KK and Chong Tang

Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

Georisk: Assessment and Management of Risk for

Engineered Systems and Geohazards

ISSN: 1749-9518 (Print) 1749-9526 (Online) Journal homepage: https://www.tandfonline.com/loi/ngrk20

Characterisation of geotechnical model


uncertainty

Kok-Kwang Phoon & Chong Tang

To cite this article: Kok-Kwang Phoon & Chong Tang (2019): Characterisation of geotechnical
model uncertainty, Georisk: Assessment and Management of Risk for Engineered Systems and
Geohazards, DOI: 10.1080/17499518.2019.1585545

To link to this article: https://doi.org/10.1080/17499518.2019.1585545

View supplementary material

Published online: 01 Mar 2019.

Submit your article to this journal

Article views: 15

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=ngrk20
GEORISK
https://doi.org/10.1080/17499518.2019.1585545

Characterisation of geotechnical model uncertainty


Kok-Kwang Phoon and Chong Tang
Department of Civil and Environmental Engineering, National University of Singapore, Singapore, Singapore

ABSTRACT ARTICLE HISTORY


The calculated response from a numerical model will deviate from the measured one given the Received 30 October 2018
presence of modelling idealizations and real world construction effects. This deviation can be Accepted 18 February 2019
directly captured by a ratio between the measured and the calculated quantity. The ratio is also
KEYWORDS
called a model factor in many design guides. The probabilistic distribution of the model factor is model uncertainty; load test
arguably the most common and simplest complete representation of model uncertainty. The database; reliability-based
characterisation of model uncertainty is identified as one of the critical elements in a design; model factor
geotechnical reliability-based design process in Annex D of ISO 2394:2015 “General Principles on
Reliability of Structures”. This Spotlight paper reviews the databases for various geo-structures
and determines their associated model statistics. Foundation load test databases are the most
prevalent. A recent effort to compile a large generic database (PILE/2739) that contains 2739
field load tests conducted on various piles and installed in different soils and countries, is
highlighted. This systematic compilation of load test data is part of a broader research agenda to
digitalise foundation design for “precision construction”, which is targeted at characterising “site-
specific” model factors and soil parameters based on both site-specific and generic data for
further customisation of design to a particular site. The mean and COV of the model factor for a
range of geo-structures, geomaterials, and limit states (both ultimate and serviceability) are
summarized in a form suitable for adoption in design and codes of practice. Based on this
summary, it is proposed that a model factor for a design model can be classified as:
(1) moderately conservative (1 ≤ mean < 2), (2) highly conservative (2 ≤ mean < 3), or (3) very
highly conservative (mean ≥ 3). The model uncertainty can be as: (1) low dispersion (COV < 0.3),
(2) medium dispersion (0.3 ≤ COV < 0.6), (3) high dispersion (0.6 ≤ COV < 0.9), and (4) very high
dispersion (COV ≥ 0.9). This summary represents the most extensive and significant update of
Table 3.7.5.1 in the 2006 JCSS Probabilistic Model Code.

Introduction frequently considered as a serviceability limit state (SLS).


ULS and SLS are commonly calibrated within the frame-
For practical convenience and because of the historical work of limit state design (LSD) (Becker 1996). Bolton
development of the mechanics of deformable solids, (1981) discussed a more comprehensive range of limit
the problems in geotechnical engineering are often cate- states: (1) unserviceability through soil strain, (2) unser-
gorised into two distinct groups, namely elasticity and viceability through concrete deformation, (3) collapse of
stability (Chen 1975). The elasticity problems deal with structure through soil failure alone, (4) collapse of struc-
stress or deformation of soil without failure. We use ture with both soil and concrete failure, and (5) collapse
the term “elasticity” in the sense described by Chen of structure arising without soil failure. Vardanega and
(1975), which covers all elasto-plastic problems prior Bolton (2016) cited Burland, Broms, and DeMello
to ultimate failure. Point stresses beneath a footing or (1977) to describe a category of damage as “disappoint-
behind an earth retaining wall, deformations around ing” in that “later develop into serviceability issues, and
tunnels or excavations, and all settlement problems then ultimately threaten structural collapse only if noth-
belong in this category (Chen 1975). The stability pro- ing has been done to interrupt the loading process or
blems are associated with the determination of a load enhance the soil-foundation system”. Phoon (2017)
that will cause the failure of soil in modes such as bearing noted that demarcation between ULS and SLS is intro-
capacity, passive and active earth pressure, and stability duced by LSD, which is distinct from reliability-based
of slopes. In many design codes, stability is usually con- design (RBD). RBD only requires a performance func-
sidered as an ultimate limit state (ULS), while elasticity is tion that is calibrated by load tests (the focus of this

CONTACT Chong Tang [email protected] Department of Civil and Environmental Engineering, National University of Singapore, Block E1A, #07-03,
1Engineering Drive 2, Singapore 117576, Singapore
Supplemental data for this article can be accessed at https://doi.org/10.1080/17499518.2019.1585545.
© 2019 Informa UK Limited, trading as Taylor & Francis Group
2 K.-K. PHOON AND C. TANG

paper) and it can handle ULS, SLS, or any intermediate 1997-1:2004 (CEN 2004). On the other hand, Table 6.2
damage limit states discussed elsewhere. in the most recent edition of the Canadian Highway
In general, it is difficult to exactly predict the behav- Bridge Design Code – CAN/CSA-S6-14 (Canadian Stan-
iour of a geo-structure due to complicated soil-structure dards Association 2014) recommended the ULS and SLS
interaction (including interface behaviour), complex resistance factors for various geo-structures to be
geological conditions, and construction effects. To selected based on the degree of “site and prediction
solve these problems analytically, various empirical or model understanding”. CAN/CSA-S6-14 applies the
semi-empirical methods were proposed, which usually semi-probabilistic Load and Resistance Factor Design
entail making conservative assumptions and physical (LRFD), which is the most popular simplified RBD for-
and/or geometric simplifications (Dithinde et al. 2016). mat in North America. The state DOTs (Departments
For example, solutions to the stability problems are gen- of Transportation) in the United States are increasingly
erally obtained by the theory of perfect plasticity, which calibrating the LRFD resistance factors for pile foun-
ignores work softening (or hardening) and assumes a dations using rigorous reliability analyses (Seo et al.
continuing plastic flow at constant stress (Chen 1975). 2015). The second approach is to introduce a random
The perfect-plastic assumption differs from the reality, variable to characterise the model factor that is the
where soil often exhibits some degree of work softening focus of this Spotlight paper. This is the basis of direct
(or hardening). Solutions to the elasticity problems are probability-based design methods. Section 2.2.3.5 in
commonly solved with the theory of linear elasticity. the Recommended Practice DNVGL-RP-C207 (DNV
Nevertheless, soil behaviour is highly non-linear even 2017) stated that “model uncertainty involves two
at small strain, which has an important influence on elements, viz.: (1) a bias if the model systematically
the selection of geotechnical parameters in routine leads to overprediction or underprediction of a quantity
design (Atkinson 2000). Construction disturbance is in question and (2) a randomness associated with the
hard to model. As a consequence, the predicted value variability in the predictions from one prediction of
(failure load, stress, or deformation) will deviate from that quantity to another”. Table 3.7.5.1 in the JCSS Prob-
the measured value (typically on the safe side). This devi- abilistic Model Code (JCSS 2006) presented some indica-
ation can be directly captured by a ratio of the measured tive values for the mean and coefficient of variation
value to the predicted value. The ratio is also known as a (COV) of the model factors for several geo-structures
model factor. The probabilistic distribution of the model (e.g. embankments, sheet pile walls, shallow foundations
factor is arguably the most common and simplest com- and driven piles). Phoon (2017) opined that the design
plete representation of model uncertainty. Different point corresponding to the model factor in the First-
design scenarios will produce different values of the Order Reliability Method (FORM) can bridge the deter-
model factor, even for the same geo-structure and the ministic and random approaches, i.e. it can be a rational
same design (calculation) model because detailed site choice for the model partial factor described in Figure C3
and construction effects cannot be captured. There is of EN 1990:2002 (CEN 2002).
overwhelming empirical evidence that assigning a single Furthermore, Annex D of the fourth edition of ISO
value to the model factor is grossly incomplete. The 2394 (ISO 2015) suggests that the characterisation of
model uncertainty is regarded as depending on the model uncertainty is one of the critical elements in a geo-
degree of knowledge of the simplifications, assumptions, technical RBD process (Dithinde et al. 2016; Phoon et al.
and approximations made in the design model (DNV 2016), which is closely related to risk-informed decision
2017). The need for assessment of model uncertainty making (Nadim 2017). Another critical element in the
has been emphasized and considered in the current revi- geotechnical RBD process is the characterisation of geo-
sion process of the Eurocodes (Lesny 2017a) and some technical variability (Phoon et al. 2016). The key features
design guidelines (JCSS 2006; DNV 2017). Some discus- are: (1) COV of a geotechnical parameter that does not
sions on the impact of model uncertainty on design can take a unique value due to different estimation approaches
be found in Bauduin (2003), Phoon (2005), Forrest and including the application of different transformation
Orr (2011), Teixeira et al. (2012), Burlon et al. (2014), models (Phoon and Kulhawy 1999) and (2) the multi-
Lesny (2017a), Haque and Abu-Farsakh (2018). variate correlation of soil/rock data can be exploited to
To provide some context on how model uncertainty is reduce the transformation uncertainty for a desired geo-
considered/applied in practice, several design codes/ technical parameter (Ching, Phoon, and Chen 2010a;
guidelines are reviewed. Broadly, two review approaches Ching and Phoon 2012). A recent review on characteris-
exist. The first is to introduce a deterministic partial fac- ation of geotechnical variabilities within a realistic
tor to ensure that the design calculation is sufficiently “MUSIC” (Multivariate, Uncertain and unique, Sparse,
safe as given in Annex A of Eurocode 7 – that is, EN and InComplete as coined in Phoon 2018) data context
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 3

is presented in Phoon and Ching (2018). ISO 2394:2015 is shaft resistance factors α and β) that are calibrated against
intended to be used as “code for code drafters” in load test data (typically limited). Over the past 30 years,
countries where the principles of risk and reliability are one of the major advances in pile design is the identifi-
used to design and assess the structures over the entire ser- cation of factors governing pile behaviour aided by
vice life (Phoon et al. 2016). Because probabilistic methods high-quality instrumented tests – model tests and field
and reliability theory which are sensitive to measured data tests. Examples of factors that affect piles driven into
can ensure a consistent level of safety and consider all sand include the extent of soil disturbance during pile
uncertainties in a more rational way (Phoon 2017), they installation and loading, shaft friction fatigue, pile ageing,
have been increasingly discussed and applied in geotech- soil plug in pile installation, and characteristics of the soil-
nical engineering (Casagrande 1965; Tang 1979; Whitman shaft interface (Jardine 1985; Lehane 1992; Chow 1997;
1984; Wu et al. 1989; Duncan 2000; Baecher and Christian Gavin 1998; Randolph 2003; White 2005; Flynn 2014).
2003; Christian 2004; Griffiths and Fenton 2007; Fenton On this basis, four methods were developed, which
and Griffiths 2008; Kulhawy, Phoon, and Wang 2012; directly relate pile capacity to cone penetration test
Phoon and Ching 2014; Fenton et al. 2016; Phoon and (CPT) data (Clausen, Aas, and Karlsrud 2005; Jardine
Retief 2016; Becker 2017; Nadim 2017; Phoon 2017). et al. 2005; Kolk, Baaijens, and Senders 2005; Lehane,
The Chinese Academy of Engineering (CAE) in collabor- Schneider, and Xu 2005). Because these factors were
ation with Clarivate Analytics identified “reliability of civil taken into consideration rationally, they provide better
engineering structures” as one of the ten engineering estimates of pile capacity in sand compared to traditional
research hotspots in civil, hydraulic engineering and design approaches based on the conventional lateral earth
architecture (Chinese Academy of Engineering 2017). pressure theory (Schneider, Xu, and Lehane 2008; Yang
The primary objective of this Spotlight paper is to et al. 2016). Other examples are the development of
review the characterisation of model factors in geotechni- improved calculation methods for pile capacity (Burlon
cal engineering to summarize the body of knowledge in a et al. 2014) based on pressuremeter test results, which
form suitable for adoption in design and codes of practice were calibrated against 174 full-scale static load tests
and to identify gaps/limitations for future research. This from the IFSTTAR database.
paper is organised as follows: (1) role of databases and A more recent development is the adoption of
model calibration in design; (2) representation of model reliability-based design (RBD), which is more sensitive
uncertainty for simple and complex design situations to data or information. Phoon (2017) opined that “the
and the underlying limitations; (3) extensive survey of advantage of RBD is that it can respond to a change of
databases for shallow foundations, pile foundations, COV in a rational way explicitly related to data, while
offshore spudcans, pipes and anchors, mechanically stabil- the factor of safety approach cannot”. Regardless of the
ised earth structures, soil slopes, and braced excavations; debate on the strengths and weaknesses of RBD, there
(4) statistical interference of model factors using frequen- is no doubt that any approach not directly linked to
tist and Bayesian approaches; (5) characterisation of the data cannot take advantage of the ongoing develop-
mean (bias) and COV (dispersion) of model factors for ments in Industry 4.0. An improved understanding
common resistance and deformation calculation methods of the ground and/or the structure can directly lead
recommended by design codes/guidelines; and (6) practi- to design savings within such a design framework. Fen-
cal classification of model uncertainty into broad groups ton et al. (2016) noted that “there is a real desire
according to the mean and COV of the model factor. amongst the geotechnical community to have their
The results presented in this paper constitute a significant designs reflect the degree of their site and modelling
update of Table 3.7.5.1 in the JCSS Probabilistic Model understanding”. Site understanding refers to “how
Code (JCSS 2006). well the ground providing the geotechnical resistance
is known” and model understanding refers to “the
degree of confidence that a designer has in the
Databases and model calibration
model used to predict the geotechnical resistance”. As
Load test data always play a key role in the development of suggested in CAN/CSA-S6-14, engineer can choose
geotechnical engineering: (1) to improve our understand- the ULS and SLS resistance factors with their site
ing of soil-structure interaction, (2) to propose more and model understanding flexibly. Phoon et al.
rational design models, and (3) to verify and calibrate a (2016) opined that
proposed method. One well-known example is the predic-
it is unrealistic to recommend a single value for each
tion of pile axial capacity. Most of the calculation methods resistance factor that is unrelated to the quality of the
were founded on the basic principles of soil mechanics in information at hand, including site data, performance
conjunction with empirical parameters (e.g. dimensionless of the prediction model adopted, type of calibration
4 K.-K. PHOON AND C. TANG

tests (small-scale, centrifuge, or prototype), quality con- of a pile or settlement of a shallow foundation under ver-
trol, monitoring data, etc. tical loading. It is defined as the ratio of the measured
The full value of a database is arguably best realised response to the calculated response (i.e. Eq. D.1 in
within a direct probability-based design method (Wang Annex D of ISO 2394:2015):
et al. 2016).
Finally, it is noteworthy that industries worldwide are M = Xm /Xc (1)
undergoing a rapid transformation as a result of increas-
ing convergence between cyber and physical systems. where M is the model factor, Xm is the measured
There is an explosion of interest on data and how to response, and Xc is the calculated response. This method
extract the best value from data for decision making. is practical, familiar to engineers, and grounded realisti-
One impetus is the introduction of new and powerful cally on a load test database. At the design stage, it is
ways to analyze big data that can reveal the underlying obviously relevant to consider the model factor associ-
pattern, trend, and association. Clearly, there is a press- ated with a calculation method given there is always a
ing need for the geotechnical community to engage in gap between the calculated value and the measured
this transformation. The International Society for Soil (actual) value.
Mechanics and Geotechnical Engineering (ISSMGE) Equation (1) has been widely used to evaluate the
recently established a new technical committee TC309 model uncertainty in (1) stability problems (e.g. capacity
on “Machine Learning” to hasten research and appli- of foundations, pipelines, anchors and mechanically
cations in this direction. ISSMGE TC304 (Engineering stabilised earth structures or factor of safety for soil
Practice of Risk Assessment and Management) also slopes) and (2) elasticity problems (e.g. settlement of
launched a database sharing initiative called 304dB foundations, wall and ground movements in braced
(http://140.112.12.21/issmge/tc304.html). The Insti- excavations). The model factor itself is not constant,
tution of Civil Engineers (2017) urged the industry to but takes a range of values that may depend on the geo-
“think about not only the physical asset, but also its digi- metrical and geological scenarios covered in a load test
tal twin – all the associated data and the information that database. It is customary to model M as a random vari-
this can reveal” and opined that “if we truly consider able. The simplest method to characterise a random vari-
infrastructure as a service, then making this mental able is to calculate the mean (bias) and COV
shift is essential. Delivering infrastructure based on out- (dispersion). A mean and COV of M close to 1 and 0,
comes for users drives us toward whole life decisions and respectively, would represent a near perfect calculation
recognising the value of the entire data estate”. Digital method that matches measured responses for all scen-
approaches are already changing the role of an engineer arios in the calibration database. It goes without saying
in post-construction asset management (Pathmananda- that such methods do not exist in geotechnical engineer-
vel 2018). In foundation design, Pathmanandavel ing, regardless of their numerical sophistication. More
(2018) emphasized that digital solutions should be sophisticated calculation methods typically require
explored to advance and add value for clients and com- more input parameters and some of these parameters
munities. In the opinion of the authors, the compilation are not measured in routine site investigations. Lambe
of databases in geotechnical engineering has become (1973) argued that the quality of prediction does not
more pressing and more critical when viewed within increase with the quality of the method regardless of
this broader context. There are undoubtedly strong the quality of data. The quality of prediction is optimal
interests from both academia and industry to re-examine only for a particular combination of method and data.
how all forms of geotechnical data (soil data, load test The mean of M would provide an engineer with a
data, monitoring data, and new data sources) can further sense of the hidden factor of safety that either adds or
support decision making and add value to engineering subtracts from the nominal global factor of safety,
service using probabilistic and deep learning methods. depending on whether the calculation method is conser-
vative (mean > 1) or unconservative (mean < 1) in the
average sense. It should not be inferred that a calculation
Representation of model uncertainty method is conservative or otherwise for a specific case
because M takes a range of values in actuality (hence it
Assessment of model uncertainty by a model
is random). This random nature is practically significant,
factor
because it implies that a calculation method can be
For simple limit states involving a single response, the unconservative when applied to a specific case even
model uncertainty can be characterised in a natural though the method is conservative on the average. There-
way by a model factor. Examples would include capacity fore, it is also necessary to consider the degree of scatter
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 5

