Distributed Fiber Optic Sensing of Axially

Loaded Bored Piles

Loizos Pelecanos 1; Kenichi Soga, M.ASCE 2; Mohammed Z. E. B. Elshafie 3; Nicholas de Battista 4;
Cedric Kechavarzi 5; Chang Ye Gue 6; Yue Ouyang 7; and Hyung-Joon Seo 8

Abstract: Instrumented pile tests are vital to establish the performance of a pile and validate the assumptions made during initial design.
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

Conventional instrumentation includes vibrating wire strain gauges and extensometers to measure the change in strain or displacements
within a pile. Although these strain and displacement gauges are very accurate, they only provide strain/displacement readings at discrete
locations at which they are installed. It is therefore common to interpolate between two consecutive points to obtain values corresponding to
the data gaps between points; in practice, these discrete instrumented points could be tens of meters apart, at depths corresponding to different
soil layers, and hence simple interpolation between the measurement points remains questionable. The Brillouin optical time-domain re-
flectometry fiber optic strain sensing system is able to provide distributed strain sensing along the entire length of the cable, enabling the full
strain profile to be measured during a maintained pile load test. The strain data can also be integrated to obtain the displacement profile. This
paper presents three case studies which investigate the performance of three concrete bored piles in London using both conventional vibrating
wire strain gauges and distributed fiber optic strain sensing during maintained pile load tests, which enable comparisons made between the
two instrumentation systems. In addition, finite-element analyses show that the ability to measure the full strain profiles for each pile is highly
advantageous in understanding the performance of the pile and in detecting any abnormalities in the pile behavior. DOI: 10.1061/(ASCE)
GT.1943-5606.0001843. © 2017 American Society of Civil Engineers.
Author keywords: Piles; Field monitoring; Fiber optic sensors; Load transfer; Pile load test; Finite-element analysis; Pile instrumentation.

