Asem 2021

Rock Mechanics and Rock Engineering
https://doi.org/10.1007/s00603-021-02485-0
ORIGINAL PAPER
On the Use and Interpretation of In Situ Load Tests in Weak Rock

Masses
Pouyan Asem1 · Paolo Gardoni2
Received: 14 November 2020 / Accepted: 22 April 2021

© The Author(s), under exclusive licence to Springer-Verlag GmbH Austria, part of Springer Nature 2021
Abstract
Sockets are often constructed into weak rock mass to improve their axial response to applied compressive or tensile loads.
This paper first establishes two separate load test databases for the side and base resistances in in situ socket load tests in
weak rocks. The databases include the rock socket geometry, rock socket load–displacement response and the rock mass
properties. The paper then reviews the available models for the prediction of (i) the axial resistance (compression or uplift),
and (ii) axial deformation. The databases are used to evaluate the existing predictive models for axial resistance and axial
deformation, and to evaluate the behavior of drilled shafts in weak rock mass. The load test databases are also used to evalu-
ate the effect of rock mass variability on the reliability of foundations in weak rock. Based on the results of the analysis
presented, the applicability of the existing models and approaches to the study of socket behavior in weak rocks are discussed.
Keywords Side resistance · Base resistance · Settlement · Rock socket · Weak rock · Load test database · Reliability
analysis
List of Symbols Ddist Rock disturbance factor
A Rock socket cross-sectional area DL Dead load
a Empirical factor to account for the socket dc Depth factor
sidewall roughness Ec Concrete Young’s modulus
af = bf Empirical load displacement constant for Ei Intact rock deformation modulus
side resistance Em Rock mass deformation modulus
aq = bq Empirical load displacement constant for f′c Compressive strength of concrete
base resistance fs Unit side shear stress
ASD Allowable stress design fsu Unit ultimate side shear stress
B Rock socket diameter fsp Unit peak side shear stress
ci Cohesion intercept of intact rock fe Elastic side stress
cm Cohesion intercept of rock mass fp Plastic side stress
C Settlement constant FS Safety factor
DGS Depth of embedment from ground surface GSI Geological strength index
DTOR Depth of embedment from top of rock Gm Rock mass shear modulus
g Limit state function
h Rock socket interface roughness height
* Pouyan Asem hc Head of concrete
[email protected] I Embedment influence factor
Paolo Gardoni kf A correction factor that accounts for mobili-
[email protected] zation of unit side
1
Ksi Initial shear stiffness of rock socket
Department of Civil, Environmental and Geo‑Engineering,
sidewalls
University of Minnesota Twin Cities, 500 Pillsbury Drive
S.E., Minneapolis, MN 55455, USA Kn Base initial normal stiffness
2 Kns Initial normal stiffness of rock socket
Department of Civil and Environmental Engineering,
University of Illinois at Urbana-Champaign, 205 N. Mathews sidewall
Ave., Urbana, IL 61801, USA
13
Vol.:(0123456789)
P. Asem, P. Gardoni
Ksp Empirical factor accounting for joint spac- δλ C.O.V. of bias

ing and joint aperture η Rock socket construction factor
L Rock socket length Θf A facto that accounts for mobilization of
lc Rock socket roughness cord angle unit side resistance
l Span length λ Bias
lt Total travel distance along socket sidewall λR Bias for resistance
LRFD Load and resistance factor design λDL Bias for dead load
LL Live load λLL Bias for live load
mi Material constant for Hoek–Brown failure λ Mean of bias
criterion, intact material ν Poisson’s ratio
m Material constant for Hoek–Brown failure σh Standard deviation of rock socket roughness
criterion, rock mass height
M′ Measured quantity σi Standard deviation of rock socket roughness
M = 1/SJ Discontinuity frequency angle
n Number of observations σp Atmospheric pressure
Nγ = Nq = Nc Bearing capacity factors σnp Normal stress on rock socket sidewall at
P Total axial load mobilization of peak shear stress
Ps Axial load from side resistance σno Initial normal stress on rock socket sidewall
Pb Axial load from base resistance σv Normal total stress
Pi ith load effect σλ Standard deviation of bias
P′ Predicted quantity ϕi Internal friction angle of intact rock
Pf Probability of failure ϕm Rock mass friction angle
Q Q-system rock mass classification system ϕs Resistance factor for side resistance
q Unit base normal stress ϕb Resistance factor for base resistance
qall Allowable base contact pressure ϕint Rock/concrete interface friction angle
qy = q1 Base contact pressure at initial yield ω Joint aperture
qf = q2 Base contact pressure, interpreted failure
qu Unconfined compressive strength
qe Elastic base pressure 1 Introduction
qe–j Elastic base pressure at jth load increment
qp Plastic base pressure Rock sockets are being increasingly used for the support of
qp–j Plastic base pressure at jth load increment structures that carry heavy loads (Horvath and Kenney 1979;
qj Base pressure at jth axial load O’Neill and Reese 1999; Seidel and Collingwood 2001;
r Rock socket radius Brown et al. 2010; Dai et al. 2016; AlKhafaji et al. 2020).
RF Roughness factor Rock socketed foundations often interact with weak rocks
Rf Empirical fitting ratio for hyperbolic model (i.e., a rock mass with unconfined compressive strength, qu,
RMR Rock mass rating from ~ 0.5 to ~ 30 MPa; geological strength index, GSI < ~ 70
RQD Rock quality designation after Hoek and Brown 1997; or equivalently, rock mass rat-
SRC Socket roughness ratio ing, RMR < ~ 40). This is because the upper zone of most
s Material constant for Hoek–Brown failure rock formations hosting foundations is altered by physical
criterion, rock mass and chemical weathering, especially in tropical and humid
SJ Joint spacing areas (Goodman 1993). A weak rock mass consists of rel-
sγ = sq = sc Shape factors atively intact rock blocks that are separated by joint sets
(Hoek 1983; Jaeger et al. 2007). The behavior of foundations
Greek Symbols
in weak rock, therefore, is complex due to the existence of
α = fsp/qu Normalized peak side stress
rock mass secondary structures and their influence on rock
β Reliability index
mass mechanical properties.
βT Target reliability index
Some researchers have based their work on the results
γ Unit weight
of axial drilled shaft load tests in weak rock (e.g., Teng
γi ith load factor
1962; Coates 1967; Rosenberg and Journeaux 1976; Hor-
δ Axial socket displacement
vath and Kenney 1979; Meigh and Wolski 1979; Williams
δj Axial socket displacement at jth axial load
1980; Kaderabek and Reynolds 1981; Gupton and Logan
increment
1984; Rowe and Armitage 1987; Carter and Kulhawy 1988;
13
On the Use and Interpretation of In Situ Load Tests in Weak Rock Masses
Toh et al. 1989; Kulhawy and Phoon 1993; Zhang and Ein- by Rosenberg and Journeaux (1976) only contains eight
stein 1998; Miller 2003; Abu-Hejleh et al. 2003; Paikowsky drilled shaft load tests. The accuracy of estimation of
et al. 2010; Stark et al. 2013; Stark et al. 2017; Asem 2018; model parameters is dependent on the number of obser-
Asem et al. 2018; Baghdady 2018; Asem et al. 2019a, b; vations (e.g., load tests) (Gardoni et al. 2002). Use of
Asem 2019a, b; Asem and Gardoni 2019a, b, c; Tabandeh small databases for prediction of model parameters
et al. 2020). These studies have led to the development of could result in inaccurate predictions and large uncer-
empirical design methods where the observed side and base tainties in the estimated parameters. Some databases
resistances, and the deformational properties of rock sockets only focus on side or base resistance (Zhang and Ein-
have been related to the rock properties (e.g., unconfined stein 1998) and not both.
compressive strength, qu and rock mass deformation modu- 2. Availability of complete load–displacement information:
lus, Em) and to the rock socket geometry (e.g., rock socket In the majority of the current databases (see Table 1), the
length, L and rock socket diameter, B). These methods are information regarding the load–displacement response
sometimes site specific (e.g., Williams 1980; Miller 2003; (i.e., side shear stress-displacement and base contact
Abu-Hejleh et al. 2003). pressure–displacement, f s–δ and q–δ relationships,
Other researchers (e.g., Lam 1983; Haberfield 1987; respectively) are not reported. Therefore, such databases
Johnston and Lam 1989; Kodikara 1989; Seidel 1993; Has- cannot be used for a proper study of the load–displace-
san 1994; Collingwood 2000; Seidel and Collingwood 2001; ment mechanism of drilled shafts in weak rock mass that
Haberfield and Lochaden 2018; Johnson 2020) have used is a fundamental aspect of drilled shaft behavior.
a combination of theoretical and empirical methods for 3. Lack of basic rock mass mechanical properties: In most
the analysis of the behavior of rock sockets. These semi- of existing drilled shaft load test databases (e.g., Abu-
empirical methods are applicable to a broader range of field Hejleh et al. 2003; Stark et al. 2013, 2017; Baghdady
conditions, because they are mechanistic where the axial 2018), only the unconfined compressive strength (qu) for
behavior of drilled shafts is related to the fundamental vari- weak rock is reported. Moreover, the weathering condi-
ables that characterize the weak rock mechanical and engi- tion of weak rock is not discussed and the engineering
neering properties, the rock socket geometry and properties properties of the weak rock “mass” are not reported or
of mobilized shear surfaces on the rock socket sidewalls and estimated. Therefore, the actual problem involving the
the boundary conditions at the base of rock sockets. interaction of a drilled shaft with the weak rock “mass”
This paper first establishes two databases for drilled shafts is simplified to a problem where the rock mass is repre-
in weak rock namely (i) an in situ side-resistance database, sented by the properties of intact rock (i.e., unconfined
and (ii), an in situ base-resistance database. The databases compressive strength qu and intact rock deformation
include information on the rock socket geometry, rock socket modulus, Ei).
load–displacement response, and rock mass properties. The 4. Range of rock compressive strength: The range of rock
paper discusses the test methods and their interpretation unconfined compressive strength (qu) in some of the
technique. These databases are used to evaluate the exist- existing databases is well beyond the upper bound for
ing predictive models for axial resistance and settlement of the compressive strength of weak rock (i.e., 30 MPa)
rock foundations. The use of in situ databases in develop- (e.g., see database of Zhang and Einstein 1998).
ment of reliability-based methods is also discussed. Based 5. Range of rock socket size: The drilled shaft diameters (B)
on the results of the analyses provided, the applicability of reported in the existing databases cover a wide range.
the existing models to the study of drilled shaft behavior in However, drilled shaft diameters are not distributed uni-
weak rock is discussed. formly. Therefore, the effect of drilled shaft diameter
(i.e., size) cannot be properly investigated. For example,
data in Williams (1980) consist mostly of small diameter
(B) drilled shafts and the rock socket lengths (L) were
2 In Situ Load Tests usually short.
2.1 Existing Databases
The existing load test databases which are relevant to this 2.2 New Side‑Resistance and Base‑Resistance
study are summarized in Table 1. The following should be Databases
noted:
Two databases are given in the Supplemental Data Section.
1. Number of load tests: The number of load tests in some One is a side-resistance database, and one is a base-resist-
databases is limited. For example, the database collected ance database. The rock types include weak shale, siltstone,
13
Table 1 Summary of the existing load test databases
Database Number of load tests Location of load tests Rock type Unconfined compres- Rock mass Load–displace- Load test type
in rocks sive strength range, qu properties ment relationship
13
(MPa) reported reported
FHWA deep founda- Limited United States Sedimentary rocks No No

tion load test database
(DFLTD)
NCHRP report 507, 83 Sedimentary rocks No
Paikowsky et al.
(2004)
NCHRP report 651, 122 United States, United Claystone, mudstone, 0.2–103 Yes No Tip resistance
Paikowsky et al. Kingdom, Italy, Rus- sandstone, shale,
(2010) sia, Taiwan, Japan, till, chalk, diabase,
Canada, Australia, limestone, mudstone,
Singapore, South gypsum, granite, tuff
Africa
Rosenberg and 8 Interbedded shale and 0.5–55 No No Side and tip resistances
Journeaux (1976) sandstone, shale,
sandstone, limestone
and andesite
Horvath and Kenney 76 Canada, United States, Shale, clayshale, 0.35–110 No No Side resistance
(1979) United Kingdom, mudstone, siltstone,
Australia, Czecho- sandstone, chalk,
slovakia, Germany, limestone, marl,
South Africa, New diabase, basalt, schist
Zealand, Switzerland and slate
Williams (1980) Australia Siltstone 0.2–80 No Yes Side and tip resistances
Williams and Pells 71 Australia, Ireland, Mudstone, marl, clay- 0.48–62 No No Side resistance
(1981) South Africa, United stone, clayshale, shale
States, and sandstone
Rowe and Armitage 158 Australia, United States, Sandstone, mudstone, 0.412–110 Yes No Side resistance
(1984) Canada, United King- shale, claystone, silt-
dom, South Africa, stone, chalk, andesite,
Czechoslovakia, schist, limestone and
Ireland slate
Hassan (1994) 139 South Africa, United Shale, diabase, clay- 0.14–37.23 No No Side and tip resistances
States, Singapore, stone, marl, till, hard
Ireland, Canada, Bra- clay, hardpan, basalt,
zil, Australia, United siltstone, mudstone,
Kingdom, Hong Kong chalk, marl and
limestone
Zhang and Einstein 39 Australia, United States, Mudstone, shale and 0.52–55 No No Tip resistance
(1998), Zhang (1999) South Africa, Italy, clayshale, gypsum,
Singapore till, diabase, hardpan,
sandstone, siltstone,
P. Asem, P. Gardoni
marl and limestone

claystone, mudstone, limestone, sandstone, tuff, and granite.
Side and tip resistances
Side and tip resistances

The case histories where qu was larger than 30 MPa (that
is the selected upper bound for weak rock after Deere and
Side resistance
Load test type
Miller 1966; Barton et al. 1978; Rowe and Armitage 1987;

Kanji 2014) are not used in the development of the database
mainly because the aim of development of the databases
in this section is to study the performance of the available
design models with particular attention to weak rock. The
main contributions of the these databases include reporting:
ment relationship
Load–displace-
(i) the load-deformation properties for the rock sockets, and

(ii) the rock mass properties (e.g., deformation modulus
reported
Em). Additionally, the load test databases only concentrate

No
No
No on the axial behavior of rock sockets in weak rocks. The

side- and base-resistance databases are discussed below.
Rock mass
properties
2.2.1 Side‑Resistance Database
reported
No
No
No
The side-resistance database includes 317 drilled shaft load

tests. The peak side resistance (fsp) and shear stress–shear
sive strength range, qu
Unconfined compres-
displacement (fs–δ) relationship for each socket sidewall is

directly measured using in situ instrumentation. Therefore,
it was not necessary to assume a value of base resistance to
United states, Australia, Shale, clayshale, lime- 0.375–110
Shale, mudstone, clay- 0.12–4.45
obtain the fs–δ relationships. The side-resistance database

0.12–6.6
(MPa)
contains the following information:
1. Load-transfer function (fs–δ relationship) for rock socket

stone, sandstone, mica
sidewalls: The fs–δ relationships are obtained based on

siltstone, marl and
andesite, diabase,
schist, mudstone,
chalk, hard clay,
stone, mudstone
Shale, mudstone,
the measured drilled shaft load–displacement relation-

ship and the load-distribution versus depth measure-
claystone
Rock type
ments that are gathered using strain gauges or load cells

granite
(see Fig. 1). The displacements reported( in fs)–δ relation-

ships excludes the elastic deformation AE PL
of the test
c
shaft to give the net rock socket deformation, where P is
South Africa, Singa-
Location of load tests
United States, South
United States, South
the axial load, L is the rock socket length, A is socket

United Kingdom,
Africa, Australia,
Africa, Australia,
Canada, Ireland,
Ireland, Canada
Ireland, Canada
cross-sectional area and Ec is Young’s modulus of con-

crete. A typical fs–δ relationship is shown in Fig. 2b.
2. Properties of intact rock: The unconfined compressive
pore
strength (qu) of the intact rock specimens is reported for

all cases. The internal friction angle (ϕi), shear strength
intercept (ci), and intact rock deformation modulus (Ei)
54 (side) and 33 (tip)
93 (side) and 62 (tip)

Number of load tests
were not available for majority of case histories.