(dispersion) in M and to ensure probability of a statistics to cater to different permissible movements)


measured value being lower than the calculated value is would be tedious if not impossible to apply when the per-
capped at a known value say p% (Lesny 2017a). This missible movement is a random variable rather than a
idea is conceptually similar to EN 1997-1:2004, 2.4.5.2 constant. An alternate approach is the parameterisation
(11) which recommends a cautious estimation (or of measured load-movement curves to handle SLS is a
characteristic value) for a geotechnical design parameter more realistic way (Phoon and Kulhawy 2008). This
can be “derived such that the calculated probability of a involves fitting the measured curve to a simple empirical
worse value governing the occurrence of the limit state model, thus replacing the continuous curve (a function)
under consideration is not greater than 5%.” As dis- by two correlated model factors. It is applicable to all the
cussed previously, it is possible to handle a random foundation types studied so far. Popular models include
model factor more rationally in a partial factor design the hyperbolic model in Eqs. (2) and (3) and the power
format by selecting the FORM design point for the law in Eq. (4), where the normalisation of load by the
model partial factor (Phoon 2017). measured capacity is employed to reduce the scatter in
Lesny (2017a) stated that with field load test data, the the measured curves:
model uncertainty characterised by Eq. (1) is intrinsically
Q s
tied to: (1) inherent variability of soil parameters (esti- = (2)
mated by theoretical formulae, indirect correlations, or Rum a + bs
direct measurements) used within the design model;
Q (s/B)
(2) measurement errors in soil investigations and load = (3)
Rum a + b(s/B)
tests; and (3) definition and determination of Xm from
load test results. Q  s b
=a× (4)
Rum B
Parameterisation of load-movement curves for
where Q is applied load, s is movement, a and b are load-
SLS model uncertainty
movement model factors determined by least-squares
Equation (1) is applicable for settlement predictions as regression fitting to measured curves, and Rum is an inter-
presented in Muganga (2008), Zhang, Xu, and Tang preted failure load from the measured load-movement
(2008), Gurbuz (2007), Zhang and Chu (2009b), and curve with a proper definition. To be “proper”, a definition
Abchir et al. (2016) for a specified movement or working of capacity must be based on some mathematical rule and
load level. Nevertheless, Phoon and Kulhawy (2008) sta- with due consideration of the shape of the load-move-
ted that Eq. (1) can be problematic even for the relatively ment curve (e.g. Davisson offset limit, Brinch-Hansen
simple SLS. Although it is possible to replace the capacity 80% criterion, or Chin’s method) (Fellenius 2018).
by a well-defined permissible load that depends on the The bivariate load-movement model is an improve-
permissible movement (Phoon, Kulhawy, and Grigoriu ment to Eq. (1), because a and b allows the entire load-
1995; Paikowsky and Lu 2006), the chief drawback is movement curve to be simulated, rather than one fixed
that the distribution of this SLS model factor has to be “working load” point on the load-movement curve. For
re-evaluated when a different permissible movement is a prescribed permissible movement sa, the resulting per-
prescribed. It is tempting to argue that the load-move- missible load Qa can be easily obtained from the bivariate
ment behaviour is linear at the SLS, and subsequently load-movement model in Eqs. (2) – (4). The probabilistic
the distribution of the model factor for a given permiss- distribution of Qa can be derived from the bivariate dis-
ible movement can be extrapolated to other permissible tribution of the load-movement model factors a and b
movements by simple scaling. However, for reliability (Phoon and Kulhawy 2008; Uzielli and Mayne 2011;
analysis, the applied load follows a probability distri- Stuedlein and Reddy 2013; Huffman and Stuedlein
bution, and it is possible for the load-movement behav- 2014; Najjar, Shammas, and Saad 2014, 2017; Stuedlein
iour to be nonlinear at the upper tail of the distribution and Uzielli 2014; Huffman, Strahler, and Stuedlein
(corresponding to high loads). Moreover, it may be more 2015, 2016; Reddy and Stuedlein 2017b; Tang and
realistic to model the permissible movement as a random Phoon 2018c; Chahbaz, Sadek, and Najjar 2019). In the
variable (Zhang and Ng 2005), given that it is affected by absence of site-specific load tests, the measured capacity
many interacting factors, such as the type and size of the Rum in the bivariate load-movement model could be
structure, the properties of the structural materials and replaced with the calculated capacity Ruc. In this situ-
the underlying soils, and the rate and uniformity of the ation, the uncertainty in capacity calculation should
movement. The standard definition of a model factor also be incorporated into design through the model fac-
in the form of Eq. (1) (a table of SLS model factor tor M in Eq. (1).
6 K.-K. PHOON AND C. TANG

The bivariate load-movement models in Eqs. (2) – (4) Woeller (2014), and Mayne and Dasenbrock (2018) cali-
are not sufficiently general to handle multiple and typi- brated a direct design method based on CPT against a set
cally correlated limit states. It is possible to consider of 130 full-scale field load tests on clay, sand, and silt.
improving the above simple approach, such as normali- It suffices to say that more research is needed to
sation using a quantity related to the soil-pile stiffness. In generalize Eq. (1) so that the model uncertainties
this way, other relevant parameters pertinent to the associated with more complex soil-structure interaction
deformation mechanism such as pile stiffness, soil stiff- problems can be characterised at the system level
ness, and pile-soil interaction behaviour can be included, where multiple limit states involving multiple
but this direction of research can be practically con- responses are operational. The Bayesian approach
strained by the range of soil parameters measured during briefly described below is a possible solution. Varda-
the load test. nega and Bolton (2016) noted that the “great majority
of RBD approaches focus on ultimate failure, whereas
the onset of excessive deformations leading to struc-
Model uncertainty evaluation for more complex
tural unserviceability is widely accepted to be a more
design situations
critical issue in foundation engineering”. This may
For more complex design scenarios, design models or arise in part from this research gap. Phoon (2017)
calculation methods such as the finite element method clarified that reliability theory is sufficiently general
(FEM) can produce a vector of responses that vary to accommodate geotechnical needs. The primary gap
with time or loading stages. It is difficult to use Eq. (1) between research and practice is introducing greater
for model uncertainty evaluation in these scenarios realism into reliability analysis that responds to current
(Lesny 2017a). needs in practice rather than making further improve-
One example is the design of braced excavation. The ments in theory.
maximum deflection and bending moment should be
considered, which do not occur at a fixed depth and
Survey of databases
change with the excavation stage. Lesny (2017a) pointed
out that the maximum deflection and bending moment Database plays a key role in the characterisation of model
are likely to be correlated and there is a need to examine factors. This section presents a comprehensive survey of
how the model uncertainty related to both quantities can the databases available in the literature for (1) shallow
be updated using available monitoring data. Another and deep foundations, (2) offshore spudcans, (3)
example is pile design under lateral loading, which mechanically stabilised earth and soil nail walls, (4)
assumes a range of behaviour depending on the pile- pipes and anchors (plate, helical, and shoring), (5) slopes
soil stiffness (Lesny 2017b). For short piles, rigid body and base heave, (6) cantilever walls, and (7) braced
deformation is adopted and passive and active earth excavations. A summary of these databases is given in
pressures are mobilised along the complete length of Table 1. Another ongoing database effort to develop
the pile. This represents a typical ULS, where the defor- the Databases to Interrogate Geotechnical Observations
mation is of minor importance and the model uncer- (DINGO) was reported by Hancock (2018). The
tainty can be directly quantified by Eq. (1) as DINGO database project led by Dr. Paul Vardanega in
conducted by Phoon and Kulhawy (2005). However, partnership with Arup, Atkins and Fugro will compile
for long piles (flexible) it is questionable if a single repre- pile load test data for UK and will make the database
sentative quantity (e.g. pile head movement or rotation) freely available for research and for pile design/perform-
is sufficient to define the limit state or if the whole deflec- ance regularisation, taking account of evolving ground
tion curve is a better choice to control performance. conditions and performance-based design (Pathmanan-
Lesny (2017a) suggested that it may be possible to com- davel 2018). The survey presented in Table 1 is global
pare the entire measured and calculated load-movement and covers a broad range of geo-structures at ULS and
curves. Examples can be found in the design of shallow SLS (beyond pile load tests in one country).
foundations. Briaud and Gibbens (1999) and Briaud For load test databases to be useful, they should con-
(2007) observed that a characteristic load-movement tain information beyond just the foundation geometry,
curve can be used to represent the response of footing soil profile, and load-movement curves. Below is a review
load tests, where the data are presented in terms of of some recommendations made in the literature. Due to
applied pressure versus movement normalised by the page limit, an important element relating to the esti-
footing width. Using the concept of a characteristic mation of design parameters from the soil profile is not
load-movement curve, Mayne and Illingworth (2010), covered below. Details are given elsewhere (Kulhawy
Mayne, Uzielli, and Illingworth (2012), Mayne and and Mayne 1990; Phoon and Ching 2018).
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 7

Table 1. Summary of database in literature for various geotechnical structures.


Geo-structure Database/reference Data source Test type Geomaterial N
Shallow foundations UML-GTR ShalFound07 Global Laboratory/field Cohesionless 549
(Paikowsky et al. 2010)
UML-GTR RockFound07 Global Field Rock 122
(Paikowsky et al. 2010)
Akbas (2007) Global Field Cohesionless 400
Mayne and Dasenbrock (2018) Global Field Sand 130
Patra et al. (2012a, 2012b) — Laboratory Sand 192
Okamura, Takemura, and Kimura (1997) Japan Centrifuge Sand over clay 31
Tang and Phoon (2017) — Centrifuge Dense sand 53
Samtani and Allen (2018) USA/Europe Field Cohesionless 71
Spudcans Teh (2007) NUS Centrifuge Sand over clay 14
Hossain (2014) UWA Centrifuge Clay with sand 14
Hossain and Randolph (2010) UWA Centrifuge Layered clay 40
Lee (2009) UWA Centrifuge Sand over clay 35
Hu (2015) UWA Centrifuge Sand over clay 32
Ullah (2016) UWA Centrifuge Clay-sand-clay 27
Drilled shafts Ng et al. (2001) Hong Kong Field Rock/saprolite 38
(vertical load) AbdelSalam, Baligh, and El-Naggar (2015) Egypt Field Various 318
Asem, Long, and Gardoni (2018) Global Field Soft rock 190
DSHAFT (Garder et al. 2012) Iowa, USA Field Various 38
Motamed, Elfass, and Stanton (2016) Las Vegas Valley Field Caliche 41
Stark et al. (2017) Illinois, USA Field Weak rock 155
TxDOT (Moghaddam et al. 2018) Texas Field Various 27
Drilled shafts EPRI (Chen and Kulhawy 1994) Global Field Clay/sand 88
(lateral load) Chen and Lee (2010) Global Field Clay/sand 99
Chen, Lin, and Kulhawy (2011) Global Field Clay/sand 40
Marcos and Chen (2013) Global Field Gravel 24
Augered cast-in-place piles Reddy and Stuedlein (2017a) USA Field Cohesionless 112
McVay et al. (2016) Florida, USA Field Various 78
Driven piles AAU-NGI (Augustesen 2006) Global Field Various 420
Zhang et al. (2006) Hong Kong Field (static/dynamic) Weathered granite 1514
Long, Hendrix, and Jaromin (2009) Wisconsin, USA Field (dynamic) Various 316
PILOT (Roling, Sritharan, and Suleiman 2011) Iowa, USA Field Various 275
PSU (Smith et al. 2011) Global Field Various 322
Long and Anderson (2014) Illinois, USA Field (dynamic) Various 111
ZJU-ICL (Yang et al. 2016) Global Field Sand 117
Long (2016) Wisconsin, USA Field (static/dynamic) IGM 215
Lehane et al. (2017) Global Field Various 120
Adhikari et al. (2018) Wyoming, USA Field Soft rock 25
TxDOT (Moghaddam et al. 2018) Texas Field Various 33
Helical piles Tang and Phoon (2018a) Canada/USA Field Various 1010
Driven cast-in-situ piles Long (2013) Wisconsin, USA Field Various 182
Flynn (2014) United Kingdom Field Sand 116
Pile foundations FHWA DFTLD Mainly in USA Field Various 1567
Dithinde et al. (2011) South Africa Field Various 174
IFSTTAR (Burlon et al. 2014) France Field Various 174
Niazi (2014) Global Field Various 330
Galbraith, Farrell, and Byrne (2014) Ireland Field Various 175
AUT-CPT (Moshfeghi and Eslami 2018) Global Field Various 466
WBPLT (Chen et al. 2014) Global Field Various 613
LADOTD (Rauser and Tsai 2016) Louisiana, USA Field (static/dynamic) Various 1465
Foundations EPRI (Kulhawy et al. 1983) USA Field Various 804
Mechanically Stabilised Huang and Bathurst (2009) — Laboratory Cohesionless 318
Earth walls Miyata and Bathurst (2012b) Japan Laboratory/in situ Cohesionless 652
Miyata and Bathurst (2012c) Japan Laboratory Various 503
Miyata, Bathurst, and Allen (2014) Japan Laboratory N/A 362
Miyata and Bathurst (2015) Japan Field Various 520
Miyata and Bathurst (2018) Global In situ Cohesionless 113
Allen and Bathurst (2018) — Field Various 378
Miyata, Yu, and Bathurst (2018) — In situ/laboratory Various 202
Wood et al. (2012a, 2012b) Texas, USA Laboratory Cohesionless 650
Soil Nail Walls Lazarte (2011) — Field — 166
Cheung and Shum (2012) Hong Kong Field CDG/CDV 913
Lin, Bathurst, and Liu (2017a) Global In situ — 123
Liu et al. (2018a) — In situ — 95
Yuan et al. (2019) China In situ Cohesive/cohesionless 147
Multi-anchor walls Miyata, Bathurst, and Konami (2011) Japan In situ Various 28
Slopes Travis, Schmeeckle, and Sebert (2011a) Global Field Various 157
Bahsan et al. (2014) — Field Clay 43
Excavations (base heave) Wu, Ou, and Ching (2014) Global In situ Cohesive 24
Pipes White, Cheuk, and Bolton (2008) — Small/full-scale Sand 61

(Continued )
8 K.-K. PHOON AND C. TANG

Table 1. Continued.
Geo-structure Database/reference Data source Test type Geomaterial N
Stuyts, Cathie, and Powell (2016) — Small/full-scale Sand 108
Ismail, Najjar, and Sadek (2018) — Small scale/centrifuge Sand 143
Plate anchors White, Cheuk, and Bolton (2008) — Small/full-scale Sand 54
Stuyts, Cathie, and Powell (2016) — Small/full-scale Sand 192
Helical anchors Tang and Phoon (2016) — Laboratory Cohesive 78
Field Cohesive 25
Shoring anchors Chahbaz, Sadek, and Najjar (2019) Beirut Field Clay/marl/limestone 70
Cantilever wall Phoon, Liu, and Chow (2009) — Centrifuge Sand 20
Excavation Marsland (1953) — Small-scale Loose/dense sand 23
(stability) Large-scale 10
Excavation Long (2001) Global Field Various 296
(wall deflection) Moormann (2004) Global Field Soft soil 530
Wang, Xu, and Wang (2010) Shanghai Field Soft soil 300
Wu, Ching, and Ou (2013) Taipei Field Soft clay 22
Note: CDG = completely decomposed granite; CDV = completely decomposed volcanic; IGM = intermediate geomaterial; N = number of load tests; NUS =
National University of Singapore; UWA = University of Western Australia; ZJU = Zhejiang University; ICL = Imperial College London.

Background information piles, the capacity is mainly derived from the adhesion
or friction along the soil-shaft interface and the load-
Phoon and Kulhawy (2005) emphasized that a compre-
movement curve typically demonstrates a plunging
hensive database with high-quality field and laboratory
behaviour. The shaft capacity can be fully mobilised at
tests is necessity for characterising geotechnical model
a smaller level of settlement than that required to fully
uncertainty. Abu-Hejleh et al. (2015) and Lesny
mobilise the base resistance. For friction and end bearing
(2017a) recommended that the following background
piles, the capacity is dependent on both shaft and base
information beyond the load-movement test results
resistances. The definition of a plunging capacity is
should be included within a pile load test database:
inadequate because large movements are required for
pile plunge. In the majority of load tests, a distinct plun-
(1) Subsurface data at the test site. These data should
ging capacity is not observed. To standardise the evalu-
contain geological profile, in-situ and/or laboratory
ation procedure, various failure criteria were proposed
test results which are required to estimate the geo-
to interpret or define the “capacity” as presented in Fell-
technical parameters.
enius (2018). The section 1810.3.3.1.3 on “Load test
(2) Data on the tested geo-structures. These data should
evaluation methods” in the 2018 International Building
contain test location, material type, dimensions, and
Code (IBC 2018) explicitly recommends three methods:
the installation or placement procedure.
(1) Davisson offset limit, (2) Brinch-Hansen 90% cri-
(3) Load test data. These data should contain the test
terion, and (3) Butler-Hoy criterion. Note that this inter-
scale, type, and procedure, loading type, and load
preted capacity may not be the actual value (if it exists at
test results.
all). Figure 1 shows different criteria lead to various esti-
mations of the capacity for a drilled shaft load test in
Ideally, the database should cover the whole range of
Federal Highway Administration (FHWA) (1993). The
design scenarios. In practice, the characterisation of
difference in the interpreted capacity will affect the
model factors is complicated by the availability of a lim-
model statistics as discussed by Phoon and Kulhawy
ited number of load tests from different sources with
(2005). This introduces an extraneous degree of uncer-
each covering a limited range of possible design situ-
tainty to the capacity evaluation that is related to incon-
ations only (Lesny 2017a). Application of the derived
sistent interpretation. This is undesirable. Fellenius
model statistics beyond the calibration domain must be
(2018) opined that the interpretation of a static load
verified (Phoon et al. 2016; Lesny 2017a). It is not sur-
test will be significantly affected by the presence of a
prisingly that foundation load test database is the most
residual force in the test pile. If present, it must not be
ubiquitous.
omitted from the analysis.
Interpretation of test results
Quality and classification of data
For stability problems, the measured quantity Xm often
refers to the capacity. For pile design and analysis, The indiscriminate adoption of all load test data, regard-
capacity corresponds to the load beyond which move- less of quality, can affect the model uncertainty evalu-
ment becomes excessive or progressive for little incre- ation process. The terminology of Roling, Sritharan,
ment in the applied load – the pile plunges. For friction and Suleiman (2011) is adopted to classify load test
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 9

robust and can be applied confidently to the full


range of design situations that lie outside the cali-
bration database. This method is reasonable under
the ideal case. In the worst case, the statistics are
very sensitive to all of the design parameters. For
example, the statistics for short piles will differ
from those for long piles because of physical reasons
(e.g. side resistance dominates total resistance in
long piles) or statistical reasons (e.g. spatial aver-
aging of soil strength is more significant in long
piles). In this context, it is questionable that the stat-
istics derived from a load test database are applicable
to circumstances not covered by the database. Pai-
kowsky et al. (2004) presented model statistics that
are dependent on some design parameters.