1
Lecturer in Geotechnical Engineering, Dept. of Architecture and
Civil Engineering, Univ. of Bath, Claverton Down, Bath BA2 7AY,
U.K.; formerly, Research Associate, Dept. of Architecture and Civil
Engineering, Univ. of Cambridge, Cambridge CB2 1TN, U.K. (correspond-
ing author). ORCID: https://orcid.org/0000-0001-6183-1439. E-mail:
L.Pelecanos@bath.ac.uk
2
Chancellor’s Professor of Civil Engineering, Dept. of Civil and Envir-
onmental Engineering, Univ. of California, 760 Davis Hall, Berkeley, CA
94720-1710; formerly, Professor, Dept. of Civil and Environmental Engi-
neering, Univ. of Cambridge, Cambridge CB2 1TN, U.K. E-mail:
Soga@berkeley.edu
3 soil investigation [e.g., cone penetration test (CPT) and standard pen-
Lecturer in Construction Engineering, Centre for Smart Infrastructure
and Construction, Dept. of Engineering, Univ. of Cambridge, Trumpington
St., Cambridge CB2 1PZ, U.K. E-mail: ME254@cam.ac.uk
4
Research Associate, Centre for Smart Infrastructure and Construction,
Dept. of Engineering, Univ. of Cambridge, Trumpington St., Cambridge
CB2 1PZ, U.K. E-mail: N.Debattista@eng.cam.ac.uk
5
Training and Knowledge Transfer Manager, Centre for Smart Infra-
structure and Construction, Dept. of Engineering, Univ. of Cambridge,
Trumpington St., Cambridge CB2 1PZ, U.K. E-mail: CK209@cam.ac.uk
6
Ph.D. Research Student, Centre for Smart Infrastructure and Con-
struction, Dept. of Engineering, Univ. of Cambridge, Trumpington St.,
Cambridge CB2 1PZ, U.K. E-mail: CYG20@cam.ac.uk
7
Project Manager, Cementation Skanska, Neelands House, Pipering
Lane, Doncaster DN5 9NB, U.K.; formerly, Univ. of Cambridge,
Cambridge CB2 1PZ, U.K. E-mail: Echo.Ouyang@skanska.co.uk
8
Lecturer, Dept. of Civil Engineering, Xi’an Jiaotong–Liverpool Univ.,
Shaanxi Sheng 710048, China; formerly, Research Associate, Dept. of
Civil Engineering, Univ. of Cambridge, Cambridge CB2 1PZ, U.K. E-mail:
Hyungjoon.Seo@xjtlu.edu.cn
Note. This manuscript was submitted on March 7, 2016; approved on
September 1, 2017; published online on December 28, 2017. Discussion
period open until May 28, 2018; separate discussions must be submitted for
individual papers. This paper is part of the Journal of Geotechnical and
Geoenvironmental Engineering, © ASCE, ISSN 1090-0241.
Introduction
The overall geotechnical capacity of a pile is derived from the skin
friction and the base resistance. The design process begins with
evaluating moderately conservative soil parameters based on site
investigation test results. Depending on the type of soil, different
equations and methods for pile capacity can be used. For example,
for piles in clay the α-method and the method proposed by
Meyerhof (1965) are commonly used (e.g., in the United Kingdom)
to predict the ultimate skin friction and end bearing resistance re-
spectively. Other methods adopt direct correlations based on in situ
etration test (SPT)] (Eslami and Fellenius 1997), Laboratoire Central
des Ponts et Chausees (LCPC) (Bustamante and Gianeselli 1982),
and the Imperial College (IC) method (Jardine and Chow 1996).
More complex and rigorous numerical methods can also be used
for complicated pile problems such as piled groups (Kraft et al.
1981; Poulos 1989; Randolph 2003) and piled raft (Poulos and
Davis 1974; Kitiyodom and Matsumoto 2003) foundations. Never-
theless, all these methods are used in the design stage and therefore
they only provide an estimate of a pile’s behavior. As such, instru-
mented pile tests are recommended by standard codes of practice
[e.g., Clause 7.5 of Eurocode 7 (BSI 1995)] to quantify the perfor-
mance of a pile in order to validate the initial design assumptions.
General preliminary pile tests (McCabe and Lehane 2006) in-
clude a number of vibrating wire strain gauges (VWSG), either
in pairs or in threes at several levels within the pile, along with
a measurement of pile head settlement measured from an indepen-
dent reference beam by linear voltage distance transducers (LVDT).
This instrumentation scheme offers very useful but discrete data
points (Lehane et al. 1993). Data from appropriately monitored pile
load tests can provide a means to assess the behavior of the pile and
develop pile behavior models (Comodromos and Bareka 2009), such