3. Properties of the rock mass: The rock mass modu-
lus is estimated using the method of Pells and Turner
(1979)—that is essentially a collection of solutions
in rocks
based on theory of elasticity for calculation of socket

162
settlement, based on the numerical integration of

Mindlin’s equation—as Em = 2𝜋LKsi I , where L is the
rock socket length, I is the embedment influence fac-
Collingwood (2000)
Table 1 (continued)
Stark et al. (2013)
Baghdady (2018)
tor (Pells and Turner 1979), and Ksi is the initial shear
stiffness (Fig. 2b) that is obtained from load tests.
Other properties of the rock mass are estimated using
Database
13
P. Asem, P. Gardoni
the same approach explained for base-resistance data- (drilled shaft and plate load tests) and contains the follow-
base. ing information:
4. Side resistance: Fig. 2b shows a typical fs–δ relationship
that illustrates the definitions of initial shear stiffness 1. Load-transfer function (q–δ relationship) for drilled
(Ksi), peak shear stress (fsp) and ultimate shear stress shaft base: The q–δ relationships (e.g., Fig. 2a) are
(fsu). The initial shear stiffness (Ksi) is the initial slope developed based on the top of rock socket load–displace-
of the fs–δ relationship. ment relationship and load distribution measured by
5. Drilled shaft geometry and material: The diameter for strain gages or tell-tales in conventional top-loaded load
each rock socket (B), and the depths of embedment from tests or the load–displacement relationship obtained
the ground surface ( DGS) and from the top of rock forma- from the O-cell in Osterberg load
( tests.
) The elastic com-
tion (DTOR) to the center of the shear profile are obtained pression of the test shaft AE PL
is considered and
based on the idealized site stratigraphy and drilled shaft excluded in the determination of deformation of rock
c
reported dimensions. The length of rock socket L is also mass under the rock socket base in top-loaded load tests.
provided for each case history. The method of construc- 2. Properties of the intact rock: The database contains
tion is summarized, when reported. The drilled shaft drilled shaft case histories that include the measured
concrete compressive strength (f′c), concrete modulus unconfined compressive strength (qu) of weak rock spec-
of elasticity (Ec), and concrete slump are summarized imens obtained from rock cores extracted from each load
when they are available. test site. The unconfined compressive strength (qu) of the
intact rock specimens is reported for all cases. The fric-
tion angle (ϕi), shear strength intercept (ci), and modulus
of deformation (Ei) for intact rock are not reported in the
majority of cases.
2.2.2 Base‑Resistance Database
3. Properties of the rock mass: The modulus of deformation
of the rock mass (Em) is estimated by the Authors from
Figure 1 shows the load test methods that are commonly
the load test results in each case history using the method
used to measure the base resistance of rock sockets. The
of Pells and Turner (1979) where the rock mass deforma-
base-resistance database includes 190 axial load tests
tion modulus (Em) is related to the initial normal stiffness
Reaction frame
Reaction frame Reaction frame
Top load Ground

cell Top load cell Ground Top load cell Ground
surface surface
surface
Top of soft Top of soft Top of soft

rock rock rock
Test shaft Reaction Reaction

Reaction DGS DTOR
DGS DTOR Test shaft shafts Strain
shafts DGS DTOR Test shaft shafts
gauge
Base load Void or

cell compressible
B
B base
B
Method (a) Method (b) Method (c)
P Reaction frame
Ground surface
Reaction frame
Ground surface
Strain gauge
Ground surface
Reaction Loading Reaction

Top of soft rock rod
shafts shafts
Top of soft rock
DGS DTOR Test shaft Top of soft rock

DGS DTOR Loading plate
Plug or anchor
DGS DTOR B
O-cell
B
B Method (f)
Method (d) Method (e)
Fig. 1 Load test methods used in the in situ test database
13
Fig. 2 Typical side resistance

and base displacement relation-
ships for rock sockets in weak
rock
of the rock socket base (Kn) as Em = 0.5𝜋BKn I (B is the where Em and qu are in units of MPa. The predicted
drilled shaft diameter and I is depth influence factor and is values of Em and the method of Hoek and Diederichs
based on the results of finite element models by Pells and (2006)—Em = 1+e 100,000
[75-GSI]
—are used to infer the geological
Turner 1979). For design purposes, Em can be estimated strength index (GSI). Rock mass Mohr–Coulomb
using a linear model of Em as a function of qu (Rowe strength parameters (Hoek and Brown 1997; Labuz and
and Armitage 1984; Asem 2018). This model is devel- Zang 2012), namely the drained friction angle (ϕm) and
oped using linear regression analysis of (i) the results of failure envelope shear strength intercept (cm) for rock
in situ plate load tests reported by Chern et al. (2004) for mass are estimated based on (i) the inferred values of
weak rocks in China and Taiwan, (ii) back-analysis of GSI as explained above, (ii) approximate material con-
q–δ relationships from 190 drilled shaft load tests from stant (mi) that relates to the type of rock and frictional
the United States, Puerto Rico, Canada, Australia, South characteristics of rock minerals based on values reported
Africa, Italy, United Kingdom and Singapore, and (iii) in Hoek (1983) and Marinos and Hoek (2001) and (iii)
back-analysis of 340 fs–δ relationships from 292 drilled the method of Hoek and Brown (1997) for estimation of
shaft load tests in weak rock as discussed in subsequent ϕm and cm. The predicted values of ϕm and cm are not
sections. This linear model may be expressed as: sensitive to mi (Marinos and Hoek 2001) and thus esti-
mation of mi using qualitative description of rock will
Em = 150q1.1
u
, (1) not significantly affect the predicted values of ϕm and
13
P. Asem, P. Gardoni
cm, especially for sedimentary and metamorphic rocks Therefore, significant effort has been devoted to the devel-
in the side-resistance database. opment of predictive models for side resistance. Design
4. Base resistance: The q–δ relationships (Fig. 2a) usu- methods for side resistance are commonly empirical and
ally exhibit a strain hardening response and a failure have been developed using load test databases that have a
point is not apparent. Therefore, the L1–L2 approach limited number of tests in a wide range of rock types with
(Hirany 1988) is used to interpret two failure points, a broad range of rock compressive strengths. Exceptions
i.e., the values of qy (i.e., contact pressure at initial yield include those of Rowe and Armitage (1984), Stark et al.
that is analogous to Hirany’s q1 stress) and qf (i.e., the (2013), and Asem (2018). Some methods, however, are
interpreted failure pressure that is analogous to Hirany’s based on laboratory constant normal stress and constant
q2 stress for Terzaghi 1943 failure pressure). When a normal stiffness (CNS) direct shear tests on synthetic rock/
distinct failure point was reported, it is included in the concrete interfaces where bonding between rock and con-
database. crete is prevented (e.g., Seidel 1993; Collingwood 2000;
5. Initial normal stiffness (Kn): The initial normal stiffness Seidel and Collingwood 2001). These methods commonly
(Kn) is defined as the slope of the tangent line to the relate the measured peak shear stress that is mobilized on
initial portion of q–δ relationship as shown in Fig. 2. the rock/concrete interface to the properties of the rock
6. Drilled shaft geometry: As shown in Fig. 1, the base (e.g., the unconfined compressive strength of intact rock, qu
diameter for each rock socket (B), and the depths of and rock mass deformation modulus, Em) and the properties
embedment of rock socket base from the ground sur- of the rock/concrete interface such as interface roughness
face (DGS) and from the top of rock formation (DTOR) height (h).
are obtained based on the reported site stratigraphy and
drilled shaft reported dimensions. The method of con- 3.1 Empirical Side‑Resistance Models
struction and condition of the drilled shaft base are sum-
marized, when reported. Rosenberg and Journeaux (1976) is among one of the first
models for peak side resistance. It is based on a total of
eight drilled shaft load tests where intact rock compressive
3 Review of Side‑Resistance Methods strength for four cases are reported. The load tests were con-
ducted on drilled shafts in sandstone, shale, limestone and
Rock sockets carry their loads in side resistance, base resist- andesite where 0.5 MPa < qu < to 34 MPa. Rosenberg and
ance, or combination of both (Horvath and Kenney 1979; Journeaux (1976) did not provide a mathematical expression
Rowe and Armitage 1987; Hassan et al. 1997; Zhang and for their proposed method. Kulhawy et al. (2005) provided a
Einstein 1998; Seidel and Collingwood 2001; CGS 2006; best-fit equation for the method of Rosenberg and Journeaux
Turner 2006; Asem 2018; Haberfield and Lochaden 2018). (1976) as:
Osterberg and Gill (1973) and Rowe and Armitage (1987)
using finite element analysis and Goodman (1980, 1989)
)0.56
fsp
(
q
and Seo and Prezzi (2008) using solutions based on the the- = 1.1 u , (2)
𝜎p 𝜎p
ory of elasticity showed that the portion of the load trans-
ferred to the rock socket base is only a small percentage of where fsp is the peak shear stress and σp is the atmospheric
the total applied load when the ratio of socket length to the pressure (σp = 0.101 MPa). In the method of Rosenberg and
drilled shaft diameter, L/B, is greater than ~ 4. Therefore, Journeaux (1976), (i) some of the load tests did not mobilize
rock sockets with L/B > ~ 4 and especially those socketed their peak side shear stress, (ii) the effect of rock type on
in stiff rock masses (i.e., Em/Ec > ~ 5) carry most of their side resistance is not evaluated, (iii) the effect of construc-
service load using side resistance (Seidel and Collingwood tion and shaft geometry on the development of shear stress
2001; Seo and Prezzi 2008). Additionally, the design of is not studied, and (iv) the effect of rock mass properties on
foundations for base resistance requires inspection pro- side resistance is not considered. The method of Rosenberg
cedures that guarantee the cleanness of the rock socket and Journeaux (1976) is evaluated using the new load test
base which has been proven to be expensive and difficult, database in Fig. 3a.
especially for the case of deep sockets in weathered rock The method of Horvath (1978) is based on the results
mass (Seidel and Collingwood 2001; Brown et al. 2010). of six drilled shaft load tests in Queenston Shale Forma-
The influence of the construction procedure and the det- tion where (i) all test shafts are constructed at one test
rimental effects of soft material near the rock socket base site near Ontario, Canada, (ii) the shale is horizontally
are discussed in the literature (e.g., Holden 1984; Turner bedded and 4.7 < q u < 11.1 MPa, and (iii) B = 710 mm.
2006; Brown et al. 2010; Haberfield and Lochaden 2018). The effect of drilled shaft geometry and embedment depth
13
Fig. 3 Evaluation of the empirical side resistance models using the in situ load test database (data from Asem 2018)
could not be investigated using the database of Horvath and

(1978) as all drilled shafts had the same diameter and the
depth of embedment was 1.37 m in all cases. Horvath fsp = 0.8qu RF0.45 , (4)
(1978, 1982), based on the measurement of socket wall
where h represents the average height of asperities that is
roughness, developed a roughness factor (RF) which is
measured with respect to the surface of the largest imaginary
used to account for the effect of the socket wall rough-
cylinder that can fit inside the rock socket, lt is the total travel
ness on the peak side stress (fsp). The predictive models
distance along the socket wall which accounts for the irregu-
for RF and fsp are:
lar texture of the rock socket sidewalls (e.g., see Seidel and
h lt Collingwood 2001 for additional discussion on characteriza-
RF = , (3) tion of sidewall roughness), r is the socket radius and L is
rL
13
P. Asem, P. Gardoni
the socket length. The roughness of the interface (h) can be where L is the length of the rock socket. Alternatively,
evaluated using the following expression (Asem 2018) that a more refined approach in Asem (2018) is based on
was obtained based on rock socket roughness measurements Mohr–Coulomb model and is written as:
compiled by Collingwood (2000) (see Fig. 4):
fsp = 𝜎np tan 𝜙m , (7)
50
h=
qu + 6.53
, (5) where σnp is the mobilized normal stress on the rock socket
sidewalls at the instant of mobilization of fsp and may be
where h is in mm and qu is in MPa. The average socket side- obtained as:
wall roughness height (h) is inversely related to the uncon-
0.37
fined compressive strength (qu), because a greater qu indi- 0.15Em
(8)
𝜎np = .
cates a greater resistance to abrasion and thus cutting tools L0.77 B0.01
are expected to create a smoother sidewall. The method of
These models (for Asem 2018, Eq. 7) are evaluated
Horvath is evaluated in Fig. 3. The results of recent research
in Fig. 3.
(e.g., Lam 1983; Haberfield 1987; Johnston and Lam 1989;
Kodikara 1989; Seidel 1993; Hassan 1994; Collingwood
3.2 Semi‑empirical Side‑Resistance Models
2000; Haberfield and Lochaden 2018; Johnson 2020) show
that drilled shaft/rock “interface” roughness dominates the
Hassan (1994) and Hassan et al. (1997) proposed that rock/
behavior of drilled shafts as it governs the development
concrete interface has a sinusoidal roughness pattern when
and maintenance of the normal stresses on the sidewalls of
the drilled shaft hole is constructed with rock auger in clay-
rock sockets. Field observations (Williams 1980), however,
shale and distinguished between smooth and rough sockets.
showed that the shear surface most likely forms within the
Hassan (1994) proposed that for the case of rough sockets,
rock mass and not at the drilled shaft/rock interface, espe-
the shear plane is within the weak rock material and thus the
cially in weak rocks. Therefore, the drilled shaft/rock inter-
unit side resistance (fsp) is equal to the average drained shear
face roughness may not be as important as is implied from
strength of geomaterial. Hassan (1994) and Hassan et al.
the laboratory tests conducted by previous investigators.
(1997) proposed the following expression for obtaining the
Other investigators have proposed similar empirical models
peak side resistance (fsp) of drilled shafts with rough rock/
to that of Rosenberg and Journeaux (1976) for side resist-
concrete interfaces:
ance of drilled shafts in weak rocks. These models are sum-
marized in Table 2 and evaluated in Fig. 3. fsp = cm + 𝜎no tan 𝜙m , (9)
Asem (2018) proposed two alternative models that
account for the rock socket geometry and rock mass prop- where it is assumed that cm and ϕm are the drained shear
erties. The model in Asem (2018) is written as: strength intercept and friction angle of rock mass, and σno
is the initial normal stress on the shear surface. For smooth
q0.22 0.31
Em rock/concrete interfaces, Hassan (1994) and Hassan et al.
fsp = u
, (6)
12.8L0.79 (1997) proposed that the peak unit side resistance (fsp) is
Fig. 4 Predictive model for

