loaded piles in rock.” J. Rock Mech. Geotech. Eng. 11 (3): prediction of rock socket shaft resistance.” Can. Geotech. J. 38 (1):
535–548. https://doi.org/10.1016/j.jrmge.2018.10.001. 138–153. https://doi.org/10.1139/t00-083.
Horvath, R. G. 1980. Research project report on load transfer for rock- Standards Australia. 2009. Piling, design and installation. AS 2159-2009.
socketed drilled pier foundations. DSS File No. 10x5.31155-9-4420, Sydney: Standards Australia.
Contract Serial No. 1SX79.00053. Ottawa: National Research Council Walter, D. J., W. J. Burwash, and R. A. Montgomery. 1997. “Design of
of Canada. large-diameter drilled shafts for the Northumberland Strait bridge
Horvath, R. G., and T. C. Kenney. 1979. “Shaft resistance of rock-socketed project.” Can. Geotech. J. 34 (4): 580–587. https://doi.org/10.1139/t97
drilled piers.” In Proc., Symp. on Deep Foundations, 182–214. -036.
New York: ASCE. Williams, A. F. 1980. “The design and performance of piles socketed into
Horvath, R. G., T. C. Kenney, and P. Kozicki. 1983. “Methods of improving weak rock.” Ph.D. thesis, Dept. of Civil Engineering, Monash Univ.
the performance of drilled piers in weak rock.” Can. Geotech. J. 20 (4): Williams, A. F., I. W. Johnston, and I. B. Donald. 1980. “The design
758–772. https://doi.org/10.1139/t83-081. of socketed piles in weak rock.” In Proc., Int. Conf. on Structural
Horvath, R. G., T. C. Kenney, and W. A. Trow. 1980. “Results of tests to Foundations on Rock, edited by P. J. N. Pells, 327–347. Rotterdam,
determine shaft resistance of rock-socketed piers.” In Proc., Int. Conf. Netherlands: A.A. Balkema.
on Structural Foundations on Rock, edited by P. J. N. Pells, 349–361. Williams, A. F., and P. J. N. Pells. 1981. “Side resistance rock sockets in
Rotterdam, Netherlands: A.A. Balkema. sandstone, mudstone and shale.” Can. Geotech. J. 18 (4): 502–513.
Johnston, I. W. 1977. “Rock socketing down–under.” Contract J. 279: https://doi.org/10.1139/t81-061.
50–53. Wong, P. K., and D. Oliveira. 2012. “Class A prediction versus performance
Johnston, I. W., and S. K. Choi. 1986. “A synthetic soft rock for laboratory of O-Cell pile load tests in Sydney Sandstone.” Austr. Geomech. 47 (3):
model studies.” Géotechnique 36 (2): 251–263. https://doi.org/10.1680 87–94.
/geot.1986.36.2.251. Zhan, C., and J.-H. Yin. 2000. “Field static load tests on drilled
Johnston, I. W., T. S. K. Lam, and A. F. Williams. 1987. “Constant shaft founded on or socketed into rock.” Can. Geotech. J. 37 (6):
normal stiffness direct shear testing for socketed pile design in weak rock.” 1283–1294. https://doi.org/10.1139/t00-048.
Géotechnique 37 (1): 83–89. https://doi.org/10.1680/geot.1987.37.1.83. Zhang, L., and H. H. Einstein. 1998. “End bearing capacity of drilled shafts
Johnston, I. W., A. F. Williams, and H. K. Chiu. 1980. “Properties of soft in rock.” J. Geotech. Geoenviron. Eng. 124 (7): 574–584. https://doi.org
rock relevant to socketed pile design.”. In Proc., Int. Conf. on Structural /10.1061/(ASCE)1090-0241(1998)124:7(574).

© ASCE 04020144-13 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(12): 04020144