© ASCE 04017122-1 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 as load transfer curves (Ménard 1963; Butterfield and Banergee
1971; Kraft et al. 1981; Frank and Zhao 1982; Poulos 1989; Lee
1993; Klar et al. 2006; Abchir et al. 2016; Seo et al. 2017).
Recent advances in geotechnical instrumentation include
fiber optic (FO) technology such as fiber Bragg gratings (FBG)
(Kersey and Morey 1993; Lee et al. 2004; Liu and Zhang 2012;
Doherty et al. 2015) and distributed Brillouin optical time-domain
reflectometry (BOTDR) (Kurashima et al. 1993; Soga 2014;
Pelecanos et al. 2017). The latter technology offers near spatially
continuous strain data along the entire length of the pile, which can
be further processed to provide detailed information regarding the
pile behavior, integrity, and load-transfer properties (Pelecanos and
Soga 2017; de Battista et al. 2016). The BOTDR technique has
been successfully used to monitor various soil–structure interaction
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.
problems (Acikgoz et al. 2016, 2017), including piles (Klar et al.
2006; Ouyang et al. 2015; Pelecanos et al. 2016), shafts/retaining
walls (Mohamad et al. 2011; Schwamb et al. 2014; Schwamb and
Soga 2015), tunnel linings (Mohamad et al. 2010, 2012; Cheung
et al. 2010; de Battista et al. 2015; Di Murro et al. 2016; Soga et al.
2017), tunneling and other geotechnical process-induced surface
settlements (Hauswirth et al. 2014; Klar et al. 2014; Linker and
Klar 2017), concrete cracking (Goldfeld and Klar 2013), soil
slopes, and so on.
This paper briefly discusses the BOTDR distributed monitoring
technology and explores its application in a number of pile load test
cases (both top-loaded using an external reference frame and bidir-
ectionally loaded using an Osterberg cell) in London. The monitor-
ing data from the distributed BOTDR and discrete VWSG
technologies in the three case studies are analyzed and compared
to shed light on the relative merits of each approach (continuous
and discrete) and to highlight their necessity in future reliable pile
load testing. Finally, numerical analyses are conducted for each of
the three piles; the results enable a better understanding of pile
behavior under loading.
Distributed Fiber Optic Monitoring
This section provides a brief description of the principles of
BOTDR. However, the complete description of the method and
Fig. 1. Principle of distributed fiber optic sensing using BOTDR
© ASCE 04017122-2
J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122
the associated experimental approaches required for calibration
are well beyond the scope of this paper and they are therefore
not included, because they are available elsewhere in great detail
(Mohamad 2007; Iten 2011; Soga 2014; Soga et al. 2015). More
information about the fundamentals of light propagation are avail-
able from relevant literature in the area of photonics (Horiguchi
et al. 1995), because this is out of scope of this paper. Kechavarzi
et al. (2016) gave a detailed description of the theory of distributed
FO strain sensing and its applications in civil and geotechnical
infrastructure.
Principle of Brillouin Optical Time-Domain
Reflectometry
A fiber optic cable allows light waves from a FO analyzer to travel
along its entire length through total internal reflection, irrespective
of the orientation of the cable itself. This allows a signal to be
carried over very long distances, such as for broadband Internet.
Backscattered signals are generated as the light wave passes
through the optical fiber and presents itself as Rayleigh, Raman,
and Brillouin spectrum. Within the Brillouin backscatter, the peak
frequency experiences a shift that is generally considered to be lin-
early proportional to applied strain. Using the measured time
required for the backscattered signal to return to the analyzer, the
specific location at which this frequency shift is observed can
be estimated accurately. Therefore the entire fiber optic cable
essentially serves as a distributed strain sensor (Fig. 1).
The FO analyzer sends light with a wavelength of 1,550 nm into
an optical fiber and the generated Brillouin spectrum of the back-
scattered light has 25–27 MHz bandwidth and approximately
11 GHz central peak frequency when no strain is applied on the
fiber. The backscattered Brillouin central frequency, vb , pro-
vided directly from the FO analyzer, is related to the input light
according to
2 · nf · va
vb ¼ ð1Þ
λl
where nf = fiber core refractive index; vα = acoustic velocity in the
fiber; and λl = wavelength of the input light.
J. Geotech. Geoenviron. Eng.
 Changes in temperature and/or strain induce a density change in
the cable and therefore a change in the acoustic velocity, vα , of the
light. As the strain or temperature at a given location change, the
frequency of the backscattered light is shifted by an amount which
is approximately linearly proportional to the applied strain, Δε, or
temperature, ΔT, according to Eq. (2)
Δvb ¼ Δvb0 þ M · Δε þ N · ΔT
where vb0 = central Brillouin peak frequency at zero strain and at a
given temperature; Δε = applied strain; ΔT = temperature change;
and M and N ¼ coefficients for strain and temperature change, re-
spectively. For an incident wavelength of 1,550 nm, the Brillouin