the drilled shaft/rock interface
roughness (data from Colling-
wood 2000; Asem 2018)
13
Table 2 Summary of the existing empirical methods for side resistance of rock sockets
References Equation Comments
Horvath and Kenney (1979) (i) Large- and small-scale drilled shafts and rock anchors, (ii) Rocks
√
fsp = a qu (MPa)
include shale, sandstone, limestone, chalk, and igneous and meta-
morphic rocks, (iii) 0.35 < qu < 110 MPa, and (iv) 0.2 < a < 0.25
(large diameter) and 0.25 < a < 0.33 (small diameter), where
parameter a is an empirical factor to account for √ the socket side-
wall roughness and is obtained by fitting fsp = a qu (MPa) to fsp
and qu data
Pells et al. (1979) fsp = 𝛼qu (i) Load tests in Hawkesbury sandstone in the Sydney, Australia, (ii)
6 < qu < 40 MPa, (iii) 75 < B < 710 mm, (iv) 240 < DGS < 1370 mm,
(v) 1 < h < 10 mm, and (vi) 0.1 < α (= fsp/qu) < 0.3, where α is the
ratio of fsp to qu
Williams (1980) fsp
( )0.28
q (i) weak Melbourne siltstone at Stanley Avenue, Middleborough
= 2.69 𝜎u
𝜎p p Road and Westgate Highway test sites, (ii) 0.4 < qu < 80 MPa, (iii)
0.335 < B < 1.35 m, and (iv) DGS = DTOR < 2 m
Kaderabek and Reynolds (1981) fsp = 0.3qu (i) Miami limestone, (ii) qu < 8.6 MPa, and (iii) 50 < rock quality
designation (RQD) < 80% in the upper part of the formation and
usually is near 0 in the lowest portion of the formation
Rowe and Armitage (1984) fsp = 0.45 √qu (MPa) for smooth and (i) Sandstone, mudstone, shale, siltstone, chalk and andesite, (ii)
√
fsp = 0.6 qu (MPa) for rough rock 0.55 < qu < 32 MPa, (iii) 0.1 < B < 1.12 m, and (iv) 0.2 < DGS < 10
sockets m
Stark et al. (2013), Stark et al. fsp = 0.3qu (i) Stark et al. (2013) developed a database of 54 drilled shaft load
(2017), and Baghdady (2018) tests mostly in Intermediate Geomaterial (IGMs) (i.e., shale,
mudstone, siltstone and claystone with 0.13 < qu < 3.2 MPa), (ii)
0.3 < B < 2 m, and (iii) Stark et al. (2017) and Baghdady (2018)
updated the method of Stark et al. (2013) using new load tests
(e.g., Vu 2013) where fsp = 0.31qu
mobilized at very small displacements (i.e., local displace- show no relationship between peak side resistance and ini-
ments of 5–10 mm). They also assumed that the shear sur- tial normal stress on the mobilized shear surface especially
face for smooth interfaces is at the rock/concrete interface when the rock socket sidewalls are relatively rough and
and suggested that the fsp in smooth sockets is governed by large dilations are expected to take place as a result. The
initial normal stress on the interface and interface friction method of Hassan (1994) is evaluated in Fig. 5 that shows
angle. For smooth rock/concrete interface, Hassan (1994) data do not provide a clear trend as that suggested by Has-
suggested the following expression: san (1994) method.
Alternatively, the design model of Seidel and Colling-
fsp = 𝛼qu , (10) wood (2001) is semi-empirical, uses analysis from the
where α is the adhesion factor written as ROCKET computer program, which is based on the
mechanistic models of Seidel (1993) and other research-
ers (e.g., Lam 1983; Haberfield 1987; Johnston and Lam
( )d−1
q tan 𝜙int
𝛼 = (5 − 8.8d) u , (11) 1989; Kodikara 1989; Seidel 1993; Hassan 1994; Colling-
𝜎p tan 30◦
wood 2000; Seidel and Collingwood 2001; Haberfield and
in which ϕint is the interface friction angle of rock/concrete Lochaden 2018). These mechanistic models were devel-
interface, and oped based on the observations of the behavior of natural
𝜎no
and synthetic rock/concrete interfaces in drained constant
15 − normal stiffness (CNS) direct shear tests. It is noted that
𝜎p
(12)
d=
27
. the bonding between the synthetic rock and the concrete
was sometimes prevented in the preparation of the speci-
In relation to the effect of stresses on side resistance, mens (e.g., Seidel 1993). The method accounts for the
Vesic (1963) indicated that the side resistance is a “linear effects of socket geometry (that is modeled using the nor-
function of vertical stress at failure.” Vesic (1963) fur- mal stiffness, Kns, Seidel and Collingwood 2001), qu, the
ther maintained that “this stress is not necessarily equal equivalent elastic properties of rock mass (i.e., modulus
nor proportional to the overburden pressure.” Contrary to of deformation, Em, and Poisson’s ratio, ν) and the socket
the proposal of Hassan (1994), drilled shaft load test data sidewall average roughness height (h). The method of
13
P. Asem, P. Gardoni
Fig. 5 Evaluation of the model

of Hassan (1994) and Hassan
et al. (1997) using the in situ
drilled shaft load test database
(data collected by Asem 2018)
Seidel and Collingwood (2001) is shown in Fig. 6. The stress (fsp) of rock sockets. For each model, we compute the
shaft resistance coefficient (SRC) in Fig. 6 is computed as: bias (λ) defined as:
Em M�
hqu (13) λ= , (15)
SRC = 𝜂 , P�
1+𝜈 B
where M′ is the measured quantity of interest and P′ is the
where η is the construction method reduction factor. One predicted quantity of interest. The bias (λ) for the method
of the important parameters that is accounted for in this of Rowe and Armitage (1984) is plotted versus various rock
model is the roughness of rock/socket interface. This model mass and rock socket properties in Fig. 8 as an example that
is based on idealizing the interface roughness by a series shows that λ varies with B, qu, Kns, the shear length (L) and
of interconnected cords. The assumption is made that the ϕm that suggests λ = f (B, qu, Kns, L). Further, because qu is
distribution of the cord angle is Gaussian and thus standard included in the method of Rowe and Armitage (1984), the
deviation of roughness height, cord angle and roughness variation of λ with qu indicates that the model form is not
angle are related as: appropriate. The existing models are also evaluated using
𝜎h = lc sin 𝜎i , (14) other databases compiled by other researchers (Horvath and
Kenney 1979; Rowe and Armitage 1984; Stark et al. 2013);
where σh is the standard deviation of roughness height, lc the results are presented in Table 3. The analyses based on
the cord length, and σi is the standard deviation of rough- our database are summarized in Table 4. The mean of bias
ness angle. The model of Seidel and Collingwood (2001) (μλ) is:
is evaluated using the in situ load test data in Fig. 7 which n ( )
shows a reasonable degree of agreement between measured 1 ∑ M�
𝜇𝜆 = , (16)
and predicted values. n i=1 P� i
where n is the number of observations in each database.

3.3 Discussion Similarly, the variance of M′/P′ (or other quantities of
interest such as initial normal stiffness or yield pressure as
The previous sections presented a review of the existing defined later in this paper) is estimated using the following
literature on the predictive models for the peak side shear expression (Ang and Tang 2007):
13
Fig. 6 Design method of

Seidel and Collingwood (2001)
(reproduced form Seidel and
Collingwood 2001)
M′/P′ ratio about its mean which is indicative of the model

accuracy:
𝜎𝜆
𝛿𝜆 =
𝜇𝜆
. (18)
The data presented in Tables 3 and 4 suggest that most of

the available models are biased and have high c.o.v.
4 Base‑Resistance Methods
In weak rock masses, a significant portion of the axial load

is carried in base resistance in short sockets (e.g., when
Em/Ec < 0.02; Seo and Prezzi 2008) after peak side resist-
ance (fsp) is mobilized and full-slip condition has material-
ized. Mobilization of failure mechanism in base resistance
requires a significant amount of displacement at the base of
drilled shaft that is significantly greater than that required for
mobilization of the peak side resistance (fsp). The available
Fig. 7 Evaluation of design method of Seidel and Collingwood
(2001) using the in situ load test data (data collected by Asem 2018) design models are of empirical and semi-empirical nature.
Empirical models are based on in situ or laboratory load
tests (e.g., Teng 1962; Coates 1967; Rowe and Armitage
1987; Carter and Kulhawy 1988; ARGEMA 1992; Zhang
n
and Einstein 1998; CGS 2006; Stark et al. 2013; Stark et al.
[( ) ]2
1 ∑ M�
𝜎𝜆2 = − 𝜇𝜆 . (17) 2017; Baghdady 2018; Asem 2018; Johnson 2020). Due to
n − 1 i=1 P� i
the large displacement required for the development of the
failure mechanism in the base resistance, most of the load
The square root of the variance is the standard deviation tests that were used in the development of empirical models
(σλ) of λ for each predictive model. The coefficient of varia- did not reach failure and thus the maximum mobilized base
tion (c.o.v.) for bias (δλ) is a measure of “dispersion” of the contact pressure was often used to develop predictive models
13
P. Asem, P. Gardoni
Fig. 8 Evaluation of the variation of bias (λ) for the design method of Rowe and Armitage (1984) using the in situ load test data (data collected
by Asem 2018)
for base pressure (q) (e.g., Zhang and Einstein 1998). There- developed semi-empirical predictive models for the base
fore, the failure mode varies from one predictive model to resistance based on an assumed failure mechanism (e.g.,
another. In many cases, the failure is based on serviceability Bishnoi 1968; Turner 2006) or by inconsistently combin-
requirements, where the resulting stresses correspond to dif- ing an assumed failure mechanism and the results of in situ
ferent drilled shaft axial displacements. Thus, the resulting load tests (e.g., Stark et al. 2013, 2017; Baghdady 2018).
design models cannot be readily compared and the predicted Therefore, a clear understanding of the assumptions and
base resistance using these models for one drilled shaft and failure criteria (i.e., definitions) or assumed mode of fail-
a given in situ condition can vary by several orders of mag- ure in the existing methods is a prerequisite to their evalu-
nitude (Zhang and Einstein 1998). Other researchers have ation and use. Some of the suggested predictive models are
13
Table 3 Evaluation of the side resistance predictive models using Table 4 Evaluation of the side resistance predictive models using
databases of Horvath and Kenney (1979), Rowe and Armitage (1984) new databases (this study) using 279 measurements at peak side shear
and Stark et al. (2013) stress
Design methods Databases Design Database (Asem 2018)

methods
Horvath and Rowe and Stark 35 > GSI 35 < GSI < 50 50 < GSI < 70 Entire database
Kenney (1979) Armitage (1984) et al.
(2013) Rosenberg and Journeaux (1976)
μλ 0.77 2.41 1.12 1.40
Rosenberg and Journeaux (1976)
σλ 0.63 1.94 0.78 1.50
μλ 0.83 1.15 1.08
δλ 0.82 0.81 0.69 1.07
σλ 0.45 0.58 0.44
Horvath and Kenney (1979)
δλ 0.55 0.5 0.41
μλ 1.11 3.60 1.71 2.07
Horvath and Kenney (1979)
σλ 0.91 2.91 1.19 2.25
μλ 1.24 1.73 1.57
δλ 0.82 0.81 0.70 1.09
σλ 0.68 0.87 0.65
Pells et al. (1979)
δλ 0.55 0.5 0.41
μλ 1.89 2.08 0.62 1.88
Pells et al. (1979)
σλ 1.55 1.54 0.42 1.54
μλ 0.89 1.07 1.78
δλ 1.89 0.74 0.68 0.82
σλ 0.72 0.95 0.88
Williams (1980)
δλ 0.8 0.89 0.5
μλ 0.61 2.57 1.39 1.38
Williams (1980)
σλ 0.52 2.12 0.99 1.65
μλ 0.89 1.3 0.96
δλ 0.85 0.83 0.71 1.20
σλ 0.54 0.71 0.42
Kaderabek and Reynolds (1981)
δλ 0.6 0.54 0.44
μλ 1.26 1.38 0.41 1.25
Kaderabek and
σλ 1.03 1.03 0.28 1.02
Reynolds (1981)
δλ 0.82 0.74 0.68 0.82
μλ 0.59 0.71 1.18
Rowe and Armitage (1984)
σλ 0.48 0.63 0.59
μλ 0.62 2.00 0.95 1.15
δλ 0.8 0.89 0.5
σλ 0.51 1.62 0.66 1.25
Rowe and Armitage (1984)
δλ 0.82 0.81 0.70 1.09
μλ 0.69 0.96 0.87
Carter and Kulhawy (1988)
σλ 0.38 0.48 0.36
μλ 1.39 4.50 2.14 2.58
δλ 0.55 0.5 0.41
σλ 1.14 3.64 1.49 2.81
Carter and Kulhawy (1988)
δλ 0.82 0.81 0.70 1.09
μλ 1.54 2.16 1.96
Hassan (1994) for rough interfaces
σλ 0.85 1.09 0.81
μλ 0.75 0.83 0.25 0.75
δλ 0.55 0.5 0.41
σλ 0.62 0.62 0.17 0.61
Hassan (1994) for rough interfaces
δλ 0.82 0.74 0.68 0.82
μλ 0.36 0.43 0.71
Hassan (1994) for smooth interfaces
σλ 0.29 0.38 0.35
μλ 3.35 9.47 6.47 5.55
δλ 0.8 0.89 0.5
σλ 3.59 9.43 6.26 6.87
Miller (2003)
δλ 1.07 1.00 0.97 1.24
μλ 0.77 1.08 0.98
Seidel and Collingwood (2001)
σλ 0.43 0.54 0.4
μλ 1.25 1.65 0.64 1.36
δλ 0.55 0.5 0.41
σλ 1.11 1.14 0.58 1.13
Kulhawy et al. (2005)
δλ 0.89 0.69 0.90 0.83
μλ 0.97 1.08 1.24
Kulhawy et al. (2005)
σλ 0.54 0.54 0.51
μλ 0.87 2.84 1.35 1.63
δλ 0.55 0.5 0.41
σλ 0.72 2.30 0.94 1.78
Stark et al. (2013)
δλ 0.82 0.81 0.70 1.09
μλ 0.59 0.71 1.18
Stark et al. (2013), Stark et al. (2017), Baghdady (2018)
σλ 0.48 0.63 0.59
μλ 1.26 1.38 0.41 1.25
δλ 0.8 0.89 0.5
σλ 1.03 1.03 0.28 1.02
13
P. Asem, P. Gardoni
Table 4 (continued) where qall is the allowable unit base resistance. Teng (1962)
Design Database (Asem 2018) pointed out that the evaluation of the effect of embedment
methods on the base resistance becomes exceedingly difficult in weak
35 > GSI 35 < GSI < 50 50 < GSI < 70 Entire database
rock because of the presence of joints, and bedding planes
δλ 0.82 0.82 0.68 0.82 and fissures. Therefore, the effect of embedment depth is
usually conservatively discounted. Additionally, it is unclear
what displacement is required for mobilization of the base
shown in Figs. 9 and 10 and are compared with qy and qf. resistance proposed by Teng (1962). Other investigators have
Additionally, the assumed failure mechanism (e.g., general proposed similar models for predicting the bearing capac-
shear failure mode) is often not mobilized due to restrictions ity of foundations on rock. These models are reviewed in
on the amount of allowable deformation (Terzaghi 1943) or Table 5 and are evaluated in Fig. 11.
increase in confining pressure that promotes a local type of
failure. Such change in failure mode with change in confin- 4.2 Semi‑empirical Base‑Resistance Models
ing pressure has been observed in the literature.
Coates (1967) developed a model based on the Griffith’s
4.1 Empirical Base‑Resistance Models fracture theory (Griffith 1921; Hoek 1965) suggesting that
microcracks in rock mass migrate and expand when the
Teng (1962) discussed the base resistance of drilled shafts stresses along their edges grow and become equal to the
in terms of allowable bearing pressure as: tensile strength of the material. Coates (1967) used a wedge
analysis in combination with Griffith’s fracture theory for
1 1
qall = to qu , (19) determination of bearing capacity of strip foundations to
5 8
write:
Fig. 9 Variation of the available