Characterisation of model factors


Figure 1. Interpretation of a drilled shaft load test by various fail-
ure criteria (FHWA 1993). Data outliers
Outliers are extreme values that deviate markedly from
data as “reliable” and “usable”. A load test is considered the main trend of a dataset. The presence of outliers
“reliable” if it contains sufficient data to produce Xm which could be attributed to human error, instrument
without any extrapolation. This is due to the lack of error and/or natural deviations in the population may
established guideline for extrapolation and considerable affect the model statistics leading to biased estimations.
uncertainty in extrapolating data (Lesny 2017a). A load Data outliers can be detected according to either univari-
test is considered “usable” if it contains sufficient infor- ate (e.g. the sample z-score and box-plot) or bivariate
mation to predict the response of a geo-structure (e.g. approach (e.g. scatter plot of predicted versus measured
resistance or settlement). Bauduin (2002) discussed values) (Dithinde et al. 2011). Yuen and Mu (2012) pro-
three main philosophies for using databases to character- posed a novel method for robust outlier detection. This
ise the model uncertainty in pile design: approach calculates the probability of outlier that incor-
porates the size of a dataset. It quantifies how probable a
(1) The model factors are evaluated for specific soil con- data point is an outlier.
ditions and specific pile types. Minor differences in
the pile installation and construction procedure,
Correlation between model factor and input
pile tip condition, and material can be accommo-
parameters
dated. This approach is ideal, but it needs a signifi-
cant number of load tests on each pile type for Because some design models provide an oversimplified
each soil. tool to predict the response of geo-structures, the vari-
(2) The model factors are evaluated by grouping similar ation in a model factor may be explainable by input par-
soils and pile types into categories. This approach is ameters such as the problem geometry (e.g. thickness of a
frequently used in practice. It requires a fairly large soil layer and dimension of the geo-structure) and soil
and comprehensive database. The same model stat- properties (e.g. stiffness and strength parameters)
istics are then applied to similar piles installed under (Phoon and Tang 2015, 2017; Zhang et al. 2015; Lin,
similar geological conditions. Bathurst, and Liu 2017a; Tang, Phoon, and Akbas
(3) A single model factor is evaluated for all piles in all 2017a; Tang et al. 2017b; Miyata, Yu, and Bathurst
soil types. In the ideal case, the statistics of a 2018; Liu et al. 2018a). This further reinforced the initial
“lumped” model factor are completely insensitive observation by Phoon and Kulhawy (1996) that model
to the design parameters, so they can be applied to factors derived from undifferentiated databases [Bauduin
all possible problem geometries, geologic for- (2002)’s third approach] are typically not robust. This
mations, soil properties, etc. In this ideal case, stat- lack of robustness may be due in part to the over simplifi-
istics estimated from a load test database cations of complex real behaviour. The degree of depen-
regardless of the range of situations covered are dency can be measured by the correlation coefficient (e.g.
10 K.-K. PHOON AND C. TANG

Pearson, Kendall, or Spearman). It is inappropriate to (3) Correction factor from numerical analysis (Zhang
treat the model factor as a random variable in this situ- et al. 2015; Phoon and Tang 2017; Tang and
ation (Phoon et al. 2016). Although correlation can be Phoon 2016, 2017; Tang et al. 2017b). A more phys-
incorporated into reliability analysis, it involves trans- ically meaningful way is to perform regression
forming correlated to uncorrelated variables. analysis of the model factor against each influential
For pile design, this issue can be partially addressed by parameter explicitly. Unfortunately, it is not an
grouping similar piles installed under similar geological easy task because load tests in the calibration data-
conditions together and determining a distinct model base are usually limited and the influential par-
factor for each group [Bauduin (2002)’s second ameters cannot be varied systematically for
approach]. Kulhawy, Phoon, and Wang (2012) opined regression analysis. As illustrated in Figure 2, the
that this grouping strategy is only partially effective key step is to use a mechanically consistent numeri-
when the model factor is not highly sensitive to a particu- cal method to establish a correction factor Mc, which
lar design parameter (e.g. pile depth to diameter ratio, D/ is defined as a ratio between the numerical result Xp
B). The reason is that it is important to ensure that the D/ and the calculated result Xc from the design model.
B ratios are fairly uniformly distributed over any one Because all practical design scenarios can be simu-
group. This requirement may be difficult to satisfy in a lated, these numerical results would be a beneficial
load test database because the records are collected supplement to the limited load test data. Regression
from the literature rather than a single comprehensive analysis is performed to capture the systematic vari-
research programme. For example, in a database group ation of Mc by a function f of the influential par-
defined for D/B > 10, it could be that D/B actually is lar- ameters. The resulting residual η is no longer
gely concentrated between 30 and 50, which may not be dependent on the input parameters. The database
sufficiently representative of the range that the group is is applied to evaluate the model factor Mp of the
supposed to represent. There are three approaches to adopted numerical method, which is likely to be ran-
remove the statistical dependency by regression analysis, dom. The model factor M’ for the method modified
as briefly described below: by the regression equation f is characterised as a pro-
duct of the residual η and the model factor Mp of the
(1) Direct regression against the database for the most numerical method. This method is currently the best
influential parameters (Lin, Bathurst, and Liu in terms of providing physical insights, correcting
2017a; Lin et al. 2017b; Tang, Phoon, and Akbas the bias in the original calculation method, handling
2017a; Liu et al. 2018a). This approach is direct problems with highly sensitive input parameters,
and simple. The primary limitation is that the and making efficient use of limited load test data.
most influential parameters are not varied systema-
tically in the database. The dependency may not be
completely captured by the regression equation. For Frequentist and Bayesian inference
the cases beyond the calibration domain, the model
For a design model in geotechnical engineering, a set of
factor might still depend on the input parameters
model factors (population) will be obtained from a data-
and its randomness should be verified prior to
base. Statistical inference concerns estimating unknown
reliability calibration.
parameters that describe the characteristics of the popu-
(2) Generalized model factor approach (Huang and
lation (e.g. mean, COV, probability distribution func-
Bathurst 2009; Dithinde et al. 2016). It entails per-
tion, and even higher order information involving
forming regression using the calculated values as
dependencies between variables, etc.). Inferences are
the predictor variable (Huang and Bathurst 2009;
Dithinde et al. 2016). The generalised model factor
is derived from the regression of lnXm on lnXc,
where ln(·) is a natural logarithm. The relation
between lnXm and lnXc is given by Xm=M’×(Xc)m,
where M’ is the generalised model factor (not
dimensionless) and m = regression constant. One
possible way to transform M’ to be dimensionless
is the normalisation of Xm and Xc (Dithinde et al.
2016). This method is purely empirical and does
not provide a physical insight on the sources of Figure 2. Correcting model factor for systematic dependencies
the statistical dependency. on input parameters.
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 11

made using data and a statistical model that links the capacity. This section summarises model statistics in
data to the parameters. There are two distinct “philos- the literature for the calculation of (1) capacity of shallow
ophies” to inference – frequentist and Bayesian. The fun- foundations under vertical, inclined, eccentric, or
damental difference between frequentist and Bayesian inclined-eccentric loading; (2) peak penetration resist-
approaches is how the concept of probability is inter- ance of offshore spudcans in sand overlying clay; (3)
preted. The frequentist approach interprets probability capacity of driven piles (including static and dynamic
as the relative frequency in a large number of trials, analysis methods) and drilled shafts under axial loading;
while the Bayesian approach interprets probability as a (4) capacity of rigid drilled shafts under lateral loading;
measure of the degree of belief about the value of an (5) maximum tensile load or pullout capacity of
unknown parameter. Currently, the frequentist approach mechanically stabilised earth and soil nail walls (that
dominates the characterisation of model factors, as it is are related to internal stability checks); (6) uplift capacity
easy to implement in the computational sense. of pipes and anchors; (7) embedment depth of cantilever
Baecher (2017) discussed that most geotechnical retaining walls; and (8) factor of safety for slopes and
engineers are intuitive Bayesians and practical examples base heave stability.
of such “Bayesian” thinking in site characterisation, data
analysis and reliability are common in practice. The
Shallow foundations
emblematic observational approach of Terzaghi is a
pure Bayesian concept although in a qualitative form Model uncertainties in the calculation of foundation
(Lacasse 2016). A practical guide for Bayesian methods capacity in cohesionless soil were evaluated by: (1)
applied in geotechnical practice was given by Juang Akbas and Kulhawy (2009) using 167 full scale field
and Zhang (2017). Moreover, the model uncertainty load tests (N = 167); (2) Paikowsky et al. (2010) using
for a generic database can be excessively large, because two databases that are UML-GTR ShalFound07 (N =
it is intended to accommodate a wide range of soil 269) for sand and UML-GTR RockFound07 (N = 58
types and site conditions. However, if we narrow down for footings in rock and N = 61 for rock sockets); (3)
a database to a specific site, the remaining calibration Phoon and Tang (2015, 2017) with 192 laboratory
data could be too limited to characterise the model factor model tests in Patra et al. (2012a, 2012b) (footing
with any degree of statistical significance. Zhang, Zhang, width B = 0.1 m); and (4) Tang and Phoon (2017) with
and Tang (2009) and Zhang et al. (2012) presented Baye- 53 centrifuge tests. The model statistics are given in
sian calibration of geotechnical model uncertainty. Sev- Table A1, where mean = 1–3.43 and COV = 0.22–0.52.
eral studies demonstrated that the Bayesian approach Some observations can be summarized as:
provides a useful tool to construct site-specific prob-
ability model of the model factor for pile capacities (1) The mean model factors for positive eccentricity are
(Park, Kim, and Chung 2012; Zhang et al. 2014) and smaller than that for negative eccentricity. This is
site-specific multivariate probability model of soil par- because the failure load in the case of positive eccen-
ameters (Ching and Phoon 2019). The Bayesian tricity is smaller than the failure load in the case of
approach can deal with the more complex design situ- negative eccentricity (Paikowsky et al. 2010), which
ations mentioned by (Lesny 2017a), but it is not simple is not incorporated into current design methods.
and not familiar to engineers. (2) The observed model factors in Figure 3 decrease
The model factor statistics for various design/calcu- nonlinearly as γB/pa (γ is soil weight and pa is atmos-
lation methods are presented in Tables A1–A13 in pheric pressure≈101 kPa) increases (Paikowsky
Appendix (see supplementary data). In the presence of et al. 2010; Phoon and Tang 2015, 2017; Tang and
correlations between a model factor and some input par- Phoon 2017; Tang, Phoon, and Akbas 2017a). This
ameters in the design/calculation method, the model fac- is primarily attributed to the scale effect in which
tor statistics may be partially or fully corrected using the the friction angle of sand depends somewhat signifi-
procedure explained in Figure 2. These corrected model cantly on the stress level (Bolton 1986). The statisti-
factor statistics are parenthesised in Tables A2 and A7 to cal dependency is removed by regressing M against
distinguish them from the conventional statistics pro- γB/pa with the database in Figure 3. The improved
duced by Eq. (1). performance of the design method corrected by the
regression function can be seen from the smaller
COV associated with mean around 1.
Ultimate limit state (ULS) model statistics
(3) Carter and Kulhawy (1988) substantially underesti-
The ultimate limit state (ULS) in geotechnical engineer- mates (on the safe side) the capacity of foundations
ing entails checking whether the load exceeds the in rock (mean = 11.9 and COV = 1.07 for footings
12 K.-K. PHOON AND C. TANG

(Tang et al. 2017b; Tang and Phoon 2019b). The


model statistics in Table A2 (mean = 0.76–2.71 and
COV = 0.12–0.42) suggest:

(1) Design methods in ISO 19905-1:2016 underestimate


the peak penetration resistance, where mean = 1.8–
2.7 and COV = 0.27–0.46. The approaches that
solely express the ratio of peak resistance to the
capacity of the underlying clay as a quadratic func-
tion of the ratio of the thickness of the upper sand
layer to the foundation diameter is likely to be over-
simplified (Lee 2009). The resulting model factor is
found to be a function of the soil parameters (e.g.
relative density of sand and strength of the under-
lying clay) and problem geometries (e.g. thickness
of the sand layer and spudcan diameter) (Tang
et al. 2017b), where the correction factor approach
shown in Figure 2 was applied to remove the statisti-
cal dependency.
Figure 3. Correlation between the capacity model factor and (2) When more realistic failure mechanisms are assumed
footing width. for stress dependent methods (Okamura, Takemura,
and Kimura 1998; Hu 2015), the model factor is no
and mean = 4.29 and COV = 0.72 for rock sockets). longer influenced by the soil parameters and
The model statistics were differentiated according to problem geometries. The mean and COV values are
the rock mass rating (RMR) by Paikowsky et al. around 1 and 0.2. This supports the hypothesis that
(2010). It was found that the model statistics (mean the lack of robustness in the model factor (or the
= 2.93-23.6 and COV = 0.46–0.65) vary with RMR. degree of dependency on the input parameters) is
(4) Goodman (1989) is much more accurate than Carter related to over-simplifications in the design method.
and Kulhawy (1988) and the performance is consist-
ently more reliable for all ranges of rock quality
(mean = 1.24–1.55 and COV = 0.43–0.6). The
results support the need to incorporate a rock qual- Axially loaded piles
ity index in the bearing capacity calculation. Pile axial capacity can be calculated by static or dynamic
analysis methods as detailed in AASHTO (2014). Static
analyses include indirect and direct methods. Indirect
methods, although involving some degree of empiricism,
Offshore spudcans in sand overlying clay
are built on established equations of bearing capacity and
Over the past two decades, the offshore industry devel- skin friction used to describe the behaviour of piles and
oped rapidly due to a rising demand for oil and gas buried anchors in compression and uplift (e.g. α- and β-
worldwide. The majority of the world’s offshore drilling methods). Direct methods eliminate the need to estimate
in water depth up to about 150 m is conducted by jack- geotechnical input parameters by transforming the in-
ups. The footing of jack-ups rig is called a spudcan, situ test results to the bearing capacity and skin friction
which is an inverted conical structure with an approxi- (e.g. CPT- and SPT-based methods) without the
mately circular plane area. When a spudcan is being mediation of a physical model. Dynamic methods
installed in stiff to weak soil layers, a sudden rapid pen- include (1) Wave Equation Analysis Programme
etration may occur if the underlying soil layer cannot (WEAP); (2) dynamic equations such as Engineering
resist the applied load that exceeds the peak resistance News Record (ENR), the formulae of Gates (1957) and
mobilised within the top layer. Design models in ISO its modified version; and (3) dynamic measurements
19905-1:2016 (i.e. punching shear and load spread) such as Case Pile Wave Analysis Program (CAPWAP)
and stress dependent methods (Okamura, Takemura, and Case method.
and Kimura 1998; Hu 2015) to calculate the peak resist- Over the past few years, the authors compiled a gen-
ance of spudcans in sand overlying clay were evaluated eric database (PILE/2739) covering many foundation
with centrifuge tests (N = 159) and numerical analyses types installed in different soil types and different
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 13

countries (see Figure 4). It includes 2739 full-scale field and Tsai 2009; Zhang and Chu 2009a; Dithinde et al.
load tests, which are compiled from several existing data- 2011; AbdelSalam et al. 2012, 2015; Ng and Fazia 2012;
bases (e.g. AAU-NGI, DSHAFT, FHWA DFLTD, Abu-Farsakh, Chen, and Haque 2013; Lacasse et al.
PILOT, WBPLT, and ZJU-ICL), real construction pro- 2013a, 2013b; Stark, Long, and Asem 2013, 2017; Burlon
jects, published technical reports and papers. The geo- et al. 2014; Long and Anderson 2014; Ng, Sritharan, and
graphical regions cover Australia, Belgium, Brazil, Ashlock 2014; McVay et al. 2016; Motamed, Elfass, and
Canada, China, Denmark, France, Germany, Indonesia, Stanton 2016; Yang et al. 2016; Lehane et al. 2017; Reddy
Iraq, Iran, Ireland, Italy, Kuwait, Mexico, Netherlands, and Stuedlein 2017a; Adhikari et al. 2018; Asem, Long,
Norway, Singapore, Sweden, Taiwan, Thailand, Turkey, and Gardoni 2018; Tang and Phoon 2018a, 2018b,
United Kingdom and United States. The geomaterial 2018c, 2018d, 2019c; Tang, Phoon, and Chen 2019).
covers soft to stiff clay, loose to dense sand, silt, and Ching, Lin, and Yen (2011) calibrated the LRFD resist-
gravel. The pile types are also diverse, covering small ance factors of bored piles with incomplete load test
to large displacement piles (e.g. steel H-piles, torque-dri- results and hybrid Gibbs-Metropolis algorithm. The
ven helical piles, driven cast-in-situ piles, and driven model statistics are given in Table A3 for driven piles
closed/open-end concrete/steel piles) and non-displace- and Table A4 for drilled shafts. Chen and Gilbert
ment piles [e.g. drilled shafts and augered cast-in-place (2017) presented a Bayesian calibration of model factors
(ACIP) piles]. for pile systems and reliability assessment. Some efforts
Because of the availability of a large number of load have been made to evaluate dynamic analysis methods
tests (see Table 1), static analysis methods have been (Paikowsky et al. 2004, 2009, 2014; Long, Hendrix, and
extensively calibrated by many researchers (Gilbert and Jaromin 2009; Long and Anderson 2014; Penfield et al.
Tang 1995; Lacasse and Nadim 1996; Eslami and Felle- 2014; Bougataya and Stuedlein 2016; Tavera et al. 2016;
nius 1997; Paikowsky et al. 2004; Abu-Farsakh, Yoon, Adhikari et al. 2018). For consistency, only statistics

Figure 4. Distribution of load test data by pile type.