frequency shift can vary from 9 to 13 GHz depending on the differ-
ent fiber properties. Therefore knowledge about this frequency
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.
difference can provide information about the applied strain and
temperature changes at the location at which the backscattered light
was generated. Because the speed of light is constant, the location
can be evaluated by measuring the time because the light was ini-
tially sent into the fiber. Backscattered light is generated at every
point along the entire length of the fiber, and therefore resolving
both time and frequency can determine a continuous strain profile
along the fiber.
The case studies in this paper used either an AQ8603 analyzer
(Yokogawa Electric Corporation, Tokyo, Japan) or a NeubreScope
NBX-5000 analyzer (Neubrex, Kobe, Japan). These provide a min-
imum readout resolution between 0.05 and 0.1 m with a spatial
resolution of 0.5–1.0 m. Spatial resolution implies that the analyzer
4x125µm optical fibres
Steel wire with 250µm coating Strong nylon
1.2mm
5.2mm
(a)
Gel-filled tube Water blocking
Glass Yarns
Optical fibre
PE
(polyethylene)
outer Sheath
6.9mm
(b)
Fig. 2. Fiber optic cables used at the pile cases studied: (a) Fujikura
reinforced strain cable; (b) unitube temperature cable
© ASCE
J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122
produces a weighted average strain reading over 0.5 or 1 m at every
0.05 m length of the cable (this is considered spatially continuous
or distributed data). These settings can be changed depending on
the time allocated for the specific test. Essentially, the technology
offers a large number of strain data (every 0.05 to 0.1 m in this case)
along a structure embedded with fiber optic cables.
In addition to the clear advantage of measuring a full strain pro-
ð2Þ file, the simplicity of this method lies in the fact that only a single
cable is required for the entire system, enabling its use in small-
diameter piles and eliminating the time and effort for cable man-
agement that would be required for conventional strain gauges.
No electricity is required other than to power the analyzer itself,
which could be located much farther away in a safe and convenient
location on the construction site, because light waves travel effi-
ciently through the fiber optic cables. The result is an instrumen-
tation system which can provide a full strain profile of the pile.
Fiber Optic Cables
Strain on an optical fiber can be generated from two sources,
mechanical or thermal. Therefore two types of optical fiber cables
were installed (Fig. 2): Fujikura (Tokyo, Japan) four-core single-
mode fiber reinforced ribbon cable for strain sensing (strain sensing
cable) and Excel four-core single-mode fiber loose tube for temper-
ature compensation (temperature cable). Although they are both
attached to a reinforcement cage, the fiber optic cores of the tem-
perature cable sit in a gel which isolates any transfer of mechanical
strains from the outer coating. Thus it is only subjected to thermal
changes. These measurements are used to compensate the readings
measured from the strain cables to provide an accurate reading of
interest, the actual mechanical strain.
Installation of FO Instrumentation
Installation of FO cables is usually done on site (Fig. 3). Long pile
foundations typically consist of a number of steel reinforcement
cage segments, and therefore the bottom steel cage is instrumented
on the ground. The FO cables are run along the entire length of the
bottom segment on two opposite sides of the pile and a loop of FO
cable is made close to the bottom of the segment. The longitudinal
cables are prestrained (i.e., a tensile strain is applied) using cable
clamps at the two ends of the steel cage. Once the borehole is dug,
the bottom cage is inserted, and while the other cages are spliced
onto the bottom cage and the whole pile is lowered into the bore-
hole, the remaining FO cable is attached to them. Finally, the two
ends of the FO cable run from the top of the pile to the FO analyzer.
With the pile loaded axially, it is assumed that the concrete pile
will have negligible hoop strain across its cross section, and there-
fore a 10-m loop cable for both strain and temperature is prepared
and secured at the end of the bottom reinforcement cage to serve as
a zero-strain loop for referencing and compensation purposes.
For the ease of data interpretation, a prestrain of approximately
1,000–2,000 με is often introduced to the strain cable. Anchorage
is provided on the bottom loop end by cable wire clamps before
stretching the strain cable to the predetermined prestrain. Strain ca-
ble is then secured with another set of cable wire clamps at the top
of the reinforcement cage before supplementing the anchorage by
either spot gluing with epoxy glue or using cable ties at approxi-
mately every 0.5–1.0-m interval. Temperature cables are loosely
secured next to the strain cables with cable ties as they are routed
to the top of the cage. Figs. 4(a and b) show the installed FO cables
and sister-bar VWSGs on a foundation pile.
Once the bottom cage has been instrumented, it is lowered into
the borehole. The fiber optic cables are then unwound from the
04017122-3 J. Geotech. Geoenviron. Eng.
 higher
steel
cage
bottom steel cage FO analyser
machine FO cables