theories for base resistance (q)
and comparison with measured
yield pressure (qy) (data col-
lected by Asem 2018)
13
Fig. 10 Variation of the avail-

able theories for base resist-
ance (q) and comparison with
measured fracture initiation
pressure (qf) (data collected by
Asem 2018)
Table 5 Summary of empirical base-resistance models

Rowe and Armitage (1987) q = 2.5qu (i) Based on in situ load tests by Horvath (1982), Glos and Briggs (1983) and Wil-
liams (1980), (ii) 0.3 < B < 1 m, (iii) 0.65 < qu < 9.26 MPa, and (iv) rock types in
this database include siltstone, sandstone and shale sandstone
Zhang and Einstein (1998) (i) Based on a database of 39 drilled shaft load tests in mudstone, shale, gypsum,
√
q = 4.8 qu (MPa)
till, diabase, hardpan, sandstone, siltstone, marl and limestone, (ii) 0.3 < B < 1.92
m, (iii) 1 < DGS < 20.7 m, (iv) 0.52 < qu < 55 MPa, (v) 6 load tests reported are
model centrifuge tests (Leung and Ko 1993) on synthetic rock, and (vi) 3 load
tests from Carrubba (1997) were not instrumented and base resistance had to be
back-analyzed using elastic theory
Stark et al. (2013), Stark q = 2.5qu dc (i) Based on 33 drilled shaft load tests where the base stress–displacement rela-
et al. (2017), Baghdady depth factor (dc) (Vesic 1963) tionships (q–δ relationships) were back-calculated from the results of drilled
(2018) shaft load tests, (ii) rocks include shale and clayshale, claystone and siltstone,
(iii) 0.34 < qu < 4.2 MPa, (iv) 0.3 < B < 1.82 m, and (v) failure mechanism in
the base resistance is developed at a displacement of 5% of the drilled shaft tip
diameter
mobilized along the entire failure surface at the same time.

q = 3qu , (20)
In reality, however, the failure in brittle materials (e.g., rock)
where q is the base resistance of a long foundation. would initiate from regions of stress concentration and sub-
The model of Coates (1967) is developed for hard rock sequently propagate to less severely stressed zones (Hoek
and is based on the assumption that the strength of rock is 1965; Coates 1967). The method of Coates (1967) assumes
13
P. Asem, P. Gardoni
Fig. 11 Evaluation of empiri-

cal models for base resistance
with measure base resistance in
in situ load tests (data collected
by Asem 2018)
that the surcharge loads representing the embedment depth

are small compared to the applied loads, and do not con-
tribute to the bearing capacity. Hence, they do not account
for the effect of foundation depth on the bearing capacity
of foundations. Other researchers (e.g., Zhang and Einstein
1998) have made similar conclusions about the tip resistance
of drilled shafts in weak rock and in sand (Vesic 1963).
The method of Carter and Kulhawy (1988) is based on
the Hoek–Brown failure criterion for jointed rock mass and
was developed for foundations placed on the surface of rock
mass (Fig. 12). It is written as:
(21)
√
q = (m + s)qu ,
where q is the lower bound estimate of base resistance for

drilled shafts, and “m” and “s” are the material constants
from Hoek–Brown (Hoek and Brown 1980, 2019) crite- Fig. 12 Method of Carter and Kulhawy (1988) for the end resistance
rion. The method in Carter and Kulhawy (1988), however, of foundations on rock mass. Figure reproduced from Carter and
is not based on drilled shaft load tests but may be classi- Kulhawy (1988)
fied as a semi-empirical method only due to material con-
stants “m” and “s” which are determined empirically. As for side resistance cannot be checked. Other investigators have
the previous methods, it is unclear what displacements are developed semi-empirical models which are summarized
required for mobilization of the base resistance proposed in Table 6. The base-resistance models are evaluated in
by Carter and Kulhawy (1988) and thus compatibility with Table 7.
13
Table 6 Summary of semi-empirical base-resistance models

Canadian Geotechni- q = Ksp qu (i) q is the allowable bearing capacity of foundation, (ii) Ksp is an empirical coefficient that
3+S ∕B
cal Society (2006) Ksp = √ J includes a factor of safety of 3.0 (Ksp that ranges from 0.1 to 0.4), and (iii) CGS defines the
10 1+300𝜔∕SJ
“favorable conditions” as conditions, where “(i) the rock surface is perpendicular to the
foundation, (ii) the foundation load has no tangential component and (iii) the rock mass has
no open discontinuities.” 0.05 < SJ/B < 2 and 0 < ω/SJ < 0.02
Lee et al. (2013) (
q S
)0.315 (i) Parametric studies using numerical models, and (ii) in situ load test data from South Korea
q = 79.2 𝜎u BJ 𝜎p
p
Table 7 Evaluation of base-resistance models using existing models Table 8 Relationship between RQD (%) Allowable
allowable foundation pressure pressure
Design meth- Databases and rock quality designation
ods (MPa)
Paikowsky et al. Stark et al. Zhang and (RQD) for unweathered rock
(2010) (2013) Einstein masses (table after Peck et al. 0 1
(1998) 1974 or Peck 1976)
25 3
Teng (1962) 50 6.5
μλ 24.64 20.32 17.79 75 12
σλ 24.98 10.97 10.86 90 20
δλ 1.01 0.54 0.61 100 30
Coates (1967)
μλ 1.64 1.35 1.19
σλ 1.67 0.73 0.72 et al. 1974; Kulhawy and Goodman 1980) results in founda-
δλ 1.01 0.54 0.61 tion settlements that are less than 13 mm (Peck et al. 1974;
Rowe and Armitage (1987) Kulhawy and Goodman 1980). GSI of rock mass for each
μλ 1.97 1.63 1.42 load test is converted to RQD based on the recommenda-
σλ 2 0.88 0.87 tions of Hoek et al. (2013). The relationship proposed by
δλ 1.01 0.54 0.61 Peck et al. (1974) is plotted in Fig. 13. qy and qf from the
Carter and Kulhawy (1988) base-resistance database case histories interpreted using the
μλ 23.46 19.35 16.94 Terzaghi (1943) and Hirany (1988) methods and the esti-
σλ 23.79 10.45 10.34 mated RQD values are also shown. The comparison of the
δλ 1.01 0.54 0.61 data from 190 load tests (i.e., plate, O-cell and drilled shaft
ARGEMA (1992) load tests) with the proposed method of Peck et al. (1974)
μλ 1.09 0.9 0.79 in Fig. 13 shows that this method only provides crude esti-
σλ 1.11 0.49 0.48 mates of qf and qy and that significant scatter exists around
δλ 1.01 0.54 0.61 the mean correlation proposed by Peck et al. (1974). The
Zhang and Einstein (1998) data scatter in Fig. 13 indicates that in addition to the RQD,
μλ 1.64 1.01 1.06 that is mainly a measure of the joint frequency (Peck et al.
σλ 2.82 0.44 0.34 1974; Pells et al. 2017) in the weak rock mass, other rock
δλ 1.72 0.44 0.32 mass properties (i.e., rock mass friction angle ϕm) as well
Canadian Foundation Engineering Manual (2006)
as joint and fissure alteration, drilled shaft geometry (i.e.,
μλ 7.51
drilled shaft diameter) and displacement to mobilize qf may
σλ 6.92
also be important (Asem 2018; Tang et al. 2020). These
δλ 0.92
parameters are missing in Peck et al. (1974) predictive
equation. Moreover, it must be noted that the method of
Peck et al. (1974) was originally developed for “unweath-
4.3 Relationships with Rock Mass Properties ered and sound rock masses” and should be used with cau-
tion in weak rocks.
Peck et al. (1974) proposed a relationship between the Sowers (1979) solutions for the case of shallow founda-
allowable bearing pressure of foundations on unweathered tions on rock masses for situations where joint spacing (SJ)
and sound rock masses and the rock quality designation is less than the foundation size (i.e., drilled shaft or ingeneral
(RQD). The use of this correlation (see Table 8 or Peck foundation diameter, B) (see Fig. 14) are discussed below.
13
P. Asem, P. Gardoni
Fig. 13 Peck et al. (1974)

method for prediction of base
resistance based on rock quality
designation (RQD)
Fig. 14 Failure modes proposed

by Sowers (1979) for shallow
foundations on rock masses (fig-
ure is reproduced from Kulhawy
and Goodman 1980)
When rock joints are vertical, open and SJ < B, the bearing obtained from Sowers (1979). The models proposed by Sow-
capacity is obtained as: ers (1979) are plotted in Figs. 15 and 16 for different values
( ) of rock mass cohesion intercept (cm) and friction angle (ϕm).
𝜙 The qf values (obtained from q–δ relationships collected by
q = qu = 2cm tan 45 + m , (22)
2 Asem 2018) are also superposed on these figures. Examina-
tion of Figs. 15 and 16 shows that:
and when the joints are vertical, closed and SJ < B, the bear-
ing capacity is evaluated as:
1. These models cannot properly represent the actual vari-
B ation of qf with the rock mass friction angle for drilled
q = cm Nc sc + 𝛾N s + 𝜎v Nq sq , (23)
2 𝛾 𝛾 shafts in weak rock masses.
2. The significant scatter in the plots of qf versus friction
where all the material parameters in Eqs. (22) and (23) are
angle indicates that other parameters in addition to fric-
that of the rock mass. Nc, Nγ and Nq are the bearing capac-
tion angle and cohesion intercept must affect qf that are
ity factors and sc, sγ and sq are the shape factors and may be
13
Fig. 15 Comparison of the

model of Sowers (1979) for
foundations on rock masses
with vertical open joints when
SJ < B with the qf data from
the drilled shaft tip resistance
database
Fig. 16 Comparison of the

model of Sowers (1979) for
foundations on rock masses
with vertical closed joints when
SJ < B with the qf data from
the drilled shaft tip resistance
database
13
P. Asem, P. Gardoni
missing in the above proposed models (Asem 2018; elastic method. Williams (1980) does not propose a method
Tang et al. 2020). that could be specifically used for drilled shafts in weak
3. The inability of Sowers (1979) methods to properly rocks. Williams (1980), however, realizes that the behavior
predict the base resistance in weak rocks can also be of actual drilled shafts in weathered rocks is nonlinear.
attributed to assumed failure mechanism that is used to To account for the effect of nonlinearity observed in the
develop the above proposed design equations which is field for back-calculated fs–δ and q–δ relationships, Wil-
like to be different from that mobilized in the field. liams (1980) defined a plastic stress, fp or qp (for side and
tip, respectively), at each increment of axial displacement
which is described in mathematical form for side and tip
5 Settlement Methods resistance in Eqs. 24 and 25, respectively, as:
5.1 Load‑Transfer Methods
fp = fe − fs , (24)
The serviceability limit state is important in the design of and