14 K.-K. PHOON AND C. TANG

for the ratio between the capacity interpreted by the variability and enhance the robustness of calibrated
Davisson offset limit and the capacity calculated by LRFD resistance factors for pile foundations at a site
dynamic methods are given in Table A5. The results (Ching, Lin, and Yen 2011; Park, Kim, and Chung
indicate: 2012; Zhang et al. 2012).

(1) The evaluated model statistics are closely related to


Laterally loaded rigid drilled shafts
the calibration database including (1) number of
load tests (N), (2) geological conditions covered, Phoon and Kulhawy (2005) used the database of Chen
and (3) pile type and dimensions. Although different and Kulhawy (1994) (N = 152) to evaluate common
model statistics are obtained even for the same design methods for calculating the lateral capacity of rigid
method, the ranges of model statistics are similar, drilled shafts with D/B < 10.5. The database contains
particularly the COV. For driven piles in sand/clay, 103 laboratory small-scale model tests and 49 full-scale
mean = 0.39–1.81 and COV = 0.26–0.62 in Pai- field load tests. Two failure criteria were used to interpret
kowsky et al. (2004), mean = 0.86–1.66 and COV = the measured capacity. The first criterion is to fit the
0.24–0.56 in Lehane et al. (2017), and mean = 0.74– measured load-movement curve by a hyperbolic model
1.6 and COV = 0.13–0.64 obtained by the authors H = s/(a’+b’s), where H is applied lateral load and s is
(Tang and Phoon 2018a, 2018b, 2018c, 2018d, the resulting horizontal deflection. The hyperbolic
2019c). For drilled shafts in sand/clay (Paikowsky et capacity is obtained as the reciprocal of b’. The second
al. 2004; Tang, Phoon, and Chen 2019), mean = criterion is the lateral or moment limit, which is based
0.6–1.71 and COV = 0.16–0.67. Most COV values on the variation of apparent depth of rotation with
do not exceed 0.6. load. The apparent depth of rotation is defined as the
(2) Excessively high COV values (>1) are obtained by ratio of the butt displacement to the tangent of the
Adhikari et al. (2018) for total drilled shaft capacity butt slope (Chen and Kulhawy 1994).
and by Asem, Long, and Gardoni (2018) for drilled The model statistics are given in Table A6. In the
shaft tip capacity in soft rocks. The possible expla- undrained case, it was found that mean = 1.11–2.28
nations are: (1) pile-rock interaction needs to be bet- and COV = 0.29–0.37 for the model factor defined by
ter understood; (2) rock quality was not considered the hyperbolic capacity and mean = 0.72–1.49 and
in the design method; (3) the assumed failure mech- COV = 0.28–0.38 for the model factor defined by the lat-
anisms significantly deviate from the actual failure eral or moment limit capacity. In the drained case, it was
mechanism, and (4) each design method applies to found that mean = 0.98–1.8 and COV = 0.33–0.43 for the
different failure stages under the shaft tip and this model factor defined by the hyperbolic capacity and
scope of applicability is not clearly distinguished. mean = 0.67–1.22 and COV = 0.38–0.48 for the model
(3) The model uncertainties associated with ENR and factor defined by the lateral or moment limit capacity.
WEAP (COV > 0.7) are generally higher than those For a given method applied to a specific drainage con-
from other dynamic methods (COV = 0.3–0.6). The dition (undrained or drained), the interpretation method
time of driving such as end of driving (EOD) and for the capacity has a significant impact on the mean,
beginning of restrike (BOR) significantly affects the while the corresponding COV is almost identical. This
model statistics. The mean for EOD is larger than is reasonable, because different interpretation methods
that for BOR. This result is attributed to soil setup, will produce different values of the measured “capacity”,
i.e. the increase in pile capacity over time after pile while the uncertainty in the capacity calculation is more
driving as excess pore water pressure dissipates. The closely related to the imperfection of the adopted design
COV decreases for BOR conditions, because capacity model and possibly some site variabilities.
at BOR is more representative of pile capacity at the
time after driving and closer to the time in which
Mechanically stabilised earth structures
the static load test is conducted (Paikowsky et al.
2009; Smith et al. 2011; Paikowsky et al. 2014). Mechanically stabilised earth structures are formed
with soil filling and reinforcement used for both slope
Note that most model statistics in pile axial capacity stabilisation and retaining structures. The reinforce-
calculations were calibrated by databases covering differ- ment can be made of metallic materials (e.g. galvanised
ent sites. It was discussed by Zhang et al. (2012) that the or stainless steel in strip, grid, or bar mat) or geosyn-
associated model uncertainty can be divided into within- thetics (e.g. geotextile or geogrid). Mechanically stabil-
site and inter-site. A few site-specific load tests with ised earth and soil nail walls are cost-effective and can
Bayesian method were used to reduce the inter-site tolerate larger movements than reinforced concrete
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 15

walls. At the end of construction, mechanically stabil- facing batter, and backfill soil cohesion) which
ised earth and soil nail walls may look similar. As dis- have a considerable influence on reinforcement
cussed in Bruce and Jewell (1986), the similarities are loads (Allen and Bathurst 2015). The performance
(1) the reinforcement forces are sustained by frictional was improved by considering stiffness. The model
resistance between the soil and reinforcing elements statistics for this simplified stiffness method are:
and (2) the facing of the retained structure (e.g. prefab- mean = 0.88–1.16 and COV = 0.37–0.53 (Allen and
ricated elements in mechanically stabilised earth walls Bathurst 2015, 2018). For the other cases, the
and shotcrete in soil nail walls) does not play a major majority of the COVs vary between 0.3 and 0.6,
role in the overall structural stability. The differences while mean = 2–3.
are: (1) the construction sequence which has an impor- (2) Correlation exists between model factors and several
tant influence on the distribution of forces – soil nail parameters (e.g. vertical stress and/or tributary
walls are constructed by staged excavations from area), because the design models are oversimplified.
“top-down”, whereas mechanically stabilised earth The statistical dependency was removed either by
walls are constructed by “bottom-up”; (2) soil nail the generalised model factor approach (Huang and
wall is an in situ reinforcement technique exploiting Bathurst 2009) or by regression analyses that express
natural soil conditions which are not manufactured to a correction factor as a function of the most influen-
meet prescribed quality specifications; and (3) grouting tial parameter (Lin, Bathurst, and Liu 2017a; Lin
is usually employed in soil nail walls to bond the et al. 2017b; Liu et al. 2018a; Yuan et al. 2019).
reinforcement to soil in which load is transferred The corrected design methods is almost unbiased
along the grout to soil interface, while friction is directly (mean around 1).
generated along the soil-reinforcement interface in
mechanically stabilised earth walls. Bathurst and Javankhoshdel (2017) investigated the
Reliability analysis of internal stability checks of effect of the model type, the bias, and the variability of
mechanically stabilised earth or soil nail walls (e.g. pull- soil parameters on the reliability of steel strip reinforced
out and rupture) requires the model statistics for the soil walls against pullout failure. Liu, Tang, and Lin
maximum tensile force (Tmax) and pullout resistance (2018b) presented a maximum likelihood estimation of
(Pu) of reinforcements or nails. The structural limit the ultimate bond strength for soil nails in clayey soils.
state check requires the model statistics for the facing The design models for Tmax and Pu of multi-anchor
tensile force (Tf). Results in Table A7 include: (1) Tmax walls were evaluated by Miyata, Bathurst, and Konami
for mechanically stabilised earth walls – geosynthetics (2009) (N = 36) and Miyata, Bathurst, and Konami
[Miyata and Bathurst (2007) (N = 46); Allen and Bath- (2011) (N = 28). Yang, Ching, and Zornberg (2011) studied
urst (2015, 2018) (N = 193)] and steels [Allen and Bath- the external stability of narrow mechanically stabilised
urst (2018) (N = 185); Miyata and Bathurst (2018) (N = earth walls against sliding and overturning. The model
97)]; (2) Pu for mechanically stabilised walls – geosyn- uncertainty is captured by the variability in a reduction fac-
thetics [Huang and Bathurst (2009) (N = 318); Miyata tor and evaluated by 24 centrifuge tests. Geogrid creep and
and Bathurst (2012c) (N = 503)] and steels [Miyata and installation damage of mechanically stabilised earth struc-
Bathurst (2012a) (N = 154); Miyata and Bathurst tures were studied by Miyata, Bathurst, and Allen (2014)
(2012b) (N = 164); Huang, Bathurst, and Allen (2012) (N = 362) and Miyata and Bathurst (2015) (N = 520).
(N = 85); Miyata, Yu, and Bathurst (2018) (N = 202)];
(3) Tmax for soil nail walls [Lin, Bathurst, and Liu
Pipes and anchors
(2017a) (N = 99); Yuan et al. (2019) (N = 144)]; (4) Pu
for soil nail walls [Lin et al. (2017b) (N = 104)]; and (5) Engineering structures are usually designed to resist
Tf for soil nail walls [Liu et al. (2018a) (N = 65)]. Some uplift forces such as wind load or wave action. A com-
observations can be summarized as: mon solution is to bury a plate anchor in soil that is
fixed to the structure by a tie rod. The uplift capacity
(1) High COV values (>0.7) with mean = 0.16–0.66 are of plate anchors is commonly calculated by limit equili-
obtained for Tmax of geosynthetic in mechanically brium methods with assuming vertical or inclined slip
stabilised earth walls. This is explained as: (1) mechanisms. Buried pipes are often used for oil trans-
reinforcement loads are assumed to increase linearly portation. The complexity and uncertainty in pipe-soil
with depth below the wall top, whereas the actual interaction are significant, and simplifications and
distribution is typically trapezoidal and (2) the assumptions are made within the engineering models.
AASHTO design method does not address key vari- For example, theoretical and experimental studies on
ables (e.g. reinforcement stiffness, facing stiffness, anchor behaviour are frequently used in the
16 K.-K. PHOON AND C. TANG

development of design guidelines for pipes, assuming 2010b). The actual failure mechanism may not be
that a geometrically similar pile and anchor behave in cylindrical as assumed in two-dimensional analyses.
a similar fashion (White, Cheuk, and Bolton 2008; Roy
et al. 2018). Based on small scale and centrifuge model-
Serviceability limit state (SLS) model statistics
ling tests, model uncertainties in uplift capacity calcu-
lations of (1) pipes in sand [White, Cheuk, and Bolton This section summarises the serviceability limit state
(2008) (N = 54); Stuyts, Cathie, and Powell (2016) (N = (SLS) model statistics associated with: (1) settlement
300); Ismail, Najjar, and Sadek (2018) (N = 143)], (2) of shallow and deep foundations at a prescribed
plate anchors in sand [White, Cheuk, and Bolton working load level, (2) permissible load correspond-
(2008) (N = 61)], and (3) helical anchors in clay [Tang ing to a permissible pile head deflection, (3) wall
and Phoon (2016) (N = 103)] were evaluated. The deflection and ground movement in braced exca-
model statistics are given in Table A8, where mean = vations, and (4) parameterised load-movement curves
0.81–1.41 and COV = 0.23–0.39. in Eqs. (2) – (4).

Embedment depth of cantilever retaining walls Foundation settlement


Free embedded cantilever walls used to retain relatively Shallow foundations in granular soil
low heights of cohesionless soils are commonly designed Settlement, rather than bearing capacity, usually governs
by limit equilibrium analysis (US Army Corps of Engin- the design of shallow foundations in granular soils par-
eers 1996). It assumes that the wall rotates as a rigid ticularly for foundation width greater than 1.5 m.
body about some point in its embedded length. How- Akbas (2007) compiled an extensive database of 426
ever, such an analytical approach needs correction, case histories to examine the reliability and accuracy of
because errors arising from model idealizations are five methods. The database includes shallow foundations
ignored. Phoon, Liu, and Chow (2009) discussed with various sizes, from small plates 0.25 m wide to rafts
model factors with correcting the passive earth pressure or mats up to 135 m wide. The structures supported by
coefficient tabulated by Caquot and Kérisel, the soil fric- these foundations include test footings, bridges, build-
tion angle, and the normalised embedment depth using ings, tanks, embankments, chimneys, nuclear reactors,
20 centrifuge model tests. and silos. For those cases where complete load-move-
ment data are available, the settlement at the elastic
limit defined by the L1-L2 method was used for cali-
Factor of safety for slope and base heave stability
bration purposes. The model statistics are presented in
The level of safety for slope and base heave stability is Table A10 with respect to four foundation types as the
frequently quantified by the factor of safety (FS), soil-foundation interaction behaviour should be differ-
which is a ratio between stabilising and destabilising ent: (1) plates (B ≤ 1 m, N = 131) (mean = 1.13–4.13
forces. Two-dimensional limit equilibrium methods and COV = 0.5–1.12); (2) small footings (1 < B ≤ 3 m,
remain the most popular means for assessing slope stab- N = 90) (mean = 1.24–4.9 and COV = 0.63–1.04); (3)
ility. Because soil strength parameters, slope geometries, large footings (3 < B ≤ 6 m, N = 105) (mean = 0.97–3.3
pore-water pressures, slip surfaces, and loading con- and COV = 0.47–0.89); and (4) rafts or mats (B > 6 m,
ditions are inherently uncertain, the calculated FS is N = 100) (mean = 0.75–2.34 and COV = 0.82–1.17).
not exact. The model factor for FS is defined as the The method of Terzaghi and Peck (1948) is found to
ratio of the actual FS to the calculated FS. For a failed be the most conservative on the average (maximum
slope, the actual FS is assumed to be 1 and then, the mean). The highest variability is encountered for settle-
model factor M is simply the reciprocal of the calculated ment calculation of rafts or mats.
FS. The model statistics of slope stability (Wu 2009; Tra- Using the database of Samtani and Kulicki (2018a)
vis, Schmeeckle, and Sebert 2011b; Bahsan et al. 2014) (N = 20) for of bridge footings in the northeastern Uni-
and base heave stability (Wu, Ou, and Ching 2014) are ted States, Samtani and Kulicki (2018b) evaluated five
given in Table A9. It was found that mean = 0.89–1.27 predictive models of immediate settlement. Samtani
and COV = 0.15–0.28, with the exception of natural and Allen (2018) evaluated the methods of Hough
slopes (N = 9) analysed by Bahsan et al. (2014). The (1959) and Schmertmann, Brown, and Hartman (1978)
major limitation of limit equilibrium analyses is that with an expanded database (N = 71) and found that the
the assumed failure mechanism could significantly devi- mean = 0.9 and COV = 1.01 for Hough (1959) and
ate from the actual one especially for layered soils or in mean = 1.21 and COV = 1.13 for Schmertmann et al.
the presence of weak seams (Ching, Phoon, and Hu (1978). Very high COV values are attributed to the
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 17

large scatter at predicted settlement (sp) smaller than = 0.18 for pipe piles with diameters larger than 0.61 m.
12.7 mm (Samtani and Allen 2018). For the filtered data- In addition, the majority of the COV values are lower
base without sp<12.7 mm, the COV values for both two than 0.6. This may be a result of the smaller number of
methods decreases substantially. For example, COV for load tests and sites (i.e. less diverse site conditions).
Hough (1959) is reduced from 1.01 to 0.45. More extensive load test data are required to confirm
Based on 52 footings in rock and 37 rock sockets, the model factor statistics for calculating pile head
Muganga (2008) evaluated the Kulhawy’s (1978) method deflections.
to calculate the settlement at five working load levels of
Q = 0.25–1.75QDL. The design load QDL is equal to Limiting tolerable movement
50%QL2. QL2 is the failure load interpreted by the L1-L2 Limiting tolerable movement of building foundations is
method. It is defined as the load at point L2 (beginning another important serviceability consideration. It could
of the final linear region of load-movement curve). The be affected by the type and size of the structure, the
model statistics presented in Table A10 show: (1) intended usage of the structure, the properties of the
mean = 0.65–1.38 (increase as QDL increases) and structural materials and the underlying soils, the
COV = 1.22–1.63 for footings and (2) mean = 1.21–2.41 characteristics of the soil-foundation interface, and
and COV = 1.35–1.97 for rock sockets. the rate and uniformity of settlement. The performance
for the settlement of 171 bridges and 95 buildings and
Axially loaded piles for the angular distortion of 204 bridges and 205 build-
The model uncertainties for single pile head movement ings were evaluated by Zhang and Ng (2005). A method
in response to an applied working load were character- with a fragility curve was proposed to represent the
ised using the following databases and Eq. (1): (1) Gur- cumulative probability distribution of the limiting tol-
buz (2007) with 29 load tests on closed-end piles and erable movement.
31 load tests on continuous flight auger piles; (2)
Zhang, Xu, and Tang (2008) with 64 load tests on
Correlated load-movement model factors
steel-H piles; (3) Zhang and Chu (2009b) with 62 load
tests on large-diameter bored piles; and (4) Abchir Three load-movement models in Eqs. (2) – (4) have been
et al. (2016) for 90 load tests on driven/bored piles. applied for shallow and deep foundations. Most recently,
The model statistics are given in Table A10. The lower Chahbaz, Sadek, and Najjar (2019) utilised the hyper-
variability (COV = 0.22–0.38) reported by Zhang, Xu, bolic model in Eq. (2) to model the bond stress-displace-
and Tang (2008) and Zhang and Chu (2009b) could be ment relation of shoring anchors. The model statistics
attributed to the fact that load tests in Hong Kong for a and b are presented in Table A12. Previous studies
were carried out in only five sites of similar soil con- demonstrated that the correlated load-movement model
ditions. Hence, the range of site conditions is less diverse factors can be reasonably simulated by a (1) translation-
compared to those encountered in the studies of Gurbuz based probability model (Phoon and Kulhawy 2008;
(2007) and Abchir et al. (2016), where COV = 0.34–1.25. Dithinde et al. 2011) or (2) copula (Li et al. 2013;
Bohn, Santos, and Frank (2017) evaluated the accuracy Huffman and Stuedlein 2014; Huffman, Strahler, and
of pile settlement analysis using the proposed cubic Stuedlein 2015; Tang and Phoon 2018a, 2018b, 2018c,
root and hyperbolic t-z curve methods. The mean is 2018c, 2019c; Tang, Phoon, and Chen 2019). Some
about 1.2, while the COV was not reported. observations can be summarized as:

Laterally loaded piles (1) The power law model factors are positively corre-
In addition to the settlement ratio, the SLS model stat- lated (Figure 5). As this type of model gives an
istics can also be determined with respect to the load infinite initial curve slope, resulting in very small
ratio at a permissible movement (in service or working movements at low load levels, the hyperbolic
condition). Gurbuz (2007) calculated the statistics of model was more often utilised to simulate foun-
the ratio between measured and calculated load for later- dation load-movement curves.
ally loaded pile foundations at two pile head deflections (2) The hyperbolic parameters are negatively correlated
of 12.7 and 25.4 mm. Three methods were evaluated and for all pile types studied by the authors (see Figure 6).
the model statistics are given in Table A11. Even for the Neglecting the correlation between a and b will pro-
same design method, model statistics are significantly duce unrealistic scatter plots of a and b which can-
different for different pile types. For p-y curve methods, not capture the scatter and shape of measured
mean = 1.33 and COV = 0.65 for pipe piles with diam- curves and considerably overestimate the probability
eters smaller than 0.46 m, while mean = 1.16 and COV of failure for SLS (Phoon and Kulhawy 2008).
18 K.-K. PHOON AND C. TANG

Figure 5. Scatter plots of power law model factors for shallow footings on clay and sand.