ground
g round g r o un d

FO cables
FO cables
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Schematic illustration of FO installation and monitoring of piled foundations

reels on each side of the borehole as the cage is lowered. Prestrain-

ing is carried out for the strain cables for subsequent reinforcement
cages as well without epoxy glue due to time constraints. Concrete
is subsequently poured in the borehole and as the concrete cures the
FO cables become securely embedded within the pile. Further
details of FO cable installation in piles established at University
of Cambridge are given by Klar et al. (2006), Soga (2014), and
Soga et al. (2015).

FO Data Analysis
As described previously, applied strain causes a shift in the peak
Brillouin frequency in the optical fiber. Therefore measuring the
frequency difference obtains the applied strain on the cable. More-
over, because FO cables are able to detect strains caused by both
mechanical and thermal loads, the two components need to be an-
alyzed separately. The measured frequency difference from the
temperature cable, ΔvbT , is influenced only by changes in temper-
ature, whereas that from the strain cable, ΔvbS , is influenced by
changes in both mechanical load and temperature.
Therefore changes in temperature, ΔT, can be obtained from
ΔvbT
ΔT ¼ ð3Þ
CTT

where CTT = property of the cable obtained by calibrating the tem-

perature cable, which determines how temperature affects the
Brillouin frequency reading of the cable, and is usually approxi-
mately 1.1 × 10−3 GHz=°C.
The thermal strain, εtemp , which is the strain that corresponds to
free thermal expansion strain due to temperature change, is then
given by
εtemp ¼ ac · ΔT ð4Þ

where αc = thermal expansion coefficient of concrete, and is usu-

Fig. 4. View of installed fiber optic cables and vibrating wire strain
ally approximately 9.65 με=°C.
gauges on the pile cage: (a) detailed view of clamp; (b) general view
The real (observed) strain, εreal , which is the actual strain that the
of installed sensors
pile experiences in the field, is then given by

© ASCE 04017122-4 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 1
εreal ¼ ðΔvbS − CT · ΔTÞ ð5Þ
CE
where CE = property of the fiber obtained by calibrating the strain
cable, which determines how strain affects the Brillouin frequency,
and is usually approximately 5 × 10−4 GHz=με; and CT = property
of the fiber that determines how the Brillouin frequency is
affected by temperature difference, and is usually approximately
1.0 × 10−3 GHz=°C.
The mechanical (constrained) strain, εmech , which is the reaction
strain that is the result of both the applied mechanical load and
temperature, is then given by

 
1 ΔvbT ΔvbT
εmech ¼ εreal − εtemp ¼ ΔvbS − CT · − ac ·
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

CE CTT CTT
ð6Þ

Finally, once the strain profiles are obtained, the actual geotech-
nical response of the pile may be captured using Eqs. (7) and (8)
to determine axial force, Fa ðyÞ, and vertical displacement, u(y),
profiles, respectively
Fa ðyÞ ¼ EA · εmech ðyÞ ð7Þ (a)
Z y Broadgate pile test
uðyÞ ¼ uðy ¼ y0 Þ þ εreal ðyÞdy ð8Þ 2500
0

where EA = axial rigidity of the pile (E = Young’s modulus and A =

2000
cross-sectional area); and y = depth from the top of the pile. For the
vertical displacements, the relative displacements obtained from the
Force, P [kN]

integration of axial strains is added to available absolute displace- 1500

ment values from displacement transducers at y0 . The data profiles
obtained from BOTDR usually have a wavy nature and therefore
1000
need to be filtered prior to data analysis. The data in this study were
filtered using a second-order Savitzky–Golay (1964) filter with a
31-point frame. 500
Applied load

Case Study 1: Pile Load Test at Broadgate Road, 0

0 10 20 30 40 50 60 70
London (b) Time, t [hours]