foundations on rock masses (Goodman 1980) and should
be evaluated carefully. Some researchers developed pre- qp = qe − q, (25)
dictive models based on the theory of elasticity and some
where fe and qe are the predicted values of side and tip
used the load-transfer approach (Seed and Reese 1957) to
stresses based on the elastic theory at any given value of
develop nonlinear stress–displacement models for the side
local displacement along the side and near the tip of the
and the tip of drilled shafts. The models based on the the-
foundation, respectively, and f and q are the correspond-
ory of elasticity relate the settlement to rock socket geom-
ing values from the field “nonlinear” load-transfer functions
etry, rock mass properties, and drilled shaft concrete prop-
for side and tip resistances, respectively. The qe values in
erties. In the load-transfer approach, the stiffness of the
Figs. 17 and 18 are normalized using the yield pressure (qy)
shear surface mobilized in the socket walls and the stiff-
of the rock mass that is an alternative to what was proposed
ness of the tip of drilled shaft may be modeled using non-
by Williams (1980). A similar relationship may be devel-
linear springs (i.e., fs–δ and q–δ relationships). This model
oped for the side resistance fs–δ relationships. The method
is robust, because the effect of socket geometry, rock and
of Williams (1980) for normalization of load-transfer func-
concrete mechanical properties, local displacement, rock/
tions of drilled shafts in weak rock is a novel approach for
concrete interface characteristics (e.g., socket wall rough-
prediction of load-transfer functions. Although this method
ness), and construction method can be accounted for. The
has been mostly applied to investigation of load–displace-
most important advantage of the load-transfer method is
ment response of foundations in weak rock, it can be applied
that it accounts for the nonlinearity of the rock behavior,
to hard rock masses as was done in some of the load tests in
which is an important aspect of the behavior of weathered
relatively hard sedimentary rocks by Williams (1980). Asem
and weak rocks that cannot be properly addressed when
(2018) developed a family of equations that could be used to
theory of elasticity (e.g., Pells and Turner 1979; Kulhawy
predict the deviation of the tip response from linear elastic
and Carter 1992) is used for evaluating the settlement of
behavior written as:
foundations in soft rocks unless stress dependent elastic
properties are used. qp−j
(
qe−j
)3 (
qe−j
)2
qe−j
Williams (1980) developed a load-transfer function that = 0.007797 + 0.05209 , ≤ 5,
qy qy qy qy
is based on the results of field load tests in Melbourne
(26)
siltstone where side and tip resistances were directly
measured. The siltstone at the load test sites is commonly and
weathered to highly weathered, 0.335 < B < 1.35 m and qp−j
(
qe−j
)
qe−j
DTOR < 2 m (the ground surface and the top of weak rock = 0.9436 − 2.467, > 5. (27)
qy qy qy
coincided in most of load tests). Williams (1980) proposed
a method for normalizing the fs–δ and q–δ relationships Hirayama (1990) method is based on the modeling of the
(i.e., load-transfer functions for side and tip resistance, load-transfer function for unit side resistance and tip resistance
respectively) for drilled shafts in weak rock. The normali- using a “two-constant” Hyperbolic model written as:
zation method, with particular attention to tip resistance,
is shown in Fig. 17 and is applied to 190 load tests in weak 𝛿
fs = , (28)
rock in Fig. 18. Williams (1980) method first predicts the af + bf 𝛿
load–displacement response of the foundation using an
and
13
Fig. 17 Normalization method

developed by Williams (1980)
(Figure after Williams 1980,
with modifications)
𝛿 the back-calculated and theoretical values. Other load trans-

q=
aq + b q 𝛿
, (29) fer models have been developed by other investigators which
are summarized in Table 9.
where fs and q are the unit side and base resistances of drilled
shaft and af, bf, aq and bq are model parameters that define 5.2 Initial Shear Stiffness and Initial Normal
the shape of the hyperbolic function. Hirayama relates the Stiffness Models
model parameters to the “ultimate” side and base resistances
of drilled shafts. The peak side resistance of drilled shafts 5.2.1 Evaluation of Models for Initial Shear Stiffness in Side
in weak rock can be estimated using the available predictive Resistance (Ksi)
models for the unit side resistance of drilled shafts in weak
rock. In the case of base resistance, however, the majority The initial shear stiffness for rock joints has been studied
of the field load tests have not been carried to failure and extensively (e.g., Bandis 1980; Bandis et al. 1983; Yoshi-
thus reliable methods for the prediction of failure stress in naka and Yamabe 1986). The problem of determination of
tip resistance of drilled shafts in weak rock are not available. the initial shear stiffness (Ksi) in side resistance of rock socket
This is an important limitation of Hirayama’s method. Hiray- drilled shafts in weak rock, however, has not been explored in
ama (1990) does not account for the difference between peak detail. Therefore, it is necessary that the existing models for
side or ultimate base resistance predicted from the hyper- the prediction of initial shear stiffness be evaluated against
bolic model and the actual measured values, and thus does the back-calculated initial shear stiffness (Ksi) from the results
not use a fitting ratio (Mesri et al. 1981; Terzaghi et al. 1996) of drilled shaft load tests. One of the most widely used cor-
(Rf) that has been used to account for this difference between relations is that of Randolph and Wroth (1978):
13
P. Asem, P. Gardoni
Em
Gm = , (31)
2(1 + 𝜈)
where ν is the Poisson’s ratio. It must be noted that Eqs.

(30) and (31) are based on the theory of elasticity and are
only applicable to small axial displacements (i.e., displace-
ments less than ~ 0.5% of the socket diameter to ensure shear
stresses on the socket sidewalls are significantly less than
the peak shear strength of rock socket and hence to ensure
socket would remain in elastic range). Hirayama (1990) pro-
posed an empirical design model for estimation of the initial
shear stiffness (Ksi) in side resistance of drilled shafts that
is shown below:
fsp
Ksi = , (32)
0.0025B
where fsp is the peak side resistance and B is the diameter
of the drilled shaft. Hirayama (1990) recognizes that the
value at which peak side resistance is developed can range
from 0.1 to 0.5% of the drilled shaft diameter (B). The peak
Fig. 18 Empirical method of Williams (1980) (modified) for predic- shear stress must be determined before this model can be
tion of qp for the purpose of determination of nonlinear load-transfer
functions for tip resistance (all data from Asem 2018 that is compiled
used. This introduces additional uncertainty because current
from literature, normalization method after Williams 1980 with mod- predictive methods for peak side resistance are not accurate
ifications and updates) as suggested by comparison of these methods with other
side-resistance databases.
These predictive models are evaluated herein in terms
Gm of the mean and coefficient of variation (c.o.v.) of the
Ksi = , (30)
r ratios of measured to predicted initial shear stiffness
(μλ and δλ, respectively). It must be noted that the meas-
where Ksi denotes the initial shear stiffness in side resistance
ured (i.e., back-calculated) initial shear stiffness values
of drilled shafts, Gm is the shear modulus of the rock mass
are obtained from the back-calculated fs–δ relationships
and r is the radius of the drilled shaft. The shear modulus of
reported in the side-resistance database and following the
rock mass (Gm) can be related to its modulus of deformation
procedure that is outlined in the preceding sections of this
(Em) (Goodman 1980) as:
paper. The evaluation results are summarized in Table 10.
The bias (i.e., ratio of the measured to predicted initial
Table 9 Summary of load transfer models
Hassan et al. (1997), side resistance fs = Θf fsp for pre-peak and (i) kf and Θf are correction factors that account for mobilization of
fs = kf fsp for post-peak unit side resistance and are functions of displacement, (ii) fsp is the
peak unit side resistance, and (iii) kf and Θf could be obtained from
Hassan et al. (1997)
Hassan et al. (1997), tip resistance q = Λ𝛿 0.67 based on the results of finite element numerical models (FEM)
Gupta (2012) 𝛿
method does not provide any recommendations for the prediction of
fs = B
5L(1−𝜈)
load-transfer function for tip resistance of drilled shafts
[ ]
ln 𝛿R
2Gm
B
+ Bf f
sp
Stark et al. (2013) 3.2 B𝛿 (i) Based on of 33 drilled shaft load tests, and (ii) rock types in the
q= qu dc ≤ 2.5qu dc
𝛿
B
+1.3 database included shale and clay–shale, claystone and siltstone
Lee et al. (2013) q= 1
𝛿 (i) Was based on the parametric studies that were performed using
+ q𝛿
Kn where f numerical models, (ii) model depends on δ, B, Em and the discon-
tinuity spacing (SJ), and (iii) Bref = 1.0 m and σp is the atmospheric
( )0.4058 ( 𝜎 )
E S p
Kn = 0.00037 𝜎 mB J
p ref B pressure (0.101 MPa)
( )0.315
q S
qf = 79.2 𝜎u BJ 𝜎p
p
13
Table 10 Evaluation of the predictive models for the initial shear shaft geometry and weak rock mass mechanical properties
stiffness (Ksi) of drilled shafts in weak rock mass should be accounted for in the predictive models.
Design Database (Asem 2018) 5. In addition to the possibility of existence of missing
methods parameters in the current predictive models, the strong
correlation between the bias and weak rock properties
Randolph and Wroth (1978) and the drilled shaft geometry may results from model
μλ 2.70 0.84 0.63 1.62 inexactness, meaning that the mathematical form of the
σλ 5.39 1.23 1.01 3.77 current predictive models may not accurately describe
δλ 1.99 1.46 1.61 2.33 the actual variation of initial shear stiffness with rock
Hirayama (1990) and drilled shaft properties.
μλ 2.14 0.92 2.13 1.6
σλ 7.47 1.32 4.44 5.37
δλ 3.49 1.43 2.08 3.36
5.2.2 Evaluation of Models for Initial Normal Stiffness in Tip
Resistance (Kn)
shear stiffness) is also plotted against different variables Similar to the initial shear stiffness (Ksi) of rock socketed
from the side-resistance database and is shown in Figs. 19 drilled shafts, the initial normal stiffness (Kn) for the tip
and 20. The following observations can be made: of drilled shafts in weak rock mass has not been studied
rigorously. Lee et al. (2013) developed a model for pre-
1. Table 10 shows the statistics of bias (λ) for initial shear diction of the initial normal stiffness of drilled shaft tip
stiffness (Ksi). The statistics of bias are calculated and in rock. Pells and Turner (1979), Goodman (1980), Jeong
grouped based on the condition of soft rock mass in which et al. (2010) and Lee and Jeong (2016) developed models
the drilled shafts have been constructed. These are rock for prediction of load–displacement behavior of shallow
masses with 50 < GSI < 70, 35 < GSI < 50 and GSI < 35. and drilled shaft foundations in rocks that may be used to
2. Table 10 shows that the method of Randolph and Wroth obtain correlations for prediction of initial normal stiffness
(1978) underestimates the initial shear stiffness (Ksi) for for the tip of drilled shafts.
the drilled shafts in rocks with GSI smaller than 35. The The method of Goodman (1980) may be used to develop
analysis also shows that the variability in the predictions the following expression for prediction of initial normal
using the method of Randolph and Wroth (1978) is con- stiffness (Kn) for the tip of drilled shafts:
siderably greater for very weathered and jointed weak
Em
rock mass (i.e., GSI < 35) as compared to better qual- Kn = ( ) , (33)
ity rocks. The predictive method of Hirayama (1990) is C 1 − 𝜈2 r
unconservative for weak rock masses with GSI between
where C is a constant that depends on the boundary condi-
35 and 50. The variability of the predictions also is con-
tions (Goodman 1980) and is equal to π/2 if the foundation
siderably larger for more weathered weak rocks as is the
can be shown to behave as a rigid structure compared to
case for the method of Randolph and Wroth (1978).
the weak rock or 1.7 if the foundation is a flexible structure
3. The variations of bias for the predictions of the methods
compared to the weak rock mass (Goodman 1980), ν is the
of Randolph and Wroth (1978) and Hirayama (1990)
Poisson’s ratio (see Gercek 2007 for typical values for rocks)
with different properties of weak rock mass and drilled
and r is the radius of the foundation (i.e., B/2).
shaft geometry are shown in Figs. 19 and 20. These fig-
Pells and Turner (1979) also used a similar method to
ures show that the error of prediction of these methods
that proposed by Goodman (1980). Pells and Turner (1979),
for initial shear stiffness varies significantly with the
however, add a correction factor to account for the effect of
diameter of the drilled shaft (B), depth of embedment
embedment depth on load–displacement behavior of the tip
(DGS) (or equivalently the degree of confinement of the
of drilled shafts in rock mass as:
side of the drilled shaft), the normal stiffness of the sur-
rounding rock mass (Kn) and the unconfined compres- Em
sive strength of the weak rock (qu). Kn = ( ) , (34)
C 1 − 𝜈 2 rI
4. The correlation between bias in the predictions of the meth-
ods of Randolph and Wroth (1978) and Hirayama (1990) where I is the correction factor for embedment. This is a
and weak rock mass properties and drilled shaft charac- theoretical factor that is a function of Poisson’s ratio (ν)
teristics indicates that additional properties of the drilled of rock and depth of embedment in rock to diameter ratio
13
P. Asem, P. Gardoni
Fig. 19 Variation of bias (λ) in

the predictions of the method of
Randolph and Wroth (1978) for
initial shear stiffness (Ksi) with
properties of weak rock mass
and drilled shaft geometry
(DTOR/B). I ranges from 0.4 to 1.0 for DTOR/B ratios of 0 to RQD = 100e−0.1M (0.1M + 1), (36)
10 (Pells and Turner 1979).
Lee et al. (2013) proposed a predictive model for the where RQD is the rock quality designation and M is the
initial normal stiffness (Kn) for the tip of drilled shafts in discontinuity frequency. RQD can be converted to GSI using
rock masses. The method is semi-empirical, because it was the recommendations of Hoek et al. (2013).
formulated based on parametric studies performed using These methods are evaluated herein using the mean (μλ)
numerical models and was calibrated using a drilled shaft and coefficient of variation, c.o.v., (δλ) of ratios of back-
load test database that was reported by Lee et al. (2013) and calculated to predicted initial normal stiffness for the tip of
is written as: drilled shafts in weak rock masses. It must be emphasized
)0.4058 ( that the back-calculated initial normal stiffness values
𝜎p
( )
Em SJ are obtained from the back-calculated q–δ relationships
Kn = 0.00037 , (35)
Bref 𝜎p B reported in the tip resistance database. The analysis results
are summarized in Table 11 and the diagnostic plots are
where Kn is in units of MPa/mm, SJ is the discontinuity spac- shown in Figs. 21 and 22. The findings and summary of
ing, Bref = 1.0 m, 𝜎p is the atmospheric pressure (= 0.101 observations are discussed below:
MPa) and B is the diameter of the drilled shaft. The mean
discontinuity frequency per meter (i.e., M = 1/SJ) may be 1. The review of statistics of bias shows that all three initial
obtained from the following expression proposed by Priest normal stiffness models that are evaluated in this section
and Hudson (1976): underestimate the initial normal stiffness by a large margin.
13

the predictions of the method
of Hirayama (1990) for initial
shear stiffness (Ksi) with prop-
erties of soft rock mass and
drilled shaft geometry
2. The c.o.v. of the bias for predictive models for initial 6 Estimation of Model Parameters
normal stiffness does not show a clear trend with the
condition of rock mass. However, study of the statis- The rock mass deformation modulus (Em) is used when
tics of bias for the models of Pells and Turner (1979) deformations in rock mass are calculated (Bieniawski 1978;
and Lee et al. (2013) suggest that these models tend to Hoek and Diederichs 2006). Previous investigators have used
become less accurate as the rock becomes more frac- published data to develop predictive models for strength and
tured (i.e., as the GSI values decrease). elastic properties of rock (e.g., Labuz et al. 2018; Guevara-
3. The examination of variation of bias (λ) with the properties Lopez et al. 2019; Asem 2020; Asem et al. 2021). Following
of rock mass and drilled shaft geometry indicates a strong this approach, we used in situ and laboratory test results to
correlation between bias and drilled shaft diameter (B), evaluate the existing methods, and to recommend predictive
unconfined compressive strength of intact rock (qu), geologi- models for the estimation of Em.
cal strength index (GSI) and rock mass friction angle (ϕm).
6.1 Unconfined Compressive Strength and Rock
Mass Deformation Modulus
A database of rock mass deformation modulus (Em) and the

unconfined compressive strength (qu) of weak rock in drilled
13
P. Asem, P. Gardoni
shaft load tests was developed by Asem (2018). These load Equation (37) can be used to predict Em based on the uncon-
tests were augmented by those collected by Chern et al. (2004) fined compressive strength (qu):
that include sedimentary, igneous, and metamorphic rocks
from China and Taiwan. These data are plotted in Fig. 23. Em = 150q1.1
u
, (37)
where Em and qu are in units of MPa.
6.2 Rock Mass Indices and Rock Mass Deformation