Figure 6. Scatter plots of hyperbolic parameters for drilled shafts and driven piles.

(3) The failure criterion used to define the measured a considerable effect on the statistics of hyperbolic
capacity and the construction procedure of piles load-movement model factors in Eq. (2) (Flynn
(e.g. driven cast-in-situ versus preformed piles) has and McCabe 2019; Tang and Phoon 2019a). This
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 19

can be easily understood because different failure and preloading, depth to an underlying hard stratum,
criteria produce different measured capacities and soil stiffness and strength distribution, dewatering oper-
hyperbolic parameters. ation, adjacent surcharge, soil consolidation and creep
(4) The variability of the hyperbolic parameter b (COV (Kung et al. 2007). It may not be practical to incorporate
< 0.2) is much smaller than the variability of the all these factors in simplified models. As a result, signifi-
other hyperbolic parameter a (COV > 0.5). Stuedlein cant uncertainty can exist in the prediction.
and Reddy (2013) and Tang and Phoon (2018b) The uncertainties in predicting maximum wall deflec-
showed that the hyperbolic parameters depend on tion were calibrated by Kung et al. (2007) and Zhang
the pile slenderness ratio of pile embedment length et al. (2015). Finite element analyses were implemented
to diameter. The observations can be understood to capture the systematic variation of the model factor
by the physical meanings, where the reciprocals of with the soil properties (e.g. at-rest lateral earth pressure
“a” and “b” are the initial slope (i.e. stiffness) and coefficient, undrained shear strength, and the Young’s
asymptote (i.e. “capacity” at infinite movement) of modulus) and the problem geometries (e.g. excavation
the normalised hyperbolic curve. Uncertainty in esti- width, depth, and wall thickness). The mean and COV
mating soil stiffness is usually greater than that in for the modified model factor are 1.02 and 0.26, respect-
soil strength parameters (Phoon and Kulhawy 1999). ively. It should be pointed out the calibration database is
relatively limited in Kung et al. (2007) (N = 33) and
For a permissible movement (deterministic or ran- Zhang et al. (2015) (N = 45). The extensive databases
dom), the conventional SLS model statistics can be calcu- compiled by Long (2001), Moormann (2004) and
lated with the hyperbolic model factors: Wang, Xu, and Wang (2010) should be used to conduct
sa more comprehensive analyses. Hsiao et al. (2008) evalu-
Ms = (5) ated the ground settlement induced by adjacent exca-
a + bsa
vations by Eq. (1). The analyses showed that the
where Ms is the SLS model factor and sa is the permiss- calculated settlement was highly sensitive to the model
ible movement. For sa = 25.4 mm, the results obtained by factor. The model factor was later revised through
Muganga (2008) and Dithinde et al. (2011) are given back-calculation from on-site measurements at various
Table A13. The COV of Ms presented by Dithinde construction stages using the Bayesian updating tech-
et al. (2011) is smaller than 0.1 which is significantly nique. The mean model factor was close to 1. A similar
lower than that of Muganga (2008) (COV = 0.43–0.64). study was conducted by Juang et al. (2011) to calibrate
The a and b values in Dithinde et al. (2011) are deter- the model factor of the damage potential index for adja-
mined from regression analysis of complete load-move- cent buildings due to excavation-induced settlements.
ment data normalised by measured/interpreted capacity. In summary, the model uncertainties for foundation
The data scatter is reduced after this normalisation, settlement (COV > 0.6) are generally higher than those
because the effects of foundation geometries and sur- for foundation capacity (COV < 0.6). Kulhawy (1994)
rounding soil profiles are accounted for in the stated that most analytical methods met with only
measured/interpreted capacity. This may explain the limited success, because they did not take all important
difference in the COV between these studies. factors into account, such as the in situ stress state, com-
plex soil behaviour, characteristics of the soil-foundation
interface, and construction effects. At this point, ULS
Wall and ground movements
model statistics are the most prevalent. The main
Excavation may cause damage to adjacent buildings challenge is to characterise model factors for other geo-
when the settlement induced by the excavation is greater structures particularly at SLS, based on limited full-
than what the buildings could tolerate. To protect build- scale test data (Phoon et al. 2016). This paper partially
ings from such damages, it is essential to estimate the addresses this challenge.
deflection of the supporting wall in a braced excavation
and the resulting ground surface settlement at the back
Classification of model uncertainty
of the wall. For practical purposes, various empirical
and semi-empirical methods were proposed. A large Table 3.7.5.1 of the JCSS Probabilistic Model Code (JCSS
number of case histories were compiled and analysed 2006) only provides first-order estimates of the expected
by Long (2001), Moormann (2004), and Wang, Xu, means and standard deviations for some commonly used
and Wang (2010). The results showed that the behaviour geotechnical calculation models. The procedure and data
of a braced excavation may be affected by the excavation sources underlying these indicative model statistics are
width and depth, wall stiffness, strut spacing, stiffness not provided. In this paper, the mean and the COV of
20
Table 2. Model statistics for various geo-structures (Source: Tables A1–A11 in Appendix [see supplementary data]).
No. of tests
per group Mean COV

K.-K. PHOON AND C. TANG


Geo-structure Limit state Geomaterial No. data groups Range Mean Design method Range Mean Range Mean
Footings Bearing Sand 6 6–138 51 Vesić (1975) 0.99–1.67 1.33 0.23–0.47 0.35
Settlement (Q = QL1) Sand 4 90–131 106 D’Appolonia, D’Appolonia, and Brissette (1970) 1.13–1.71 1.48 0.47–1 0.65
1 ─ 49 Hough (1959) ─ 0.66 ─ 0.45
Punch through Sand-clay 2 27–95 61 Load spread (1:3) 1.49–1.81 1.65 0.27–0.31 0.29
Load spread (1:5) 2.36–2.37 2.37 0.33–0.38 0.36
Punching shear 1.61–2.71 2.16 0.29–0.46 0.38
Okamura, Takemura, and Kimura (1998) 0.76–0.85 0.81 0.12–0.22 0.17
Hu (2015) 0.82–1 0.91 0.13–0.19 0.16
Bearing Rock 1 ─ 58 Goodman (1989) ─ 1.23 ─ 0.54
Settlement (Q = 0.5QL2) Rock 1 ─ 52 Kulhawy (1978) ─ 0.98 ─ 1.36
Rock sockets Bearing Rock 1 ─ 61 Goodman (1989) ─ 1.52 ─ 0.54
Settlement (Q = 0.5QL2) Rock 1 ─ 37 Kulhawy (1978) ─ 1.64 ─ 1.73
Driven piles Static Compression Clay 26 4–115 28 Total stress analysis 0.39–1.54 0.97 0.13–0.62 0.34
Sand 24 5–71 29 Effective stress analysis 0.61–1.66 1.18 0.21–0.64 0.43
Layered 17 13–80 40 Total/effective stress 0.48–1.81 0.91 0.31–0.59 0.45
Tension Clay 8 4–69 28 Total stress analysis 0.74–1.43 1.01 0.13–0.39 0.27
Sand 8 5–51 20 Effective stress analysis 0.98–1.6 1.26 0.22–0.56 0.36
Dynamic Compression Soil 1 ─ 125 CAPWAP (EOD) ─ 1.63 ─ 0.49
1 ─ 162 CAPWAP (BOR) ─ 1.16 ─ 0.34
5 34–175 90 WEAP (EOD) 1.27–1.94 1.67 0.52–0.77 0.69
4 34–175 87 WEAP (BOR) 0.9–1.12 1.03 0.36–0.55 0.47
3 90–135 102 FHWA Gates formula 0.77–1.07 0.89 0.29–0.53 0.42
Drilled shafts Compression Clay 6 13–64 41 Brown, Turner, and Castelli (2010) 0.84–1.15 0.99 0.25–0.5 0.44
Sand 11 9–46 30 Brown, Turner, and Castelli (2010) 0.48–2.57 1.35 0.24–0.74 0.48
Layered 9 10–90 28 Brown, Turner, and Castelli (2010) 0.6–1.32 1.09 0.16–0.58 0.32
Tension Clay 2 13–32 22 Brown, Turner, and Castelli (2010) 0.87–1 0.94 0.34–0.37 0.36
Sand 4 11–49 26 Brown, Turner, and Castelli (2010) 0.83–1.25 1.06 0.32–0.54 0.45
Layered 3 14 - 39 26 Brown, Turner, and Castelli (2010) 1.07–1.25 1.16 0.29–0.48 0.4
Lateral Clay 1 ─ 72 Broms (1964a) ─ 1.49 ─ 0.38
Sand 1 ─ 75 Broms (1964b) ─ 1.22 ─ 0.4
Pile foundations Settlement (Q = 0.5QDA) Soil 1 ─ 29 Poulos (1994) ─ 1.11 ─ 0.65
Settlement (Q = 0.5QDA) Soil 2 29–31 30 Load transfer 0.94–1.18 1.06 0.5–0.78 0.64
Settlement (Q = 0.4Ruc) Soil 4 22–62 44 t-z curve (’FZ’ model) 1.23–1.41 1.29 0.44–0.66 0.58
t-z curve (’AB1’ model) 0.78–1.02 0.94 0.4–0.71 0.62
t-z curve (’AB2’ model) 0.66–0.89 0.81 0.67–1.09 0.88
Plate anchors Uplift Sand 1 ─ 54 Limit equilibrium ─ 1.16 ─ 0.23
Pipes Uplift Sand 3 61–300 168 Limit equilibrium 0.81–1.41 1.06 0.23–0.39 0.3
Slopes Global stability Soil 7 24–134 51 Bishop 0.89–1.27 1.01 0.15–0.28 0.22
Soil nail walls Nail tensile load – 2 45–54 49 Lazarte et al. (2015) 0.66–0.95 0.8 0.38–0.52 0.45
Pullout capacity Weak rock 2 30–74 52 Lazarte et al. (2015) 2.98––3.58 3.3 0.36–0.43 0.4
Facing tensile force – 2 23 - 42 32 Lazarte et al. (2015) 0.77–0.85 0.81 0.43–0.67 0.55
Mechanically stabilised Reinforcement tensile load Geosynthetic 6 41–143 83 Berg, Christopher, and Samtani (2009) 0.16–0.66 0.4 0.73–1.46 1.01
earth walls Steel 4 29–104 70 Berg, Christopher, and Samtani (2009) 0.85–1.36 1.14 0.39–0.58 0.47
Pullout capacity Geosynthetic 3 25–159 106 Berg, Christopher, and Samtani (2009) 2.02–2.68 2.37 0.47–0.59 0.52
Steel 9 17–129 63 Berg, Christopher, and Samtani (2009) 1.12–2.73 1.9 0.33–0.69 0.48
Multi-anchor walls Maximum tensile load c ≥ 0, f>0 1 – 36 PWRC (2002) – 0.81 – 0.79
Pullout capacity c ≥ 0, f>0 1 – 28 PWRC (2002) – 1.21 – 0.35
Note: number of data group = number of rows in Tables A1–A11 that belong to the same geo-structure, limit state, and geomaterial; PWRC = Japanese Public Works Research Center; QL1 = load at elastic point L1; QL2 =
interpreted capacity by L1-L2 method; QDA = interpreted capacity by Davisson offset limit; Ruc = calculated capacity; c = cohesion; f = friction angle.
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 21

the model factor for a range of geo-structures, geomater- et al. 2010; Allen and Bathurst 2015, 2018; Miyata, Yu,
ials, and limit states are summarized in Table 2. The num- and Bathurst 2018). Better understanding of the geo-
ber of tests is also presented as an indicator of the degree of structural behaviour is required to reduce the COV.
statistical uncertainty associated with the mean and the The next best approach is to adopt the correction
COV. The number of data group in Table 2 refers to the approach in Figure 2 where the systematic variations of
number of rows in Tables A1–A11 that belong to the model factors with the input parameters are removed
same geo-structure, limit state, and geomaterial. using numerical analyses.
Some representative model statistics drawn from
databases with 20 load tests or more are plotted in
Figure 7. Note that the model mean (bias) is “unconser- Concluding Remarks
vative” if it is less than one and “conservative” if it is lar- This Spotlight paper presents a comprehensive review of
ger than one. It is observed that for most design models load test data for various geo-structures and determines
reviewed in this paper, the mean is 0.5–2 and the COV is the ULS and SLS model statistics in a consistent way
0.3–0.6. For “unconservative” models, they can be (1) with clear references to the supporting databases. In par-
moderate (0.5 ≤ mean < 1) or (2) high (mean < 0.5). ticular, the authors compiled a generic database (PILE/
For “conservative” models, they can be (1) moderate 2739) covering many foundation types installed in differ-
(1 ≤ mean < 2), (2) high (2 ≤ mean < 3), or (3) very ent soil types. The foundation types covered include
high (mean ≥ 3). These groupings broadly follow con- small to large displacement piles (e.g. steel H-piles, tor-
ventional factors of safety (Terzaghi and Peck 1948). que-driven helical piles, driven cast-in-situ piles, and dri-
The model COV (dispersion) is classified as (1) low ven closed/open-end concrete/steel piles) and non-
(COV < 0.3), (2) medium (0.3 ≤ COV < 0.6), (3) high displacement piles (e.g. drilled shafts and ACIP piles).
(0.6 ≤ COV < 0.9), and (4) very high (COV ≥ 0.9). This The resulting summary (Table 2) represents the most
broadly follows the three-tier classification for soil prop- extensive and significant update of Table 3.7.5.1 in the
erty variability proposed by Phoon, Kulhawy, and Gri- JCSS Probabilistic Model Code (JCSS 2006) to date. It
goriu (2003). A model factor with an excessively large should be pointed out that the model statistics are appli-
COV is highlighted in bold text in Table 2 to alert the cable to any implementations of RBD such as the LRFD,
engineer to re-examine the basis of the calculation the MRFD (Phoon, Kulhawy, and Grigoriu 2003), or the
method (e.g. pile capacity/settlement in rock and maxi- expanded RBD approach (Wang, Au, and Kulhawy
mum tensile force of geosynthetic in mechanically stabil- 2011). These model statistics can be used within the
ised earth walls). It is possible that it is not capturing the First-Order Reliability Method (FORM) to derive a deter-
key features of the problem (Akbas 2007; Paikowsky ministic model partial factor at the design point as well.

Figure 7. Classification of model uncertainty based on the mean and coefficient of variation (COV) of the model factor.
22 K.-K. PHOON AND C. TANG

It is evident from this review that model factors cali- the New FHWA Design Method.” Report No. FHWA/
brated from a generic database can only be applied to a LA.12/495, Louisiana Department of Transportation and
specific site under the same caveat as using a generic cor- Development.
Abu-Farsakh, M., S. Yoon, and C. Tsai. 2009. “Calibration of
relation to convert a field measurement to a design soil Resistance Factors Needed in the LRFD Design of Driven
parameter. There is potential to apply new deep learning Piles.” Report No. FHWA/LA.09/449, Louisiana
methods to identify “similar” load test data from a generic Department of Transportation and Development.
database to supplement limited site-specific load test data. Abu-Hejleh, N., M. Abu-Farsakh, M. Suleiman, and C. Tsai.
By doing so, “site-specific” model factors can be derived. 2015. “Development and Use of High-Quality Databases
of Deep Foundation Load Tests.” Transportation Research
This effort will contribute to a broader agenda to digitalise
Record: Journal of the Transportation Research Board
foundation design for “precision construction”, where 2511: 27–36.
“site-specific” model factors and soil parameters can poss- Adhikari, P., Y. Gebreslasie, K. Ng, T. Sullivan, and S. Wulff.
ibly customise design to a particular site and even a par- 2018. “Static and Dynamic Analysis of Driven Piles in Soft
ticular location in a site. This effort will also resolve the Rocks Considering LRFD Using a Recently Developed
current ambiguity on whether model factors derived Electronic Database.” In Installation, Testing, and Analysis
of Deep Foundations (GSP 294), 83–92. Reston, VA: ASCE.
from a specific generic database can be applied to a new Akbas, S. 2007. “Deterministic and Probabilistic Assessment of
design situation that partially lies outside the database Settlements of Shallow Foundations in Cohesionless Soils.”
(e.g. pile and soil types are comparable, but pile length Ph.D. thesis, Cornell University.
is longer than those in the database). Akbas, S., and F. H. Kulhawy. 2009. “Axial Compression of
Footings in Cohesionless Soils. I: Load-Settlement
Behavior.” Journal of Geotechnical and Geoenvironmental
Acknowledgement Engineering 135 (11): 1562–1574.
Allen, T., and R. Bathurst. 2015. “Improved Simplified Method
The authors are grateful to Zijun Cao; Yit-Jin Chen Peter Day, for Prediction of Loads in Reinforced Soil Walls.” Journal of
Mahongo Dithinde, Bengt H. Fellenius, Kerstin Lesny, Yoshi- Geotechnical and Geoenvironmental Engineering 141 (11):
hisa Miyata, Shadi Najjar, Chang-Yu Ou, Sukumar Pathma- 04015069.
nandavel, Johan V. Retief, and Limin Zhang for their Allen, T., and R. Bathurst. 2018. “Application of the Simplified
invaluable comments and assistance in the preparation of Stiffness Method to Design of Reinforced Soil Walls.”
this paper. Journal of Geotechnical and Geoenvironmental Engineering
144 (5): 04018024.
Asem, P., J. Long, and P. Gardoni. 2018. “Probabilistic Model
Disclosure statement and LRFD Resistance Factors for the Tip Resistance of
Drilled Shafts in Soft Sedimentary Rock Based on Axial
No potential conflict of interest was reported by the authors. Load Tests.” In Innovations in Geotechnical Engineering:
Honoring Jean-Louis Briaud (GSP 299), 1–46. Reston, VA:
ASCE.
ORCID Atkinson, J. 2000. “Non-linear Soil Stiffness in Routine
Kok-Kwang Phoon http://orcid.org/0000-0003-2577-8639 Design.” Géotechnique 50 (5): 487–508.
Augustesen, A. 2006. “The Effects of Time on Soil Behaviour
and Pile Capacity.” Ph.D. thesis, Aalborg University.
Baecher, G. 2017. “Bayesian Thinking in Geotechnics.” In Geo-
References
Risk 2017: Keynote Lectures (GSP 292), 1–18. Reston, VA:
AASHTO. 2014. LRFD Bridge Design Specifications. 7th ed. ASCE.
Washington, DC: American Association of State Highway Baecher, G., and J. Christian. 2003. Reliability and Statistics in
and Transportation Officials (AASHTO). Geotechnical Engineering. New York: John Wiley and Sons.
Abchir, Z., S. Burlon, R. Frank, J. Habert, and S. Legrand. 2016. Bahsan, E., H. J. Liao, J. Y. Ching, and S. W. Lee. 2014.
“t-z Curves for Piles from Pressuremeter Test Results.” “Statistics for the Calculated Safety Factors of Undrained
Géotechnique 66 (2): 137–148. Failure Slopes.” Engineering Geology 172: 85–94.
AbdelSalam, S., F. Baligh, and H. M. El-Naggar. 2015. “A Bathurst, R. J., and S. Javankhoshdel. 2017. “Influence of
Database to Ensure Reliability of Bored Pile Design in Model Type, Bias and Input Parameter Variability on
Egypt.” Proceedings of the Institution of Civil Engineers – Reliability Analysis for Simple Limit States in Soil-
Geotechnical Engineering 168 (2): 131–143. Structure Interaction Problems.” Georisk: Assessment and
AbdelSalam, S., S. Sritharan, M. Suleiman, K. Ng, and M. Management of Risk for Engineered Systems and
Roling. 2012. “Development of LRFD Design Procedures Geohazards 11 (1): 42–54.
for Bridge Pile Foundations in Iowa, Vol. 3: Bauduin, C. 2002. “Design of Axially Loaded Compression
Recommended Resistant Factors with Consideration to Piles According to Eurocode 7.” In 9th International
Construction Control and Setup.” Report No. IHRB Conference on Piling and Deep Foundations. Nice, France:
Projects TR-584, Iowa Department of Transportation. DFI.
Abu-Farsakh, M., Q. M. Chen, and M. N. Haque. 2013. Bauduin, C. 2003. “Assessment of Model Factors and
“Calibration of Resistance Factors for Drilled Shafts for Reliability Index for ULS Design of Pile Foundations.” In
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 23