Table 11 Evaluation of the predictive models for the initial normal Modulus
stiffness (Kn) of drilled shafts in weak rock mass
Design Database (Asem 2018) Bieniawski (1978) reviewed the results of in situ load tests
methods from Elandsberg scheme, Drakenbergs scheme, Orange Fish
Tunnel, Witbank Coalfield, Le Roux Dam, Dinorwic Scheme
Goodman (1980) and Gordon scheme projects and developed the following
μλ 3.65 5.32 61.39 6.72 expression for Em:
σλ 4.43 7.18 78.92 20.54
δλ 1.21 1.35 1.29 3.06 Em = 1.76RMR − 84.3, (38)
Pells and Turner (1979)
where Em is in units of GPa, and RMR is the rock mass rat-
μλ 1.91 1.66 1.48 1.82
ing that should be obtained in accordance with the Geome-
σλ 0.93 0.44 0.66 0.83
chanics Classification system (Bieniawski 1976). Serafim
δλ 0.49 0.26 0.44 0.45
and Pereira (1983) reviewed the proposed method of Bie-
Lee et al. (2013)
niawski (1978) and concluded that Bieniawski’s correlation
μλ 2.27 5.37 30.90 4.68
between Em and RMR is not applicable to very poor, poor
σλ 1.97 2.96 28.44 9.08
and fair quality rocks. Therefore, Serafim and Pereira (1983)
δλ 0.87 0.55 0.92 1.94
collected additional data from Cambambe Dam, Castelo Do
Bode Tunnel, Funcho Dam, Ilisu Dam, Keddara Dam and

of Goodman (1980) for drilled
shaft tip normal stiffness (Kn)
with properties of weak rock
mass and drilled shaft geometry
using tip resistance database
13

of Lee et al. (2013) for drilled
shaft tip normal stiffness (Kn)
with properties of weak rock
mass and drilled shaft geometry
using the tip resistance database
Sfiki Dam to improve Bieniawski’s equation. The method of statistics (i.e., μλ, σλ and δλ) of the ratios of measured to
Serafim and Pereira (1983) is written as: predicted (λ) rock mass modulus of deformation. Surpris-
ingly enough, the statistics presented in Table 14 indicate
Em = 10(RMR−10)∕40 , (39) that the inclusion of the rock mass indices in the equations
for prediction of the rock mass modulus of deformation
where Em is in units of GPa. Hoek and Diederichs (2006)
did not significantly improve the ability of these models
reviewed a large database that was collected by Chern et al.
to give accurate estimates of the weak rock mass modulus
(2004). Hoek and Diederichs (2006) model is written as:
of deformation compared to less sophisticated techniques
that are mainly functions of properties of intact weak rock
( )
1 − Ddist ∕2
Em = 100, 000 , (40) substance (e.g., Eq. 37). This may be explained as follows
1 + e[75+25Ddist −GSI]∕11
in relation to the sedimentary weak rock masses:
where Em is in units of MPa and Ddist is a rock disturbance
factor that accounts for damage and fracturing due to blast 1. Rock mass indices such as GSI and RMR have been
and stress relief (Hoek and Diederichs 2006). Figure 24 sug- developed mainly to reflect the effect of rock mass discon-
gests the method of Hoek and Diederichs (2006) is the most tinuity characteristics on the shear strength and deform-
accurate representation for the relationship between rock ability of the rock mass. These classification systems are
mass modulus of deformation and the geological strength effective when the shear strength and the deformational
index. According to Hoek and Diederichs (2006), RMR properties of the discontinuities are significantly different
and GSI can be used interchangeably, particularly if RMR from that of intact rock substance. The shear strength of
is obtained prior to the year 1990. This assumption is used the joints and the rock substance in a weak rock are com-
in plotting Eqs. (38) and (39) and Fig. 24. monly very similar due to weathering processes, erosion,
Other available predictive models for E m (e.g., see swelling and subsequent softening (Mesri and Shahien
Table 12; Hiltscher et al. 1984) are applied to the data- 2003) that occurs in these rock masses.
bases introduced in Table 13 and to that compiled by the 2. Most of the rating components (e.g., joint condition,
Authors that is reported in Asem (2018). The evalua- ground water condition, and joint orientation) that are
tion results are summarized in Table 14 in terms of the used in the evaluation of the rock mass indices (e.g.,
13
P. Asem, P. Gardoni
Fig. 23 Variation of modulus of

deformation of rock mass (Em)
with unconfined compressive
strength for different rock types
and the proposed design trend
line. All data from published
literature and summarized in
Asem (2018)
Fig. 24 Variation of modulus

of deformation of rock mass
(Em) and geological strength
index (GSI), and evaluation of
predictive methods for modulus
of deformation of rock mass.
All data from published litera-
ture and summarized in Asem
(2018)
RMR or GSI) are based on the qualitative observations

of the rock mass features. Therefore, a great amount of Based on the discussion provided above, it appears rea-
subjectivity may enter the evaluation process that will sonable to conclude that the simplified predictive model
increase the scatter of the available databases and thus based on the unconfined compressive strength that is pro-
will reduce the effectiveness of RMR and GSI. posed herein may be used in the evaluation of modulus of
deformation of the weak rock mass for design purposes
13
Table 12 Summary of Number Empirical methods (Em in units of GPa) References

predictive methods for the
modulus of deformation of rock 1 Ei Nicholson and Bieniawski (1990)
Em = , Ei = 50 GPa
mass 100[0.0028RMR2 +0.9 exp (RMR/22.82)]
2 Em =
Ei [1−cos (𝜋⋅RMR/100)]
, Ei = 50 GPa Em Mitri et al. (1994)
2
3 Em = 0.1(RMR∕10) 3 Read et al. (1999)
4 Barton (2002)
√
Qq
Em = 10 3 100u , qu = 100MPa
5 Carvalho (2004)
√ [ ]
Em = Ei ⋅ 4 exp (GSI − 100)∕9 , Ei = 50 GPa
6 Diederichs and Kaiser (1999)

√
Em = 7 10 ⋅ [RMR − 44]∕21
Table 13 Rock mass classification and rock mass deformation modulus database
Rock type Joint infilling material Plate Em (GPa) RMR-76 RMR-89 GSI Q References
diameter
(cm)
Fault breccia Sand and gravel with silt 30.2 0.25 15 18 10 0.0078 Keffeler (2014)
Oxidized limestone Silty sand with gravel 30.2 4.5 41 34 45 0.27 Keffeler (2014)
Argillized rhyolite dike Clayey sand with gravel 30.2 0.37 16 21 15 0.017 Keffeler (2014)
Argillized rhyolite dike Clayey sand 25.4 0.031 16 21 10 0.01 Keffeler (2014)
Decalcified limestone Carbon and iron oxide coatings 25.4 2.5 37 47 40 0.1 Keffeler (2014)
Decalcified limestone Carbon and iron oxide coatings 30.2 1.7 35 45 45 0.069 Keffeler (2014)
Decalcified limestone Carbon and non-plastic clay coating 30.2 1.3 34 44 30 0.067 Keffeler (2014)
Decalcified limestone Carbon coatings 30.2 1.1 40 49 40 1 Keffeler (2014)
Decalcified limestone Carbon and minor soft white infillings 30.2 1.4 41 49 40 0.83 Keffeler (2014)
Decalcified limestone Carbon and pyrite coatings 30.2 1.4 40 51 45 1.7 Keffeler (2014)
Decalcified limestone Carbon coatings and 1–3 mm calcite 30.2 0.71 47 51 50 3.3 Keffeler (2014)
Rock type Joint infilling material Plate Em (GPa) RQD RMR-76 RMR-89 GSI Q References
diameter
(cm)
Decalcified and argillized Clayey sand 30.2 0.9 20 28 15 0.02 Keffeler (2014)
limestone
Decalcified limestone Soft realgar coatings 30.2 4.2 45 54 50 0.67 Keffeler (2014)
Decalcified and argillized Soft realgar 1–3 mm 30.2 2.3 32 45 45 0.067 Keffeler (2014)
limestone
Siltstone and sandstone Calcite and iron oxide 30 84 59 66 Keffeler (2014)
Siltstone and sandstone Calcite and iron oxide 4.8 25 55 57 Keffeler (2014)
Sarvak limestone Calcite and iron oxide 0.971 14.1 5 44 50 Agharazi et al. (2012)
Sarvak limestone Calcite and iron oxide 0.971 16 14 43 52 Agharazi et al. (2012)
13
P. Asem, P. Gardoni
Table 14 Evaluation of Predictive method Database used for Number of μλ σλ δλ

empirical models for prediction evaluation cases
of rock mass modulus of
deformation (based on data Bieniawski (1978) Keffeler (2014) 5 1.35 1.04 0.77
from Keffeler 2014 and Asem
Serafim and Pereira (1983) Keffeler (2014) 30 0.81 0.74 0.91
2018)
Nicholson and Bieniawski (1990) Keffeler (2014) 30 0.83 0.79 0.95
Mitri et al. (1994) Keffeler (2014) 30 0.26 0.26 1.00
Read et al. (1999) Keffeler (2014) 30 0.52 0.45 0.86
Barton (2002) Keffeler (2014) 14 0.28 0.21 0.75
Diederichs and Kaiser (1999) Keffeler (2014) 23 0.58 0.49 0.83
Carvalho (2004) Keffeler (2014) 14 0.16 0.11 0.66
Palmstrom and Singh (2001) Keffeler (2014) 14 5.1 4.58 0.88
Gokceoglu and Zorlu (2004) RMR Keffeler (2014) 30 1.86 1.68 0.90
Gokceoglu and Zorlu (2004) GSI Keffeler (2014) 14 0.92 0.58 0.62
This study, Eq. (37) Asem (2018) 581 0.98 1.1 1.12
Fig. 25 Evaluation of normal

and lognormal distributions for
description of bias for method
of Zhang and Einstein (1998)
using tip resistance qf data (all
data from published literature,
see Asem 2018)
and that the use of more sophisticated models do not lead 7 Role of Load Test Data in Reliability‑Based
to more accurate estimates of the weak rock modulus of Design
deformation (Fig. 25).
Traditionally, designers have used a factor of safety (FS)
in the allowable stress design (ASD) approach to account
for the uncertainties in resistances and loads. The FS is
13
selected subjectively, and does not properly quantify the Equation (42) is now used in the FORM analysis to iterate
design uncertainties (Kulhawy et al. 2000; Roberts and about Rs until the calculated reliability index (β) (Ditlevsen
Misra 2009; Paikowsky et al. 2010). As an alternative to and Madsen 1996; Haldar and Mahadevan 2000; Gardoni
the ASD, designers have used the load and resistance fac- 2017) approaches the target reliability index (βT) that is con-
tor design (LRFD) framework to ensure the design is safe sistent with the acceptable level of failure probability that is
against each limit state (e.g., bearing capacity failure). The commonly used for design of the superstructure. Based on a
LRFD accounts for the uncertainty in loads and resistances review of the literature, βT is selected as follows:
separately. Additionally, the LRFD framework quantifies
the design uncertainties more rigorously using the theory 1. The redundancy and the possibility of load redistribution
of probability. (i.e., load sharing among foundation components) in the
In the design of foundations, two limit states should event of failure of one of the load bearing components in
be recognized (Meyerhof 1951; Allen 1975; Phoon et al. a system is an important criterion in determination of βT
2000): bearing capacity failure, and the foundation set- (Allen et al. 2005). Liu et al. (2001) defined a redundant
tlements. The evaluation of the strength limit state is dis- foundation system as one in which the calculated β value
cussed in the following sections to ensure a bearing capac- for the overall system is 0.5 higher than the reliability
ity failure will not materialize. In the context of the LRFD of the individual foundation members within the group.
framework, the strength limit state is determined using: Barker et al. (1991) determined that βT for drilled shafts
n ranges from 2.0 to 3.7 (from MVFOSM analysis), and
2.0 to 4.3 (from more advanced analysis). Barker et al.
∑
𝜙s Rs + 𝜙b Rb ≥ 𝛾i Pi , (41)
i=1 determined that βT is 3.5 for non-redundant systems,
2.5–3.0 for drilled shafts, and 2.0–2.5 for highly redun-
where ϕs and ϕb, respectively, are the LRFD resistance dant systems. Barker et al. (1991) relied heavily on the
factors for the side and base resistances, R s and Rb are, implied level of safety by previous design practice to
respectively, the total side and base resistances (determined provide their recommendations for the values of βT and
based on strain compatibility between side and base of rock the resulting resistance factors. Allen et al. (2005) pro-
socket), and the values of γi are the load factors for the ith posed that the ability of the soil to redistribute the load
axial load effects (Pi) at the top of drilled shaft, anchor, or allows for the foundation system to be designed for a
plug. lower reliability index compared to the superstructure.
The limit state function (g), which is required for the first- The only exception to this rule is the case of a single
order reliability method (FORM) analysis (Briaud et al. 2013; drilled shaft that supports the entire bridge pier where a
Gardoni 2017), is written as: βT value of 3.5 is desirable (Allen et al. 2005).
g = 𝜆R Rs − 𝜆DL DL − 𝜆LL LL, (42) 2. Phoon et al. (2000) recommended a βT of 3.2 for the
assessment of the strength limit state in drilled shafts.
where λR is the bias of the predictive model of interest, λDL Phoon et al. (2000) reported that the actual annual prob-
and λLL, respectively, are the bias for dead and live loads, abilities of failure for foundations is between 0.1 and
and DL and LL are, respectively, the dead and live loads. 1%. These values of probability of failure may be used
Following the recommendations of Paikowsky et al. (2004), to back-calculate βT of 2.5–3.0 for these foundations
Paikowsky et al. (2010), and Briaud et al. (2013), an HS-20 using the method of Rosenblueth and Esteva (1972),
design truck is used to represent the LL for typical bridge i.e., pf = 460 × e(−4.3𝛽).
piers (i.e., LL = 445 kN). The calibration of the LRFD resist- 3. Kulhawy and Phoon (2006) adopted the recommenda-
ance factors also requires an estimate of the dead load to live tions of the Canadian Building Code and the AASHTO
load ratio (DL/LL). The actual value of the DL/LL ratio is a bridge specifications. The Canadian Building Code rec-
function of the span length (l) of the structure. For example, ommends a βT of 3.5 for the superstructure and the foun-
McVay et al. (2000) provided the following values of DL/ dation and the AASHTO specifications recommends a
LL based on the values of l: DL/LL = 0.52, 1.06, 1.58, 2.12, βT of 3.5 for the superstructure and a βT of 2.0–3.5 for
2.64, 3.00, and 3.53, respectively, for l values of 9, 18, 27, the foundation.
36, 45, 50, and 60 m. These values are used in the calibra- 4. Nowak (1995) used a βT of 3.5 for the superstructure.
tion process. Evaluation of the models using load test data Allen et al. (2005) cited a βT of 3.5 for the superstruc-
show that a lognormal distribution is used to represent the ture and a βT of 3.0 for the foundation design from past
variation of the bias for predictive models. An example for practice, although Allen et al. (2005) indicates that it
fsp predictive model is shown in Fig. 17. Following Briaud is desirable to “maintain a consistent level of safety
et al. (2013), the statistics for λDL and λLL are also assumed across all limit states of a given type (e.g., strength limit
to follow a lognormal distribution. state).”
13
P. Asem, P. Gardoni
Fig. 26 Variation of load and

resistance factor design (LRFD)
resistance factors (ϕb) with the
bridge span length (l) and the
target reliability index (βT) for
side resistance
Fig. 27 Variation of load and

resistance factor design (LRFD)
resistance factors (ϕb) with the
bridge span length (l) and the
target reliability index (βT) for
base resistance
5. Ellingwood and Galambos (1982) recommended βT val- resistance factor based on the above recommendations and
ues of 2.5–4.0 for structural steel elements, and 2.3–3.6 engineering judgement. Therefore, the quantity Rs is cal-
for reinforced concrete elements in the building struc- culated for βT values of 2.0, 2.5, and 3.0, which fall within
tures. The βT values for foundations should be selected the range suggested in the literature. Once Rs is calculated
so as to maintain a degree of consistency between level from the FORM analyses, Eq. (41) is used to determine the
of safety of the foundation and the superstructure. values of ϕs for fsp and qf. The resistance factors obtained
from the FORM analysis are summarized in Figs. 26 and
According to the above discussion, a β T of 2.0–2.5 27 for side and base resistance.
should be selected when a possibility for load redistribu-
tion exists. If load sharing is not anticipated during an
extreme loading event, then a βT of approximately 3 should
be used. Therefore, it must be noted that βT does not take 8 Conclusions
a unique value and should be selected based on how the
structure would behave during an extreme event. Because The paper reviewed the application of load test databases in
a wide range of behaviors is possible (i.e., possibility of weak rock masses (i) to development of predictive models
load sharing or its absence based on structure type, design for peak side resistance, tip resistance and load-transfer func-
preferences), we are proposing a family of design curves tions for rock sockets in weak rock, (ii) models for prediction
that allows the designer to select the most appropriate of rock mass deformation modulus and (iii) to development
13
of reliability-based methods for design of drilled shafts in Supplementary Information The online version contains supplemen-
weak rocks. In summary, the following can be learned from tary material available at https://d oi.o rg/1 0.1 007/s 00603-0 21-0 2485-0.
the review of previous work conducted on axial behavior of
Acknowledgements Professor James H. Long of the University of
drilled shafts in weak rock and other geomaterials: Illinois at Urbana-Champaign contributed to this manuscript. His sug-
gestions and contributions are acknowledged.
1. The existing drilled shaft load test databases contain a lim-
ited number of drilled shaft load tests in weak rock masses. Author Contributions PA, and PG contributed to the writing of the
These databases commonly contain drilled shaft load tests manuscript and analyses produced therein. PA collected and analyzed
the data provided in the Supplemental Data Section.
in weak and strong rocks and have rarely been developed
for study of drilled shafts in “weak” rock masses. Moreo- Availability of Data and Materials Data are provided in the Supple-
ver, these databases commonly include drilled shaft load mental Data Section.
tests from one site and were often compiled with the pur-
pose of developing site-specific design models.
2. The drilled shaft load test results are inadequately docu-
mented in the existing databases for weak rocks. These References
databases commonly do not include the fs–δ and q–δ
relationships for their drilled shaft load tests. The rock Abu-Hejleh N, O’Neill MW, Hanneman D et al (2003) Improvement
profile and the drilled shaft geometry and the method of the geotechnical axial design methodology for Colorado’s
drilled shafts socketed in weak rocks. Colorado Department of
of construction are also not well documented. The con- Transportation, Denver
struction method and condition of the shear surface for Agharazi A, Tannant DD, Martin CD (2012) Characterizing rock mass
side of drilled shafts and tip of drilled shafts are not deformation mechanisms during plate load tests at the Bakhtiary
documented properly as well. dam project. Int J Rock Mech Min Sci 49(2012):1–11
AlKhafaji H, Imani M, Fahimifar A (2020) Ultimate bearing capac-
3. The secondary structure and the weathering condition ity of rock mass foundations subjected to seepage forces
of weak rock masses at drilled shaft load test sites in using modified Hoek-Brown criterion. Rock Mech Rock Eng
existing databases are not described in detail and are 53:251–268
not quantified. Data on rock “mass” properties are not Allen DE (1975) Limit states design—a probabilistic study. Can J Civ
Eng 2(1):36–49
often reported and in most cases, the only rock property Allen TM, Nowak AS, Bathurst RJ (2005) Calibration to determine
reported is the unconfined compressive strength (qu) of load and resistance factors for geotechnical and structural design.
intact rock. Washington, DC
4. In relation to the side resistance of sockets in weak rock: Ang AH-S, Tang WH (2007) Probability concepts in engineering:
emphasis on applications in civil and environmental engineer-
(i) the method of Seidel and Collingwood (2001) and ing, 2nd edn
Asem (2018) provide the best approach for prediction ARGEMA (1992) Design guides for offshore structures: offshore pile
of side resistance, (ii) the rock socket side wall rough- design. In: Tirant PL (ed) Association de recherché en geotech-
ness is related to the strength of rock, and (iii) the side nique marine, Paris
Asem P (2018) Axial behavior of drilled shafts in soft rock. Ph.D.
resistance is not strongly related to the initial geostatic Thesis. University of Illinois at Urbana-Champaign, Urbana
stresses. Asem P (2019a) Load-displacement response of drilled shaft tip in
5. In relation to the base resistance of sockets in weak rock: soft rocks of sedimentary origin. Soils Found 59(12):1193–1212
the base resistance of is governed by a combination of Asem P (2019b) Base resistance of drilled shafts in soft rock
using in situ load tests: a limit state approach. Soils Found
factors. Relating the base resistance to only qu or RQD 59(6):1639–1658
or similar method is often associated with modes that Asem P (2020) Prediction of unconfined compressive strength and
have large prediction uncertainties. deformation modulus of weak argillaceous rocks based on the
6. For rock sockets in weak rock, the method Williams standard penetration test. Int J Rock Mech Min Sci 133:104397
Asem P, Gardoni P (2019a) Evaluation of the peak side resistance
(1980) provides a promising approach for characterization for rock socketed shafts in weak sedimentary rock from an
of the load displacement response of the side and base. extensive database of published field load tests: a limit state
7. The resistance factors for sockets in weak rock vary approach. Can Geotech J 56(12):1816–1831
with (i) target reliability index and (ii) span length of Asem P, Gardoni P (2019b) A load-transfer function for the
side resistance of drilled shafts in soft rock. Soils Found
the structure the foundation would support. To properly 59(2019):1241–1259
define strength limit state, the base and side resistances Asem P, Gardoni P (2019c) Bayesian estimation of the normal and
should be strain compatible. shear stiffness for rock sockets in weak sedimentary rocks. Int
8. A large database of weak rock mass properties was com- J Rock Mech Min Sci 124(2019):104–129
Asem P, Long JH, Gardoni P (2018) Probabilistic model and LRFD
piled. Predictive models for prediction of Em based on resistance factors for the tip resistance of drilled shafts in soft
qu and GSI were proposed. sedimentary rock based on axial load tests. In: Innovations in
Geotechnical Engineering: Honoring Jean-Louis Briaud. X.
13
P. Asem, P. Gardoni
Zhang, P. J. Cosentino and M. H. Hussein, GSP 299. American Ellingwood B, Galambos TV (1982) Probability-based criteria for
Society of Civil Engineers, Florida, pp 1–49 structural design. Struct Saf 1(1):15–26
Asem P, Taukoor V, Kane T (2019a) Departure of soft rock mass Gardoni PE (2017) Risk and reliability analysis: theory and applica-
from undrained response in drilled shaft and plate load tests. tion. In: Risk and reliability analysis: theory and applications,
Soils Found 59(1):228–233 Urbana
Asem P, Jimenez R, Gardoni P (2019b) Probabilistic prediction Gardoni P, Kiureghian AD, Mosalam KM (2002) Probabilistic
of intact rock strength sing point load tests using a Bayesian capacity models and fragility estimates for reinforced concrete
formulation. In: 7th international symposium on geotechnical columns based on experimental observations. J Eng Mech
safety and risk (ISGSR 2019), Taipei, Taiwan 128(10):1024–1038
Asem P, Fuselier H, Labuz JF (2021) On a four-parameter linear fail- Gercek H (2007) Poisson’s ratio values for rocks. Int J Rock Mech Min
ure criterion. Rock Mech Rock Eng. https://doi.org/10.1007/ Sci 44(1):1–13
s00603-021-02451-w Glos III GH, Briggs JOH (1983). Rock sockets in soft rock. J Geotech
Baghdady AK (2018) Axial behavior of drilled shafts socketed into Eng 109(4): 525–535
weak Pennsylvanian shales. Ph.D. Thesis. University of Illinois Gokceoglu C, Zorlu K (2004) A fuzzy model to predict the uniaxial
at Urbana-Champaign, Urbana, Illinois compressive strength and the modulus of elasticity of a problem-
Bandis S (1980) Experimental studies of scale effects on shear atic rock. Eng Appl Artif Intell 17(1):61–72
strength and deformation of rock joints. Ph.D. Thesis. Univer- Goodman RE (1980) Introduction to rock mechanics. Wiley, New York
sity of Leeds, Leeds Goodman RE (1989) Introduction to rock mechanics. Wiley, New York
Bandis SC, Lumsden AC, Barton NR (1983) Fundamentals of rock Goodman RE (1993) Engineering geology: rock in engineering con-
joint deformation. Int J Rock Mech Min Sci Geomech Abstr struction, 1st edn. Wiley, New York
20(6):249–268 Griffith AA (1921) The phenomena of rupture and flow in solids. Philos
Barker RM, Duncan JM, Rojiani KB, Ooi PSK, Tan CK, Kim SG Trans R Soc Lond Ser A 221:582–593
(1991) Manuals for the design of bridge foundations (343). Guevara-Lopez F, Jimenez R, Gardoni P, Asem P (2019) Probabilistic
Washington, D.C prediction of intact rock strength using point load tests using a
Barton N (2002) Some new Q value correlations to assist in site Bayesian formulation. In: Georisk: assessment and management
characterisation and tunnel design. Int J Rock Mech Min Sci of risk for engineered systems and geohazards. https://doi.org/
39(2):185–216 10.1080/17499518.2019.1634274
Barton N et al (1978) Suggested methods for the quantitative descrip- Gupta RC (2012) Hyperbolic model for load tests on instrumented
tion of discontinuities in rock masses. Int J Rock Mech Min Sci drilled Shafts in intermediate geomaterials and rock. J Geotech
Geomech Abstr 15(6):319–368 Geoenviron Eng 138(11):1407–1414
Bieniawski ZT (1976) Rock mass classifications in rock engineering. Gupton C, Logan TJ (1984) Design guidelines for drilled shafts in
In: The symposium on exploration for rock engineering weak rocks of south Florida. In: South Florida annual ASCE
Bieniawski ZT (1978) Determining rock mass deformability: experi- meeting, Florida
ence from case histories. Int J Rock Mech Min Sci Geomech Haberfield CM (1987) The performance of the pressuremeter and
Abstr 15(5):237–247 socketed piles in weak rock. Ph.D. Thesis, Monash University,
Bishnoi BL (1968) Bearing capacity of a closed jointed rock. Ph.D. Monash University, Melbourne
Thesis. Georgia Institute of Technology, Atlanta Haberfield CM, Lochaden ALE (2018) Analysis and design of axially
Briaud J-L, Gardoni P, Yao C (2013) Statistical, risk, and reli- loaded piles in rock. J Rock Mech Geotech Eng 11(3):535–548
ability analyses of bridge scour. J Geotech Geoenviron Eng Haldar A, Mahadevan S (2000) Probability, reliability and statistical
140(2):04013011 methods in engineering design. Wiley, New York
Brown DA, Turner JP, Castelli RJ (2010) Drilled shafts: construc- Hassan KM (1994) Analysis and design of drilled shafts socketed into
tion procedures and LRFD design methods. National Highway soft rock. Ph.D. Thesis, University of Houston, Houston
Institute, U.S. Department of Transportation, Federal Highway Hassan KM et al (1997) Design method for drilled shafts in soft argil-
Administration, Washington, D.C., p 97 laceous rock. J Geotech Geoenviron Eng 123(3):272–280
Carter JP, Kulhawy FH (1988) Analysis and design of drilled shaft Hiltscher R, Carlsson A, Olsson T (1984) Determination of the defor-
foundations socketed into rock. Electric Power Research Insti- mation properties of bedrock under turbine foundations. Rock
tute, Palo Alto, p 190 Mech Rock Eng 17:37–49
Carvalho J (2004) Estimation of rock mass modulus. Pers commun Hirany A (1988) Test loading of drilled shafts. Ph.D. Thesis. Cornell
CGS (2006) Canadian foundation engineering manual University, Ithaca
Chern JC et al (2004) Correlation study on the deformation modulus Hirayama H (1990) Load-settlement analysis for bored piles using
and rating of rock mass (R-GT-97-04), from Taipei, Taiwan hyperbolic transfer functions. J Soils Found 30(1):55–64
Coates DF (1967) Rock mechanics principle. Department of Energy, Hoek E (1965) Rock fracture under static stress conditions. Ph.D. The-
Mines and Resources, Canada, p 410 sis, University of Cape Town, Cape Town
Collingwood B (2000) The effect of construction practices on the Hoek E (1983) Strength of jointed rock masses. Géotechnique
performance of rock socketed bored piles. Ph.D. Thesis, 33(3):187–223
Monash University, Melbourne Hoek E, Brown ET (1980) Empirical strength criterion for rock masses.
Dai G et al (2016) The effect of sidewall roughness on the shaft J Geotech Eng Div 106(9):1013–1035
resistance of rock-socketed piles. Acta Geotech 12(2):429–440 Hoek E, Brown ET (1997) Practical estimates of rock mass strength.
Deere DU, Miller RP (1966) Engineering classification and index Int J Rock Mech Min Sci 34(8):1165–1186
properties for intact rock. University of Illinois at Urbana- Hoek E, Brown ET (2019) The Hoek-Brown failure criterion and
Champaign/Air Force Weapons Laboratory, Urbana, p 327 GSI—2018 edition. J Rock Mech Geotech Eng 11:445–463
Diederichs MS, Kaiser PK (1999) Stability of large excavations in Hoek E, Diederichs MS (2006) Empirical estimation of rock mass
laminated hard rockmasses: the Voussoir analogue revisited. modulus. Int J Rock Mech Min Sci 43(2):203–215
Int J Rock Mech Min Sci 36(1):97–117 Hoek E, Carter TG, Diederichs MS (2013) Quantification of the geo-
Ditlevsen O, Madsen HO (1996) Structural reliability methods. logical strength index chart
Wiley, New York Holden JC (1984) Construction of board piles in weathered rocks. p 69
13
Horvath RG (1978) Field load test data on concrete-to-concrete rock Meigh AC, Wolski W (1979) Design parameters for weak rock. In: 7th
bond strength drilled pier foundations. University of Toronto, European conference on soil mechanics and foundation engineer-
Toronto, p 46 ing, Brighton, pp 59–79
Horvath RG (1982) Drilled piers socketed into weak shale-methods Mesri G, Shahien M (2003) Residual shear strength mobilized in first-
of improving performance. Ph.D. Thesis, University of Toronto, time slope failures. J Geotech Geoenviron Eng 129(1):12–31
Toronto Mesri G, Febres-Cordero E, Shields DR, Castro A (1981) Shear stress–
Horvath RG, Kenney TC (1979) Shaft resistance of rock socketed strain-time behaviour of clays. Géotechnique 31(4):537–552
drilled piers. In: Fuller FM (ed) Symposium on deep founda- Meyerhof GG (1951) The ultimate bearing capacity of foundations. Geo-
tions, Atlanta, pp 182–214 technique 2(4):301–332
Jaeger JC, Cook NG, Zimmerman RW (2007) Fundamentals of rock Miller AD (2003) Prediction of ultimate side shear for drilled shafts in
mechanics. Blackwell Publishing, Oxford Missouri shales. M.S. Thesis, University of Missouri-Columbia,
Jeong S et al (2010) Point bearing stiffness and strength of socket- Columbia
ted drilled shafts in Korean rocks. Int J Rock Mech Min Sci Mitri HS, Edrissi R, Henning J (1994) Finite element modeling of cable-
47(6):983–995 bolted stopes in hard rock ground mines. In: SME annual meeting,
Johnson IW (2020) Revisiting methods for the design of rock socketed New Mexico
piles. J Geotech Geoenviron Eng 146(12):04020144 Nicholson GA, Bieniawski ZT (1990) A nonlinear deformation modulus
Johnston IW, Lam TSK (1989) Shear behavior of regular triangular based on rock mass classification. Int J Min Geol Eng 8:181–202
concrete/rock joints-analysis. J Geotech Eng 115(5):711–727 Nowak AS (1995) Calibration of LRFD bridge code. J Struct Eng
Kaderabek TJ, Reynolds RT (1981) Miami limestone foundation design 121(8):1245–1251
and construction. J Geotech Eng Div 107(GT7):859–872 O’Neill MW, Reese LC (1999) Drilled shafts: construction procedures
Kanji MA (2014) Critical issues in soft rocks. J Rock Mech Geotech and design methods. Federal Highway Administration, Washing-
Eng 6(3):186–195 ton, D.C., p 758
Keffeler ER (2014) Measurement and prediction of in-situ weak rock Osterberg JO, Gill SA (1973) Load transfer mechanism for piers socket-
mass modulus: case studies from Nevada, Puerto Rico, and Iran. ted in hard soils or rock. In: The 9th Canadian rock mechanics
Ph.D. Thesis, University of Nevada, Reno symposium, Montreal, pp 235–262
Kodikara, J. K. (1989). Shear behaviour of rock-concrete joints and side Paikowsky SG, Birgisson B, McVay MC et al (2004) Load and resist-
resistance of piles in weak rock. Ph.D. Thesis, Monash Univer- ance factor design (LRFD) for deep foundations (NCHRP 507).
sity, Melbourne, Australia. Washington, D.C
Kulhawy FH, Carter JP (1992) Socketed foundations in rock masses. Paikowsky SG, Canniff MC, Lesny K, Kisse A, Amatya S, Muganga R
In: Bell FG (ed) Engineering in rock masses, Oxford, pp 509–529 (2010) LRFD design and construction of shallow foundations for
Kulhawy FH, Goodman RE (1980). Design of foundations on discon- highway bridge structures
tinuous rock. In: International conference on structural founda- Palmström A, Singh R (2001) The deformation modulus of rock masses
tions on rock, Sydney, pp 209–220 - comparisons between in situ tests and indirect estimates. Tunn
Kulhawy FH, Phoon KK (1993) Drilled shaft side resistance in clay Undergr Space Technol 16(2):115–131
soil to rock. In: Nelson PP, Smith TD, Clukey EC (eds) Design Peck RB (1976) Rock foundations for strucrures. In: Rock engineering
and performance of deep foundations: piles and piers in soil and for foundations and slopes, Boulder
soft rock, New York, pp 172–183 Peck RB, Hanson WE, Thornburn TH (1974) Foundation engineering,
Kulhawy FH, Phoon KK (2006) Some critical issues in geo-RBD cali- 2nd edn. Wiley, New York
brations for foundations. GeoCongress 2006:1–6 Pells PJN et al (1979) Socketed bored piles in Hawkesbury sandstone
Kulhawy FH, Phoon KK, Prakoso WA (2000) Uncertainty in basic Pells PJN, Turner RM (1979) Elastic solutions for the design and analysis
properties of geomaterials. In: Paper presented at the ISRM of rock-socketed piles. Can Geotech J 16(3):481–487
International Symposium, Melbourne Pells P, Bieniawski ZTR, Hencher S, Pells S (2017) RQD: time to rest in
Kulhawy FH, Prakoso WA, Akbas SO (2005) Evaluation of capacity peace. Can Geotech J 54(6):825–834
of rock foundation sockets. In: 40th U.S. symposium on rock Phoon K-K, Kulhawy FH, Grigoriu MD (2000) Reliability-based design
mechanics. American Rock Mechanics Association, Anchorage for transmission line structure foundations. Comput Geotech
Labuz JF, Zang A (2012) Mohr–Coulomb failure criterion. Rock Mech 26(3–4):169–185
Rock Eng 45:975–979 Priest SD, Hudson JA (1976) Discontinuity spacings in rock. Int J Rock
Labuz JF, Zeng F, Makhnenko R, Li Y (2018) Brittle failure of rock: a Mech Min Sci Geomech Abstr 13(5):135–148
review and general linear criterion. J Struct Geol 112(2018):7–28 Randolph MF, Wroth CP (1978) Analysis of deformation of vertically
Lam TSK (1983) Shear behaviour of concrete-rock joints. Ph.D. The- loaded piles. J Geotech Eng Div 104(GT12):1465–1488
sis. Monash University, Melbourne Read SAL, Richards LR, Perrin ND (1999) Applicability of the Hoek–
Lee J, Jeong S (2016) Experimental study of estimating the subgrade Brown failure criterion to New Zealand greywacke rocks. In: 9th
reaction modulus on jointed rock foundations. Rock Mech Rock international congress on rock mechanics, Paris
Eng 49:2055–2064 Roberts LA, Misra A (2009) load and resistance factor design (LRFD)
Lee J et al (2013) Proposed point bearing load transfer function in jointed of deep foundations using a performance-based design approach.
rock-socketed drilled shafts. Soils Found 53(4):596–606 J GeoEng 4(3):87–92
Liu WD, Neuenhoffer A, Ghosn M, Moses F (2001) Redundancy in high- Rosenberg P, Journeaux NL (1976) Friction and end bearing tests on bed-
way bridge substructures (NCHRP 458). Washington, D.C. rock for high capacity socket design. Can Geotech J 13(3):324–333
Marinos P, Hoek E (2001) Estimating the geotechnical properties of het- Rosenblueth E, Esteva L (1972) Reliability basis for some Mexican codes.
erogeneous rock masses such as Flysch. Bull Eng Geol Environ Special Publication, p 31
60(2):85–92 Rowe RK, Armitage HH (1984) The design of piles socketed into weak
McVay M, Birgisson B, Zhang L, Perez A, Putcha S (2000) Load and rock. National Research Council Canada, Ottawa, p 380
resistance factor design (LRFD) for driven piles using dynamic Rowe RK, Armitage HH (1987) A design method for drilled piers in soft
methods—a Florida perspective. Geotech Test J 23(1):55–66 rock. Can Geotech J 24(1):126–142
Seed HB, Reese LC (1957) The action of soft clay along friction piles.
ASCE Trans 122(2882):731–754
13
P. Asem, P. Gardoni
Seidel JP (1993) Analysis and design of pile shafts in weak rock. Ph.D. Toh CT, Ooi TA, Chiu HK, Chee SK, Ting WH (1989) Design param-
Thesis, Monash University, Melbourne eters for bored piles in weathered sedimentary formation. In: 12th
Seidel JP, Collingwood B (2001) A new socket roughness factor international conference on soil mechanics and foundation engi-
for prediction of rock socket shaft resistance. Can Geotech J neering, Rio de Janeiro, pp 1073–1078
38(1):138–153 Turner J (2006) Rock-socketed shafts for highway structure foundations.
Seo H, Prezzi M (2008) Use of micropiles for foundations of transporta- Project 20-5, Washington
tion structures (FHWA/IN/JTRP-2008/18). West Lafayette Vesic AS (1963) Bearing capacity of deep foundations in sand. Highway
Serafim, J. L., and Pereira, J. P. (1983). Consideration of the geomechani- Res Rec 39:112–153
cal classification of Bieniawski. In Symposium on Engineering Vu TT (2013) Load and resistance factor design of drilled
Geology and Underground Constructions. shafts at the service limit state. Ph.D. thesis, University of
Sowers GF (1979) Introductory soil mechanics and foundations. Macmil- Missouri-Columbia.
lan Publishing Co. Inc., New York, p 621 Williams AF (1980) The design and performance of piles into weak rock.
Stark TD, Long JH, Asem P (2013) Improvements for determining the Ph.D. Thesis, Monash University, Melbourne
axial capacity of drilled shafts in shale in Illinois. Illinois Depart- Williams AF, Pells PJN (1981) Side resistance rock sockets in sandstone,
ment of Transportation, Urbana mudstone, and shale. Can Geotech J 18(4):502–513
Stark TD, Long JH, Baghdady AK, Osouli A (2017). Modified standard Yoshinaka R, Yamabe T (1986) 3. Joint stiffness and the deforma-
penetration test-based drilled shaft design method for weak rocks tion behaviour of discontinuous rock. Int J Rock Mech Min Sci
(Phase 2 study). Illinois Center for Transportation, Urbana Geomech Abstr 23(1):19–28
Tabandeh A, Asem P, Gardoni P (2020) Physics-based probabilistic mod- Zhang L (1999) Analysis and design of drilled shafts in rock. Ph.D. The-
els: Integrating differential equations and observational data. Struct sis, Massachusetts Institute of Technology, Massachusetts
Saf 87:101981 Zhang L, Einstein HH (1998) End bearing capacity of drilled shafts in
Tang C, Phoon KK, Li D-Q, Akbas SO (2020) Expanded database rock. J Geotech Geoenviron Eng 124(7):574–584
assessment of design methods for spread foundations under
axial compression and uplift loading. J Geotech Geoenviron Eng Publisher’s Note Springer Nature remains neutral with regard to
146(11):04020119 jurisdictional claims in published maps and institutional affiliations.
Teng WC (1962) Foundation design. Prentice Hall, Englewood Cliffs
Terzaghi K (1943) Theoretical soil mechanics. Wiley, New York
Terzaghi K, Peck RB, Mesri G (1996) Soil mechanics in engineering
practice. Wiley-Interscience, New York
13