4th International Geotechnical Seminar on Deep Casagrande, A. 1965. “Role of the Calculated Risk in Earth
Foundations on Bored and Auger Piles, 119–136. Work and Foundation Engineering.” Journal of the Soil
Rotterdam: Millpress. Mechanics and Foundation Division 91 (4): 1–40.
Becker, D. 1996. “Limit State Design for Foundations. Part CEN. 2002. Eurocode: Basis of Structural Design. EN
I. An Overview of the Foundation Design Process.” 1990:2002. Brussels, Belgium: European Committee for
Canadian Geotechnical Journal 33 (6): 956–983. Standardization (CEN).
Becker, D. 2017. “Geotechnical Risk Management and CEN. 2004. Eurocode 7: Geotechnical Design – Part 1: General
Reliability Based Design-Lessons Learned.” In Geo-Risk Rules. EN 1997-1:2004. Brussels, Belgium: European
2017: Keynote Lectures (GSP 282), 98–121. Reston, VA: Committee for Standardization (CEN).
ASCE. Chahbaz, R., S. Sadek, and S. Najjar. 2019. “Uncertainty
Berg, R. R., B. R. Christopher, and N. C. Samtani. 2009. Quantification of the Bond Stress – Displacement
“Design and Construction of Mechanically Stabilized Relationship of Shoring Anchors in Different Geologic
Earth Walls and Reinforced Soil Slopes – Volume I.” In Units.” Georisk: Assessment and Management of Risk for
Publication No. FHWA-NHI-10-024. Washington, DC: Engineered Systems and Geohazards. under review.
Federal Highway Administration (FHWA). Chen, W. F. 1975. Limit Analysis and Soil Plasticity.
Bohn, C., A. Santos, and R. Frank. 2017. “Development of Amsterdam: Elsevier.
Axial Pile Load Transfer Curves Based on Instrumented Chen, J. B., and R. B. Gilbert. 2017. “Offshore Pile System
Load Tests.” Journal of Geotechnical and Model Biases and Reliability.” Georisk: Assessment and
Geoenvironmental Engineering 143 (1): 04016081. Management of Risk for Engineered Systems and
Bolton, M. 1981. “Limit State Design in Geotechnical Geohazards 11 (1): 55–69.
Engineering.” Ground Engineering 14 (6): 39–46. Chen, Y. J., and F. H. Kulhawy. 1994. “Case History Evaluation
Bolton, M. 1986. “The Strength and Dilatancy of Sands.” of Behavior of Drilled Shafts under Axial and Lateral
Géotechnique 36 (1): 65–78. Loading.” Report EPRI TR-104601. Palo Alto, CA: Electric
Bougataya, Y., and A. Stuedlein. 2016. “Region-specific Power Research Institute (EPRI).
Calibration of Resistance Factors for Use with Static and Chen, Y. J., and Y. H. Lee. 2010. “Evaluation of Lateral
Wave Equation Analyses of Driven Piles.” DFI Journal - Interpretation Criteria for Drilled Shaft Capacity.” Journal
The Journal of the Deep Foundations Institute 10 (3): 143– of Geotechnical and Geoenvironmental Engineering 136
152. (8): 1124–1136.
Briaud, J. L. 2007. “Spread Footings in Sand: Load Settlement Chen, Y. J., M. R. Liao, S. S. Lin, J. K. Huang, and M. C. M.
Curve Approach.” Journal of Geotechnical and Marcos. 2014. “Development of an Integrated Web-Based
Geoenvironmental Engineering 133 (8): 905–920. System with a Pile Load Test Database and pre-Analyzed
Briaud, J. L., and R. Gibbens. 1999. “Behavior of Five Large Data.” Geomechanics and Engineering 7 (1): 37–53.
Spread Footings in Sand.” Journal of Geotechnical and Chen, Y. J., S. W. Lin, and F. H. Kulhawy. 2011. “Evaluation of
Geoenvironmental Engineering 125 (9): 787–796. Lateral Interpretation Criteria for Rigid Drilled Shafts.”
Broms, B. B. 1964a. “Lateral Resistance of Piles in Cohesive Canadian Geotechnical Journal 48 (4): 634–643.
Soils.” Journal of the Soil Mechanics and Foundations Cheung, R. W. M., and K. W. Shum. 2012. “Review of the
Division 90 (2): 27–64. Approach for Estimation of Pullout Resistance of Soil
Broms, B. B. 1964b. “Lateral Resistance of Piles in Cohesionless Nails.” GEO Report No. 264. Geotechnical Engineering
Soils.” Journal of the Soil Mechanics and Foundation Office, Civil Engineering and Development Department,
Division 90 (3): 123–158. Hong Kong.
Brown, D. A., J. P. Turner, and R. J. Castelli. 2010. “Drilled Chinese Academy of Engineering (CAE). 2017. Engineering
Shafts: Construction Procedures and LRFD Design Focus 2017. Internal Report.
Methods.” In Publication No. FHWA-NHI-10-016. Ching, J. Y., H.-D. Lin, and M.-T. Yen. 2011. “Calibrating
Washington, DC: Federal Highway Administration Resistance Factors of Single Bored Piles Based on
(FHWA). Incomplete Load Test Results.” Journal of Engineering
Bruce, D., and R. Jewell. 1986. “Soil Nailing: Application and Mechanics 137 (5): 309–323.
Practice – Part 1.” Ground Engineering 11: 10–15. Ching, J. Y., and K. K. Phoon. 2012. “Modeling Parameters of
Burland, J. B., B. B. Broms, and V. F. B. DeMello. 1977. Structured Clays as a Multivariate Normal Distribution.”
“Behaviour of Foundations and Structures.” In 9th Canadian Geotechnical Journal 49 (5): 522–545.
International Conference on Soil Mechanics and Ching, J. Y., and K. K. Phoon. 2019. “Constructing Site-
Foundation Engineering, Vol. 2, 495–546. Tokyo: Japanese Specific Multivariate Probability Distribution Model Using
Society for Soil Mechanics and Foundation Engineering. Bayesian Machine Learning.” Journal of Engineering
Burlon, S., R. Frank, F. Baguelin, J. Habert, and S. Legrand. Mechanics 145 (1): 04018126.
2014. “Model Factor for the Bearing Capacity of Piles Ching, J. Y., K. K. Phoon, and Y. C. Chen. 2010a. “Reducing
from Pressuremeter Test Results: Eurocode 7 Approach.” Shear Strength Uncertainties in Clays by Multivariate
Géotechnique 64 (7): 513–525. Correlations.” Canadian Geotechnical Journal 47 (1): 16–33.
Canadian Standards Association. 2014. Canadian Highway Ching, J. Y., K. K. Phoon, and Y. G. Hu. 2010b. “Observations
Bridge Design Code. CAN/CSA-S6-14. Mississauga, on Limit Equilibrium–Based Slope Reliability Problems
Ontario, Canada: Canadian Standards Association. with Inclined Weak Seams.” Journal of Engineering
Carter, J. P., and F. H. Kulhawy. 1988. “Analysis and Design Of Mechanics 136 (10): 1220–1233.
Foundations Socketed into Rock.” Report No. EL-5918. Palo Chow, F. C. 1997. “Investigations Into the Behaviour of
Alto, CA: Electric Power Research Institute (EPRI). Displacement Piles for Offshore Foundations.” Ph.D. thesis,
24 K.-K. PHOON AND C. TANG

Department of Civil & Environmental Engineering, (DSHAFT).” Report No. InTrans Project 10-366, Iowa
Imperial College, London. Department of Transportation.
Christian, J. T. 2004. “Geotechnical Engineering Reliability: Gavin, K. G. 1998. “Experimental Investigations of Open and
How Well do We Know What We are Doing?” Journal of Closed Ended Piles in Sand.” Ph.D Thesis, University of
Geotechnical and Geoenvironmental Engineering 130 (10): Dublin (Trinity College), Dublin, Ireland.
985–1003. Gilbert, R. B., and W. H. Tang. 1995. “Model Uncertainty in
Clausen, C., P. Aas, and K. Karlsrud. 2005. “Bearing Capacity Offshore Geotechnical Reliability.” Offshore Technology
of Driven Piles in Sand, the NGI Approach.” In 1st Conference, OTC-7757-MS, American Petroleum Institute
International Symposium on Frontiers in Offshore (API).
Geotechnics (ISFOG), 677–681. London, UK: Taylor & Goodman, R. E. 1989. Introduction to Rock Mechanics. 2nd ed.
Francis group. New York: Wiley.
D’Appolonia, D., E. D’Appolonia, and R. Brissette. 1970. Griffiths, D., and G. Fenton. 2007. Probabilistic Methods in
“Closure of “Settlement of Spread Footings on Sand.” Geotechnical Engineering. New York: Springer Wien.
Journal of the Soil Mechanics and Foundations Division 96 Gurbuz, A. 2007. “The Uncertainty in the Displacement
(2): 754–761. Evaluation of Deep Foundations.” Ph.D. thesis, University
Dithinde, M., K. K. Phoon, J. Y. Ching, L. M. Zhang, and J. of Massachusetts Lowell.
Retief. 2016. “Statistical Characterization of Model Hancock, M. 2018. “Data Sharing: Please Feed the Dingo.”
Uncertainty.” In Reliability of Geotechnical Structures in https://www.geplus.co.uk/features/data-sharing-please-
ISO 2394, 127–158. Boca Raton, FL: CRC Press. feed-the-dingo/10030557.article.
Dithinde, M., K. K. Phoon, M. Wet, and J. Retief. 2011. Haque, M., and M. Abu-Farsakh. 2018. “Estimation of Pile
“Characterization of Model Uncertainty in the Static Pile Setup and Incorporation of Resistance Factor in Load
Design Formula.” Journal of Geotechnical and Resistance Factor Design Framework.” Journal of
Geoenvironmental Engineering 137 (1): 70–85. Geotechnical and Geoenvironmental Engineering 144 (11):
DNV. 2017. Recommended Practice: Statistical Representation 04018077.
of Soil Data. DNVGL-RP-C207. Oslo, Norway: Det Norske Hossain, M. S. 2014. “Experimental Investigation of Spudcan
Veritas (DNV). Penetration in Multi-Layer Clays with Interbedded Sand
Duncan, J. M. 2000. “Factors of Safety and Reliability in Layers.” Géotechnique 64 (4): 258–276.
Geotechnical Engineering.” Journal of Geotechnical and Hossain, M. S., and M. F. Randolph. 2010. “Deep-Penetrating
Geoenvironmental Engineering 126 (4): 307–316. Spudcan Foundations on Layered Clays: Centrifuge Tests.”
Eslami, A., and B. H. Fellenius. 1997. “Pile Capacity by Direct Géotechnique 60 (3): 157–170.
CPT and CPTu Methods Applied to 102 Case Histories.” Hough, B. 1959. “Compressibility as the Basis for Soil Bearing
Canadian Geotechnical Journal 34 (6): 886–904. Value.” Journal of the Soil Mechanics and Foundations
Fellenius, B. 2018. Basics of Foundation Design. Electronic ed. Division 85 (4): 11–40.
www.Fellenius.net. Hsiao, E. C. L., M. Schuster, C. H. Juang, and T. C. Kung. 2008.
Fenton, G., and D. Griffiths. 2008. Risk Assessment in “Reliability Analysis and Updating of Excavation-Induced
Geotechnical Engineering. New York: John Wiley & Sons. Ground Settlement for Building Serviceability Assessment.”
Fenton, G., F. Naghibi, D. Dundas, R. Bathurst, and D. Journal of Geotechnical and Geoenvironmental Engineering
Griffiths. 2016. “Reliability-based Geotechnical Design in 134 (10): 1448–1458.
2014 Canadian Highway Bridge Design Code.” Canadian Hu, P. 2015. “Predicting Punch-Through Failure of a Spudcan
Geotechnical Journal 53 (2): 236–251. on Sand Overlying Clay.” Ph.D. thesis, The University of
FHWA. 1993. “Axial Load-Displacement Behavior of Drilled Western Australia, Perth, Australia.
Shaft Foundations in Piedmont Residuum.” Reference No. Huang, B. Q., and R. J. Bathurst. 2009. “Evaluation of Soil-
41-30-2175, Georgia Tech Research Corporation. Geogrid Pullout Models Using a Statistical Approach.”
Flynn, K. 2014. “Experimental Investigations of Driven Cast- Geotechnical Testing Journal 32 (6): 489–504.
in-Situ Piles.” (Ph.D. thesis). National University of Huang, B. Q., R. J. Bathurst, and T. M. Allen. 2012. “LRFD
Ireland, Galway. Calibration for Steel Strip Reinforced Soil Walls.” Journal
Flynn, K., and B. A. McCabe. 2019. “Discussion of ‘Statistics of of Geotechnical and Geoenvironmental Engineering 138
Model Factors in Reliability-Based Design of Axially Loaded (8): 922–933.
Driven Piles in Sand.’.” Canadian Geotechnical Journal 56 Huffman, J., J. P. Martin, and A. W. Stuedlein. 2016.
(1): 144–147. “Calibration and Assessment of Reliability-Based
Forrest, W. S., and T. L. L. Orr. 2011. “The Effect of Model Serviceability Limit State Procedures for Foundation
Uncertainty on the Reliability of Spread Foundations.” In Engineering.” Georisk: Assessment and Management of Risk
3rd International Symposium on Geotechnical Safety and for Engineered Systems and Geohazards 10 (4): 280–293.
Risk (ISGSR), 401–408. Karlsruhe, Germany: Huffman, J. C., A. Strahler, and A. W. Stuedlein. 2015.
Bundesanstalt für Wasserbau. “Reliability-Based Serviceability Limit State Design for
Galbraith, A., E. Farrell, and J. Byrne. 2014. “Uncertainty in Immediate Settlement of Spread Footings on Clay.” Soils
Pile Resistance from Static Load Tests Database.” and Foundations 55 (4): 798–812.
Proceedings of the Institution of Civil Engineers - Huffman, J. C., and A. W. Stuedlein. 2014. “Reliability-based
Geotechnical Engineering 167 (5): 431–446. Serviceability Limit State Design of Spread Footings on
Garder, J., K. Ng, S. Sritharan, and M. Roling. 2012. “An Aggregate Pier Reinforced Clay.” Journal of Geotechnical
Electronic Database for Drilled SHAft Foundation Testing and Geoenvironmental Engineering 140 (10): 04014055.
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 25

IBC. 2018. International Building Code (IBC). Washington, Geotechnic, 17–42. Reykjavik: Icelandic Geotechnical
DC: Internal Code Council (ICC). Society (IGS).
Institution of Civil Engineers. 2017. State of the Nation: Digital Lacasse, S., and F. Nadim. 1996. “Model Uncertainty in Pile
Transformation. London: Thomas Telford. Axial Capacity Calculations.” In Offshore Technology
Ismail, S., S. S. Najjar, and S. Sadek. 2018. “Reliability Analysis Conference, OTC-7996-MS, American Petroleum Institute
of Buried Offshore Pipelines in Sand Subjected to Upheaval (API).
Buckling.” In Proceedings of the Offshore Technology Lacasse, S., F. Nadim, K. H. Andersen, S. Knudsen, U. K.
Conference (OTC). Houston, TX: American Petroleum Eidsvig, G. Yetginer, T. R. Guttormsen, and A. Eide. 2013a.
Institute. OTC-28882-MS. “Reliability of API, NGI, ICP and Fugro Axial Pile Capacity
ISO. 2015. General Principles on Reliability of Structures. ISO Calculation Methods.” In Offshore Technology Conference,
2394. Geneva, Switzerland: International Organization for OTC 24063-MS. American Petroleum Institute.
Standardization (ISO). Lacasse, S., F. Nadim, T. Langford, S. Knudsen, G. L. Yetginer,
Jardine, R. J. 1985. “Investigation of Pile Soil Behaviour with T. R. Guttormsen, and A. Eide. 2013b. “Model Uncertainty
Special Reference to the Foundations of Offshore in Axial Pile Capacity Design Methods.” In Offshore
Structures.” Ph.D. thesis, Imperial College London. Technology Conference, OTC-24066-MS. American
Jardine, R. J., F. C. Chow, R. Overy, and J. R. Standing. 2005. Petroleum Institute.
ICP Design Methods for Driven Piles in Sands and Clays. Lambe, T. W. 1973. “Predictions in Soil Engineering.”
London: Thomas Telford. Géotechnique 23 (2): 151–202.
JCSS. 2006. Probabilistic Model Code. Copenhagen, Denmark: Lazarte, C. A. 2011. “Proposed Specifications for LRFD Soil-
Joint Committee on Structural Safety (JCSS). Nailing Design and Construction.” In NCHRP Report 701.
Juang, C. H., M. Schuster, C. Y. Ou, and K. K. Phoon. 2011. Washington, DC: Transportation Research Board.
“Fully Probabilistic Framework for Evaluating Excavation- Lazarte, C. A., H. Robinson, J. E. Gómez, A. Baxter, A. Cadden,
Induced Damage Potential of Adjacent Buildings.” Journal and R. Berg. 2015. “Soil Nail Walls – Reference Manual.” In
of Geotechnical and Geoenvironmental Engineering 137 Publication No. FHWA-NHI-14-007. Washington, DC:
(2): 130–139. Federal Highway Administration (FHWA).
Juang, C. H., and J. Zhang. 2017. “Bayesian Methods for Lee, K. K. 2009. “Investigation of Potential Spudcan Punch-
Geotechnical Applications – A Practical Guide.” In Through Failure on Sand Overlying Clays.” Ph.D. thesis,
Geotechnical Safety and Reliability: Honoring Wilson The University of Western Australia, Perth, Australia.
H. Tang (GSP 286), 215–246. Reston, VA: ASCE. Lehane, B. M. 1992. “Experimental Investigations of Pile
Kolk, H. J., A. E. Baaijens, and M. Senders. 2005. “Design Behaviour Using Instrumented Field Piles.” Ph.D. thesis,
Criteria for Pipe Piles in Silica Sands.” In 1st International Department of Civil & Environmental Engineering,
Symposium on Frontiers in Offshore Geotechnics, 711–716. Imperial College, London.
London, UK: Taylor & Francis. Lehane, B. M., J. K. Kim, P. Carotenuto, F. Nadim, S. Lacasse,
Kulhawy, F. H. 1978. “Geotechnical Model for Rock R. J. Jardine, and B. F. J. Van Dijk. 2017. “Characteristics of
Foundation Settlement.” Journal of the Geotechnical Unified Databases for Driven Piles.” In 8th International
Engineering Division 104 (2): 211–227. Conference of Offshore Site Investigation and
Kulhawy, F. H. 1994. “Some Observations on Modeling in Geomechanics, vol. 1, 162–191. London, UK: Society for
Foundation Engineering.” In 8th International Conference Underwater Technology.
on Computer Methods and Advances in Geomechanics, Lehane, B. M., J. A. Schneider, and X. Xu. 2005. “The UWA-05
209–214. Rotterdam: A. A. Balkema. Method for Prediction of Axial Capacity of Driven Piles in
Kulhawy, F. H., and P. W. Mayne. 1990. “Manual on Sand.” In 1st International Symposium on Frontiers in
Estimating Soil Properties for Foundation Design.” Report Offshore Geotechnics (ISFOG), 683–689. London, UK:
EL-6800. Palo Alto, CA: Electric Power Research Institute Taylor & Francis group.
(EPRI). Lesny, K. 2017a. “Evaluation and Consideration of Model
Kulhawy, F. H., T. D. O’Rourke, J. P. Stewart, and J. F. Beech. Uncertainties in Reliability Based Design.” In Joint
1983. “Transmission Line Structure Foundations for Uplift- ISSMGE TC 205/TC 304 Working Group on “Discussion of
Compression Loading: Load Test Summaries. Appendix to Statistical/Reliability Methods for Eurocodes, 20–64.
EPRI Final Report EL-2870.” Report No. EL-3160-LD. Palo London, UK: International Society for Soil Mechanics and
Alto, CA: Electric Power Research Institute (EPRI). Geotechnical Engineering.
Kulhawy, F. H., K. K. Phoon, and Y. Wang. 2012. “Reliability- Lesny, K. 2017b. “Design of Laterally Loaded Piles-Limits of
Based Design of Foundations-A Modern View.” In Limit State Design?” In Geo-Risk 2017: Reliability-Based
Geotechnical Engineering State of the Art and Practice: Design and Code Developments (GSP 283), 267–276.
Keynote Lectures From GeoCongress 2012, 102–121. Reston, VA: ASCE.
Reston, VA: ASCE. Li, D. Q., X. S. Tang, K. K. Phoon, Y. F. Chen, and C. B. Zhou.
Kung, C., C. H. Juang, E. Hsiao, and Y. Hashash. 2007. 2013. “Bivariate Simulation Using Copula and its
“Simplified Model for Wall Deflection and Ground- Application to Probabilistic Pile Settlement Analysis.”
Surface Settlement Caused by Braced Excavation in International Journal for Numerical and Analytical
Clays.” Journal of Geotechnical and Geoenvironmental Methods in Geomechanics 37 (6): 597–617.
Engineering 133 (6): 731–747. Lin, P. Y., R. Bathurst, S. Javankhoshdel, and J. Y. Liu. 2017b.
Lacasse, S. 2016. “Hazard, Reliability and Risk Assessment – “Statistical Analysis of the Effective Stress Method and
Research and Practice for Increased Safety.” In 17th Modifications for Prediction of Ultimate Bond Strength of
Nordic Geotechnical Meeting – Challenges in Nordic Soil Nails.” Acta Geotechnica 12: 171–182.
26 K.-K. PHOON AND C. TANG

Lin, P. Y., R. Bathurst, and J. Y. Liu. 2017a. “Statistical Factors for Auger Cast in Place Piles.” Florida Department
Evaluation of the FHWA Simplified Method and of Transportation.
Modifications for Predicting Soil Nail Loads.” Journal of Miyata, Y., and R. J. Bathurst. 2007. “Evaluation of K-Stiffness
Geotechnical and Geoenvironmental Engineering 143 (3): Method for Vertical Geosynthetic Reinforced Granular Soil
04016107. Walls in Japan.” SOILS AND FOUNDATIONS 47 (2): 319–
Liu, H. F., L. S. Tang, and P. Y. Lin. 2018b. “Estimation of 335.
Ultimate Bond Strength for Soil Nails in Clayey Soils Miyata, Y., and R. J. Bathurst. 2012a. “Measured and Predicted
Using Maximum Likelihood Method.” Georisk: Assessment Loads in Steel Strip Reinforced c-f Soil Walls in Japan.”
and Management of Risk for Engineered Systems and Soils and Foundations 52 (1): 1–17.
Geohazards 12 (3): 190–202. Miyata, Y., and R. J. Bathurst. 2012b. “Analysis and Calibration
Liu, H. F., L. S. Tang, P. Y. Lin, and G. X. Mei. 2018a. of Default Steel Strip Pullout Models Used in Japan.” Soils
“Accuracy Assessment of Default and Modified Federal and Foundations 52 (3): 481–497.
Highway Administration (FHWA) Simplified Models for Miyata, Y., and R. J. Bathurst. 2012c. “Reliability Analysis of
Estimation of Facing Tensile Forces of Soil Nail Walls.” Soil-Geogrid Pullout Models in Japan.” Soils and
Canadian Geotechnical Journal 55 (8): 1104–1115. Foundations 52 (4): 620–633.
Long, M. 2001. “Database for Retaining Wall and Ground Miyata, Y., and R. J. Bathurst. 2015. “Reliability Analysis of
Movements due to Deep Excavations.” Journal of Geogrid Installation Damage Test Data in Japan.” Soils
Geotechnical and Geoenvironmental Engineering 127 (3): and Foundations 55 (2): 393–403.
203–224. Miyata, Y., and R. J. Bathurst. 2018. “Statistical Assessment of
Long, J. 2013. “Improving Agreement between Static Method Load Model Accuracy for Steel Grid-Reinforced Soil Walls.”
and Dynamic Formula for Driven Cast-in-Place Piles in Acta Geotechnica, accepted.
Wisconsin.” Report No. 0092-10-09, Wisconsin Miyata, Y., R. J. Bathurst, and T. M. Allen. 2014. “Reliability
Department of Transportation. Analysis of Geogrid Creep Data in Japan.” Soils and
Long, J. 2016. “Static Pile Load Tests on Driven Piles into Foundations 54 (4): 608–620.
Intermediate-geo Materials.” Report No. WHRP 0092-12- Miyata, Y., R. J. Bathurst, and T. Konami. 2009. “Measured
08, Wisconsin Department of Transportation. and predicted loads in multi-anchor reinforced soil walls
Long, J., and A. Anderson. 2014. “Improved Design for Driven in Japan.” Soils and Foundations 49 (1): 1–10.
Piles Based on a Pile Load Test Program in Illinois: phase 2.” Miyata, Y., R. J. Bathurst, and T. Konami. 2011. “Evaluation of
Report No. FHWA-ICT-14-019, Illinois Department of Two Anchor Plate Capacity Models for MAW Systems.”
Transportation. Soils and Foundations 51 (5): 885–895.
Long, J., J. Hendrix, and D. Jaromin. 2009. “Comparison of Miyata, Y., Y. Yu, and R. J. Bathurst. 2018. “Calibration of Soil-
Five Different Methods for Determining Pile Bearing Steel Grid Pullout Models Using a Statistical Approach.”
Capacities.” Report No. WisDOT 0092-07-04, Wisconsin Journal of Geotechnical and Geoenvironmental Engineering
Department of Transportation. 144 (2): 04017106.
Marcos, M. C. M., and Y. J. Chen. 2013. “Evaluation of Lateral Moghaddam, R. B., P. W. Jayawickrama, W. D. Lawson, J. G.
Interpretation Criteria for Drilled Shaft Capacity in Surles, and H. Seo. 2018. “Texas Cone Penetrometer
Gravels.” Geotechnical and Geological Engineering 31 (5): Foundation Design Method: Qualitative and Quantitative
1411–1420. Assessment.” DFI Journal – The Journal of the Deep
Marsland, A. 1953. “Model Experiments to Study the Influence Foundations Institute 12 (2): 69–80.
of Seepage on the Stability of a Sheeted Excavation in Sand.” Moormann, C. 2004. “Analysis of Wall and Ground
Géotechnique 3 (6): 223–241. Movements due to Deep Excavations in Soft Soil Based on
Mayne, P., and D. Dasenbrock. 2018. “Direct CPT Method for a new Worldwide Database.” Soils and Foundations 44
130 Footings on Sands.” In Innovations in Geotechnical (1): 87–98.
Engineering: Honoring Jean-Louis Briaud (GSP 299), 135– Moshfeghi, S., and A. Eslami. 2018. “Study on Pile Ultimate
146. Reston, VA: ASCE. Capacity Criteria and CPT-Based Direct Methods.”
Mayne, P., and F. Illingworth. 2010. “Direct CPT Method for International Journal of Geotechnical Engineering 12 (1):
Footings on Sand Using a Database Approach.” In 2nd 28–39.
International Symposium on Cone Penetration Testing Motamed, R., S. Elfass, and K. Stanton. 2016. “LRFD
(CPT’ 10), Vol. 3, 315–322. Madison: Omnipress. Resistance Factor Calibration for Axially Loaded Drilled
Mayne, P., M. Uzielli, and F. Illingworth. 2012. “Shallow Shafts in the Las Vegas Valley.” Report No. 515-13-803,
Footing Response on Sands Using a Direct Method Based Nevada Department of Transportation.
on Cone Penetration Tests.” In Geo-Congress 2012: Full- Muganga, R. 2008. “Uncertainty Evaluation of Displacement
Scale Testing and Foundation Design: Honoring Bengt and Capacity of Shallow Foundations on Rock.” Master of
H. Fellenius (GSP 227), 664–679. Reston, VA: ASCE. Science, University of Massachusetts Lowell.
Mayne, P., and D. Woeller. 2014. “Generalized Direct CPT Nadim, F. 2017. “Reliability-Based Approach for Robust
Method for Evaluating Footing Deformation Response Geotechnical Design.” In 19th International Conference on
and Capacity on Sand, Silts, and Clays.” In Geo-Congress Soil Mechanics and Geotechnical Engineering, 191–211.
2014: Geo-Characterization and Modeling for Seoul: South Korea.
Sustainability (GSP 234), 1983–1997. Reston, VA: ASCE. Najjar, S. S., E. Shammas, and M. Saad. 2014. “Updated
McVay, M., S. Wasman, L. Huang, and S. Crawford. 2016. Normalized Load-Settlement Model for Full-Scale
“Load and Resistance Factor Design (LRFD) Resistance Footings on Granular Soils.” Georisk: Assessment and
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 27

Management of Risk for Engineered Systems and Geohazards Patra, C., R. Behara, N. Sivakugan, and B. Das. 2012b.
8 (1): 63–80. “Ultimate Bearing Capacity of Shallow Strip Foundation
Najjar, S. S., E. Shammas, and M. Saad. 2017. “A Reliability- Under Eccentrically Inclined Load, Part II.” International
Based Approach to the Serviceability Limit State Design of Journal of Geotechnical Engineering 6 (4): 507–514.
Spread Footings on Granular Soil.” In Geotechnical Safety Penfield, J., R. Parsons, J. Han, and A. Misra. 2014. “Load and
and Reliability: Honoring Wilson H. Tang (GSP 286), 185– Resistance Factor Design Calibration to Determine a
202. Reston, VA: ASCE. Resistance Factor for the Modification of the Kansas
Ng, T., and S. Fazia. 2012. “Development and Validation of a Department of Transportation-Engineering News Records
Unified Equation for Drilled Shaft Foundation Design in Formula.” Report No. K-TRAN: KU-13-4, Kansas
New Mexico.” Report No. NM10MSC-01, New Mexico Department of Transportation.
Department of Transportation. Phoon, K. K. 2005. “Reliability-Based Design Incorporating
Ng, K., S. Sritharan, and J. Ashlock. 2014. “Development of Model Uncertainties.” In 3rd International Conference on
Preliminary Load and Resistance Factor Design of Drilled Geotechnical Engineering Combined with 9th Yearly
Shafts in Iowa.” Report No. InTrans Project 11-410, Iowa Meeting of the Indonesian Society for Geotechnical
Department of Transportation. Engineering, 191–203. Semarang, Indonesia: Indonesian
Ng, C. W. W., T. L. Y. Yau, J. H. M. Li, and W. H. Tang. 2001. Society for Geotechnical Engineering.
“New Failure Load Criterion for Large Diameter Bored Piles Phoon, K. K. 2017. “Role of Reliability Calculations in
in Weathered Geomaterials.” Journal of Geotechnical and Geotechnical Design.” Georisk: Assessment and
Geoenvironmental Engineering 127 (6): 488–498. Management of Risk for Engineered Systems and
Niazi, F. S. 2014. “Static Axial Pile Foundation Response Using Geohazards 11 (1): 4–21.
Seismic Piezocone Data.” Ph.D. thesis, Georgia Institute of Phoon, K. K. 2018. “Probabilistic Site Characterization.”
Technology, USA. ASCE-ASME Journal of Risk and Uncertainty in
Okamura, M., J. Takemura, and T. Kimura. 1997. “Centrifuge Engineering Systems, Part A: Civil Engineering 4 (4):
Model Test on Bearing Capacity and Deformation of Sand 02018002.
Layer Overlying Clay.” Soils and Foundations 37 (1): 73–88. Phoon, K. K., and J. Y. Ching. 2014. Risk and Reliability in
Okamura, M., J. Takemura, and T. Kimura. 1998. “Bearing Geotechnical Engineering. Boca Raton: CRC Press.
Capacity Predictions of Sand Overlying Clay Based on Phoon, K. K., and J. Y. Ching. 2018. “Better Correlations for
Limit Equilibrium Methods.” Soils and Foundations 38 Geotechnical Design.” In A Decade of Geotechnical
(1): 181–194. Advances, 73–102. Geotechnical Society of Singapore
Paikowsky, S., R. Birgisson, M. McVay, T. Nguyen, C. Kuo, G. (GeoSS).
Baecher, B. Ayyub, et al. 2004. “Load and Resistance Factor Phoon, K. K., and F. H. Kulhawy. 1996. “Practical Reliability-
Design (LRFD) for Deep Foundations.” NCHRP Report 507. Based Design Approach for Foundation Engineering.” In
Washington, DC: Transportation Research Board. Research Record 1546, Transportation Research Board,
Paikowsky, S., M. Canniff, K. Lesny, A. Kisse, S. Amatya, and 94–99. Washington: SAGE Publishing.
R. Muganga. 2010. “LRFD Design and Construction of Phoon, K. K., and F. H. Kulhawy. 1999. “Characterization of
Shallow Foundations for Highway Bridge Structures.” Geotechnical Variability.” Canadian Geotechnical Journal
NCHRP Report 651. Washington, DC: Transportation 36 (4): 612–624.
Research Board. Phoon, K. K., and F. H. Kulhawy. 2005. “Characterisation of
Paikowsky, S., M. Canniff, S. Robertson, and S. Budge. 2014. Model Uncertainties for Laterally Loaded Rigid Drilled
“Load and Resistance Factor Design (LRFD) Pile Driving Shafts.” Géotechnique 55 (1): 45–54.
Project – Phase II Study.” MnDOT Report No. MN/RC Phoon, K. K., and F. H. Kulhawy. 2008. “Serviceability-Limit
2014-16, Minnesota Department of Transportation. State Reliability-Based Design.” In Reliability-based Design
Paikowsky, S., and Y. Lu. 2006. “Establishing Serviceability in Geotechnical Engineering: Computations and
Limit State in Design of Bridge Foundations.” In Geo- Applications, 344–384. London, UK: Taylor & Francis
Shanghai International Conference: Foundation Analysis & group.
Design: Innovative Methods (GSP 153), 49–58. Reston, Phoon, K. K., F. H. Kulhawy, and M. D. Grigoriu. 1995.
VA: ASCE. “Reliability-Based Design of Foundations for
Paikowsky, S., C. Marchionda, C. O’Hearn, M. Canniff, and A. Transmission Line Structures.” Report EPRI TR-105000.
Budge. 2009. “Developing a Resistance Factor for MnDOTs Palo Alto, CA: Electric Power Research Institute (EPRI).
Pile Driving Formula.” Report No. MN/RC 2009-37, Phoon, K. K., F. H. Kulhawy, and M. D. Grigoriu. 2003.
Minnesota Department of Transportation. “Multiple Resistance Factor Design for Shallow
Park, J. H., D. Kim, and C. K. Chung. 2012. “Implementation Transmission Line Structure Foundations.” Journal of
of Bayesian Theory on LRFD of Axially Loaded Driven Geotechnical and Geoenvironmental Engineering 129 (9):
Piles.” Computers and Geotechnics 42: 73–80. 807–818.
Pathmanandavel, S. 2018. “Lecture 3 – Digital in Foundation Phoon, K. K., S. L. Liu, and Y. K. Chow. 2009.
Engineering – Case Histories.” ISSMGE – International “Characterization of Model Uncertainties for Cantilever
Seminar Foundation Design, Mexico. Walls in Sand.” Journal of GeoEngineering 4 (3): 75–85.
Patra, C., R. Behara, N. Sivakugan, and B. Das. 2012a. Phoon, K. K., and J. V. Retief. 2016. Reliability of Geotechnical
“Ultimate Bearing Capacity of Shallow Strip Foundation Structures in ISO 2394. Leiden: CRC Press/Balkema.
Under Eccentrically Inclined Load, Part I.” International Phoon, K. K., J. V. Retief, J. Ching, M. Dithinde, T.
Journal of Geotechnical Engineering 6 (3): 343–352. Schweckendiek, Y. Wang, and L. M. Zhang. 2016. “Some
28 K.-K. PHOON AND C. TANG