Asem 2021

Uploaded by

Copyright:

Available Formats

Asem 2021

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Asem 2021

Uploaded by

Copyright:

Available Formats

Rock Mechanics and Rock Engineering

On the Use and Interpretation of In Situ Load Tests in Weak Rock

Received: 14 November 2020 / Accepted: 22 April 2021

Ksp Empirical factor accounting for joint spac- δλ C.O.V. of bias

FHWA deep founda- Limited United States Sedimentary rocks No No

marl and limestone

claystone, mudstone, limestone, sandstone, tuff, and granite.

Side and tip resistances

Side and tip resistances

Miller 1966; Barton et al. 1978; Rowe and Armitage 1987;

(i) the load-deformation properties for the rock sockets, and

Em). Additionally, the load test databases only concentrate

No on the axial behavior of rock sockets in weak rocks. The

The side-resistance database includes 317 drilled shaft load

displacement (fs–δ) relationship for each socket sidewall is

Shale, mudstone, clay- 0.12–4.45

obtain the fs–δ relationships. The side-resistance database

contains the following information:

1. Load-transfer function (fs–δ relationship) for rock socket

sidewalls: The fs–δ relationships are obtained based on

the measured drilled shaft load–displacement relation-

ments that are gathered using strain gauges or load cells

(see Fig. 1). The displacements reported( in fs)–δ relation-

United States, South

United States, South

the axial load, L is the rock socket length, A is socket

cross-sectional area and Ec is Young’s modulus of con-

strength (qu) of the intact rock specimens is reported for

93 (side) and 62 (tip)

were not available for majority of case histories.

based on theory of elasticity for calculation of socket

settlement, based on the numerical integration of

Stark et al. (2013)

Top load Ground

Top of soft Top of soft Top of soft

Test shaft Reaction Reaction

Base load Void or

Method (a) Method (b) Method (c)

Reaction Loading Reaction

Top of soft rock

DGS DTOR Test shaft Top of soft rock

Fig. 1 Load test methods used in the in situ test database

Fig. 2 Typical side resistance

could not be investigated using the database of Horvath and

Fig. 4 Predictive model for

Fig. 5 Evaluation of the model

where n is the number of observations in each database.

Fig. 6 Design method of

M′/P′ ratio about its mean which is indicative of the model

The data presented in Tables 3 and 4 suggest that most of

In weak rock masses, a significant portion of the axial load

Design methods Databases Design Database (Asem 2018)

Fig. 9 Variation of the available

Fig. 10 Variation of the avail-

Table 5 Summary of empirical base-resistance models

mobilized along the entire failure surface at the same time.

Fig. 11 Evaluation of empiri-

that the surcharge loads representing the embedment depth

where q is the lower bound estimate of base resistance for

Ksp Empirical factor accounting for joint spac- δλ C.O.V. of bias

6.2 Rock Mass Indices and Rock Mass Deformation