Observations on ISO2394:2015 Annex D (Reliability of Seo, H., R. B. Moghaddam, J. G. Surles, and W. D. Lawson.
Geotechnical Structures).” Structural Safety 62: 24–33. 2015. “Implementation of LRFD Geotechnical Design for
Phoon, K. K., and C. Tang. 2015. “Model Uncertainty for the Deep Foundations Using Texas Penetrometer (TCP) test.”
Capacity of Strip Footings Under Negative and General Report No. FHWA/TX-16/5-6788-01-1. Texas Department
Combined Loading.” 12th International Conference on of Transportation.
applications of statistics and probability in civil engineering, Smith, T., A. Banas, M. Gummer, and J. Jin. 2011.
ICASP 12,, Vancouver, Canada,, July 12–15. “Recalibration of the GRLWEAP LRFD Resistance Factor
Phoon, K. K., and C. Tang. 2017. “Model Uncertainty for the for Oregon DOT.” Report No. FHWA-OR-RD-11-08,
Capacity of Strip Footings Under Positive Combined Oregon Department of Transportation.
Loading.” In Geotechnical Safety and Reliability: Honoring Stark, T., J. Long, and P. Asem. 2013. “Improvement for
Wilson H. Tang (GSP 286), 40–60. Reston, VA: ASCE. Determining The Axial Capacity of Drilled Shafts in Shale
Poulos, H. G. 1994. “Settlement Prediction for Driven Piles in Illinois.” Report No. FHWA-ICT-13-017, Illinois
and Pile Groups.” In Settlement 94: Vertical and Department of Transportation.
Horizontal Deformations of Foundations and Stark, T., J. Long, A. Baghdady, and A. Osouli. 2017. “Modified
Embankments (GSP 40), 1629–1649. Reston, VA: ASCE. standard penetration test-based drilled shaft design method
PWRC. 2002. Design Method, Construction Manual and for weak rocks (phase 2 study).” Report No. FHWA-ICT-17-
Specifications for Multi-Anchored Reinforced Retaining 018, Illinois Department of Transportation.
Wall. Ibaraki-ken, Japan: Public Works Research Center Stuedlein, A., and S. Reddy. 2013. “Factors Affecting the
(PWRC). Reliability of Augered Cast-in-Place Piles in Granular Soil
Randolph, M. F. 2003. “Science and Empiricism in Pile at the Serviceability Limit State.” DFI Journal - The
Foundation Design.” Géotechnique 53 (10): 847–875. Journal of the Deep Foundations Institute 7 (2): 46–57.
Rauser, J., and C. Tsai. 2016. “Beneficial use of the Louisiana Stuedlein, A., and M. Uzielli. 2014. “Serviceability Limit State
Foundation Load Test Database.” In Transportation Design for Uplift of Helical Anchors in Clay.”
Research Board 95th Annual Meeting, 1–17. Washington, Geomechanics and Geoengineering 9 (3): 173–186.
DC: Transportation Research Board. Stuyts, B., D. Cathie, and T. Powell. 2016. “Model Uncertainty
Reddy, S., and A. Stuedlein. 2017a. “Ultimate Limit State in Uplift Resistance Calculations for Sandy Backfills.”
Reliability-Based Design of Augered Cast-in-Place Piles Canadian Geotechnical Journal 53 (11): 1831–1840.
Considering Lower-Bound Capacities.” Canadian Tang, W. H. 1979. “Probabilistic Evaluation of Penetration
Geotechnical Journal 54 (12): 1693–1703. Resistances.” Journal of Geotechnical Engineering 105 (10):
Reddy, S., and A. Stuedlein. 2017b. “Serviceability Limit State 1173–1191.
Reliability-Based Design of Augered Cast-in-Place Piles in Tang, C., and K. K. Phoon. 2016. “Model Uncertainty of
Granular Soils.” Canadian Geotechnical Journal 54 (12): Cylindrical Shear Method for Calculating the Uplift
1704–1715. Capacity of Helical Anchors in Clay.” Engineering Geology
Roling, M., S. Sritharan, and M. Suleiman. 2011. 207: 14–23.
“Development of LRFD Procedures for Bridge Pile Tang, C., and K. K. Phoon. 2017. “Model Uncertainty of
Foundations in Iowa. Vol. 1: An electronic Database for Eurocode 7 Approach for Bearing Capacity of Circular
Pile Load Tests (PILOT).” Report No. IHRB Project TR- Footings on Dense Sand.” International Journal of
573, Iowa Department of Transportation. Geomechanics 17 (3): 04016069.
Roy, K., B. Hawlader, S. Kenny, and Ian Moore. 2018. “Lateral Tang, C., and K. K. Phoon. 2018a. “Statistics of Model Factors
Resistance of Pipes and Strip Anchors Buried in Dense and Consideration in Reliability-Based Design of Axially
Sand.” Canadian Geotechnical Journal 55 (12): 1812–1823. Loaded Helical Piles.” Journal of Geotechnical and
Samtani, N. C., and T. M. Allen. 2018. “Implementation Report Geoenvironmental Engineering 144 (8): 04018050.
– Expanded Database for Service Limit State Calibration of Tang, C., and K. K. Phoon. 2018b. “Evaluation of Model
Immediate Settlement of Bridge Foundation on Soil.” Uncertainties in Reliability-Based Design of Steel H-Piles
Report No. FHWA-HIF-18-008. Washington, DC: Federal in Axial Compression.” Canadian Geotechnical Journal 55
Highway Administration (FHWA). (11): 1513–1532.
Samtani, N. C., and J. M. Kulicki. 2018a. “Incorporation of Tang, C., and K. K. Phoon. 2018c. “Statistics of Model Factors
Foundation Movements in AASHTO LRFD Bridge Design in Reliability-Based Design of Axially Loaded Driven Piles
Process.” 2nd ed. Report No. FHWA-HIF-18-007. in Sand.” Canadian Geotechnical Journal 55 (11): 1592–
Washington, DC: Federal Highway Administration (FHWA). 1610.
Samtani, N. C., and J. M. Kulicki. 2018b. “Calibration of Tang, C., and K. K. Phoon. 2018d. “Characterization of Model
Foundation Movements for AASHTO LRFD Bridge Design Uncertainty in Predicting Axial Resistance of Piles Driven
Specifications.” Georisk: Assessment and Management of Into Clay.” Canadian Geotechnical Journal, accepted.
Risk for Engineered Systems and Geohazards, accepted. Tang, C., and K. K. Phoon. 2019a. “Reply to the Discussion by
Schmertmann, J. H., P. R. Brown, and J. P. Hartman. 1978. Flynn and McCabe on ‘Statistics of Model Factors in
“Improved strain influence factor diagrams.” Journal of Reliability-Based Design of Axially Loaded Driven Piles in
the Geotechnical Engineering Division 104 (8): 1131–1135. Sand’.” Canadian Geotechnical Journal 56 (1): 148–152.
Schneider, J., X. Xu, and B. Lehane. 2008. “Database Tang, C., and K. K. Phoon. 2019b. “Evaluation of Stress
Assessment of CPT-Based Design Methods for Axial Dependent Methods for the Punch-Through Capacity of
Capacity of Driven Piles in Siliceous Sands.” Journal of Foundations in Clay with Sand.” ASCE-ASME Journal of
Geotechnical and Geoenvironmental Engineering 134 (9): Risk and Uncertainty in Engineering Systems, Part A: Civil
1227–1244. Engineering. accepted.
GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 29

Tang, C., and K. K. Phoon. 2019c. “Statistical Evaluation of Soils.” Journal of Geotechnical and Geoenvironmental
Model Factors in Reliability Calibration of High Engineering 136 (7): 985–994.
Displacement Helical Piles Under Axial Loading.” White, D. 2005. “A General Framework for Shaft Resistance on
Canadian Geotechnical Journal, under review. Displacement Piles in Sand.” In 1st International
Tang, C., K. K. Phoon, and S. O. Akbas. 2017a. “Model Symposium on Frontiers in Offshore Geotechnics (ISFOG),
Uncertainties for the Static Design of Square Foundations 697–703. London: Taylor & Francis group.
on Sand Under Axial Compression.” In Geo-Risk 2017: White, D., C. Cheuk, and M. Bolton. 2008. “The Uplift
Reliability-Based Design and Code Developments (GSP Resistance of Pipes and Plate Anchors Buried in Sand.”
283), 141–150. Reston, VA: ASCE. Géotechnique 58 (10): 771–779.
Tang, C., K. K. Phoon, and Y. J. Chen. 2019. “Statistical Whitman, R. 1984. “Evaluating Calculated Risk in
Analyses of Model Factors in Reliability-Based Limit State Geotechnical Engineering.” Journal of Geotechnical
Design of Driled Shafts under Axial Loading.” Journal of Engineering 110 (2): 143–188.
Geotechnical and Geoenvironmental Engineering. accepted. Wood, T., P. Jayawickrama, J. Surles, and W. Lawson. 2012a.
Tang, C., K. K. Phoon, L. Zhang, and D. Q. Li. 2017b. “Model “Pullout Resistance of MSE Reinforcements in Backfills
Uncertainty for Predicting the Bearing Capacity of Sand Typically Used in Texas: Vol. 2, Test Reports for MSE
Overlying Clay.” International Journal of Geomechanics 17 Reinforcements in Type B (sandy) Backfill.” Research
(7): 04017015. Report: FHWA/TX-13/0-6493-3, Vol. 2, Texas Department
Tavera, E., G. Burnworth, G. Rix, and J. Jung. 2016. “Calibration of Transportation.
of Region-Specific Gates Pile Driving Formula for LRFD.” Wood, T., P. Jayawickrama, J. Surles, and W. Lawson. 2012b.
Report No. FHWA/LA.16/561, Louisiana Department of “Pullout Resistance of MSE Reinforcements in Backfills
Transportation and Development. Typically Used in Texas: Vol. 3, Test Reports for MSE
Teh, K. L. 2007. “Punch-Through of Spudcan Foundation in Reinforcements in Type A (gravelly) Backfill.” Research
Sand Overlying Clay.” Ph.D. thesis, National University of Report: FHWA/TX-13/0-6493-3, Vol. 3, Texas Department
Singapore, Singapore. of Transportation.
Teixeira, A., Y. Honjo, A. G. Correia, and A. A. Henriques. Wu, T. H. 2009. “Reliability of Geotechnical Predictions.” In
2012. “Sensitivity Analysis of Vertically Loaded Pile 2nd International Symposium on Geotechnical Safety and
Reliability.” Soils and Foundations 52 (6): 1118–1129. Risk, 3–10. Gifu, Japan: Taylor & Francis Group.
Terzaghi, K., and R. P. Peck. 1948. Soil Mechanics in Engineering Wu, S. H., J. Y. Ching, and C. Y. Ou. 2013. “Predicting Wall
Practice. 1st ed. New York: John Wiley & Sons Inc. Displacements for Excavations with Cross Walls in Soft
Travis, Q., M. Schmeeckle, and D. Sebert. 2011a. “Meta-analysis Clay.” Journal of Geotechnical and Geoenvironmental
of 301 Slope Failure Calculations. I: Database Description.” Engineering 139 (6): 914–927.
Journal of Geotechnical and Geoenvironmental Engineering Wu, S. H., C. Y. Ou, and J. Y. Ching. 2014. “Calibration of Model
137 (5): 453–470. Uncertainties in Base Heave Stability for Wide Excavations in
Travis, Q., M. Schmeeckle, and D. Sebert. 2011b. “Meta-analy- Clay.” Soils and Foundations 54 (6): 1159–1174.
sis of 301 Slope Failure Calculations. II: Database Analysis.” Wu, T. H., W. H. Tang, D. A. Sangrey, and G. B. Baecher. 1989.
Journal of Geotechnical and Geoenvironmental Engineering “Reliability of Offshore Foundations—State of the Art.”
137 (5): 471–482. Journal of Geotechnical Engineering 115 (2): 157–178.
Ullah, S. N. 2016. “Jackup foundation punch-through in clay Yang, K. H., J. Ching, and J. G. Zornberg. 2011. “Reliability-
with interbedded sand.” Ph.D. thesis, The University of Based Design for External Stability of Narrow Mechanically
Western Australia, Perth, Australia. Stabilized Earth Walls: Calibration From Centrifuge Tests.”
US Army Corps of Engineers. 1996. Design of Sheet Pile Walls Journal of Geotechnical and Geoenvironmental Engineering
– Design Guide 15. New York: ASCE Press. 137 (3): 239–253.
Uzielli, M., and P. W. Mayne. 2011. “Serviceability Limit State Yang, Z., R. Jardine, W. Guo, and F. Chow. 2016. A
CPT-Based Design for Vertically Loaded Shallow Footings Comprehensive Database of Tests on Axially Loaded Piles
on Sand.” Geomechanics and Geoengineering 6 (2): 91–107. Driven in Sand. London: Academic Press.
Vardanega, P. J., and M. D. Bolton. 2016. “Design of Yuan, J., P. Y. Lin, R. Huang, and Y. Que. 2019. “Statistical
Geostructural Systems.” ASCE-ASME Journal of Risk and Evaluation and Calibration of Two Methods for
Uncertainty in Engineering Systems, Part A: Civil Predicting Nail Loads of Soil Nail Walls in China.”
Engineering 2 (1): 04015017. Computers and Geotechnics 108: 269–279.
Vesić, A. 1975. “Bearing Capacity of Shallow Foundations.” In Yuen, K. V., and H. Q. Mu. 2012. “A Novel Probabilistic
Foundation Engineering Handbook,, 121–147. New York: Method for Robust Parametric Identification and Outlier
Van Nostrand Reinhold. Detection.” Probabilistic Engineering Mechanics 30: 48–59.
Wang, Y., S. K. Au, and F. H. Kulhawy. 2011. “Expanded Zhang, L. M., and L. F. Chu. 2009a. “Calibration of Methods
Reliability-Based Design Approach for Drilled Shafts.” for Designing Large-Diameter Bored Piles: Ultimate Limit
Journal of Geotechnical and Geoenvironmental Engineering State.” Soils and Foundations 49 (6): 883–895.
137 (2): 140–149. Zhang, L. M., and L. F. Chu. 2009b. “Calibration of Methods
Wang, Y., T. Schweckendiek, W. P. Gong, T. Y. Zhao, and K. for Designing Large-Diameter Bored Piles: Serviceability
K. Phoon. 2016. “Direct Probability-Based Design Limit State.” Soils and Foundations 49 (6): 897–908.
Methods.” In Reliability of Geotechnical Structures in ISO Zhang, J., J. P. Li, L. M. Zhang, and H. W. Huang. 2014.
2394, 193–226. Boca Raton, FL: CRC Press. “Calibrating Cross-Site Variability for Reliability-Based
Wang, J., Z. Xu, and W. Wang. 2010. “Wall and Ground Design of Pile Foundations.” Computers and Geotechnics
Movements due to Deep Excavations in Shanghai Soft 62: 154–163.
30 K.-K. PHOON AND C. TANG

Zhang, L. M., and A. M. Y. Ng. 2005. “Probabilistic Limiting Zhang, J., W. H. Tang, L. M. Zhang, and H. W. Huang. 2012.
Tolerable Displacements for Serviceability Limit State “Characterising Geotechnical Model Uncertainty by Hybrid
Design of Foundations.” Géotechnique 55 (2): 151–161. Markov Chain Monte Carlo Simulation.” Computers and
Zhang, D. M., K. K. Phoon, H. W. Huang, and Q. F. Hu. Geotechnics 43: 26–36.
2015. “Characterization of Model Uncertainty for Zhang, L. M., Y. Xu, and W. H. Tang. 2008. “Calibration of
Cantilever Deflections in Undrained Clay.” Journal of Models for Pile Settlement Analysis Using 64
Geotechnical and Geoenvironmental Engineering 141 (1): Field Load Tests.” Canadian Geotechnical Journal 45 (1):
04014088. 59–73.
Zhang, L. M., L. M. P. Shek, H. W. Pang, and C. F. Pang. 2006. Zhang, J., L. M. Zhang, and W. H. Tang. 2009. “Bayesian
“Knowledge-based Design and Construction of Driven Framework for Characterizing Geotechnical Model
Piles.” Proceedings of the Institution of Civil Engineers - Uncertainty.” Journal of Geotechnical and Geoenvironmental
Geotechnical Engineering 159 (3): 177–185. Engineering 135 (7): 932–940.

You might also like