Description of Pile Test
The Broadgate Road project in London was designed to house a
14-story office building with two basement levels. Because of tight
space restrictions along one side of the project, a number of mini
piles 0.305 m in diameter were constructed in close proximity to
support the superstructure. A high-strength steel reinforcing case
was inserted in the ground after the drilling process. The pile tested
is 0.305 m in diameter (0.343 m at the top 6 m because of a steel
casing around the pile) and 25 m long [Fig. 5(a)]. The figure also
includes the soil stratigraphy with some known material properties
obtained from relevant triaxial and simple shear laboratory tests.
The pile test was carried out once the concrete material achieved
a specified value of minimum strength. The pile test consisted of
three consecutive cycles of applied load (at the top of the pile) of up
to 720, 1,080, and 1,985 kN for each of the three cycles, achieved
after several loading and unloading steps [Fig. 5(b)]. The pile was
instrumented with distributed FO cables on two opposite sides of
the pile and a number of discrete VWSGs along the pile depth.
Data Interpretation
Fig. 6(a) shows the axial strain in the pile for the three peak values
of the three cycles captured by the FO cables and the VWSGs,
Fig. 5. Description of Case 1—Broadgate pile load test case: (a) pile
geometry and soil stratigraphy; (b) test schedule
whereas Fig. 6(b) shows the corresponding axial force profiles cal-
culated from strains multiplied by the pile axial rigidity, EA, as
described by Eq. (7) and using E ¼ 30,000 MPa. This value
adopted for E was obtained following the approach of Fellenius
(1989) and by using the FO strain values, ε, at the top 30 cm of
the pile, which was surrounded by soil but with insignificant
influence [Figs. 6(a), 9(a), and 12(a)] and the applied loads, P,
(E ¼ ΔP=Δε=A). The Fellenius method proposes a smooth linear
(best-fit) reduction of secant modulus with axial strains. Therefore
a representative average value of E over the dominant experienced
strains (∼300–700 με) was adopted based on that best-fit line.
There was generally good agreement between the two monitoring
technologies. No VWSG data were obtained for the largest cycle
(i.e., for loading of 1,985 kN), because the VWSG instruments mal-
functioned, and therefore only FO data were available for this load
case. There was some scatter in the FO data values which is cur-
rently a known issue with distributed FO strain-sensing systems;
the standard resolution of FO is constant, approximately 30–50 με,
and therefore this becomes relatively less significant for larger

© ASCE 04017122-5 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 applied loads (which imply larger induced strains). The waviness of nature of the soil-spring, the external load was applied incremen-
FO strains may offer a challenge when differentiating strain data tally and the equations were solved using an iterative modified
profiles to obtain shaft friction values, but their spatial continuity Newton–Raphson technique. A number of different soil layers, as-
allows for a distributed sensing of localized strains, e.g., necking, sociated with constant soil spring properties along the depth of each
fracture, and so on, whereas such localized features would not be layer, were considered based on the ground conditions, although
identified by discrete monitoring systems such as VWSGs. Fig. 6(c) they did not follow exactly the soil stratigraphy. Appendix I ex-
shows the vertical displacements, u, of the pile from the FO cables. plains this simplified FE analysis approach in more detail. The
The values from the FOs were obtained by integrating the strain behavior of the pile was back-analyzed to derive the properties
profiles and adding those to the absolute displacement measure- of the soil springs which were subsequently used in the FE analysis
ments from displacement transducers at the top of the pile to calculate the axial strain and vertical displacement profiles.
[Eq. (8)]. Specifically, the optimum set of properties of the soil springs was
Fig. 6 includes the results of a simplified numerical finite- obtained that was able to reproduce well the observed axial strain
element (FE) beam-spring model for comparison. The simplified and vertical displacement profiles from the FO readings. The values
FE analysis considered a single vertical pile loaded axially from of the model parameters were obtained through a simple optimiza-
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

the top, modeled with linear beam elements, and it represented tion algorithm, the Levenberg–Marquardt scheme (Levenberg
the surrounding soil with nonlinear springs, which is a practical 1944; Marduardt 1963), in which the changing variables were
approach as opposed to the more common way of modeling the the set of the model parameters—in this case, 20 parameters, 4
soil with solid elements. All the beam elements and nonlinear for each of the 4 layers and 1 for the pile base—and the objective
springs contributed to the global stiffness matrix and therefore function was the difference of the axial strains obtained from the
to the global FE equilibrium equations. Because of the nonlinear numerical model and those observed from the FOs. A good match

P=720kN P=1080kN P=1985kN P=720kN P=1080kN P=1985kN

0 0 0 0 0 0

5 5 5 5 5 5
Depth from pile top [m]

Depth from pile top [m]

Depth from pile top [m]

Depth from pile top [m]

Depth from pile top [m]

Depth from pile top [m]

10 10 10 10 10 10

15 15 15 15 15 15

FE
20 20 20 20 FO
20 20
FO
FE
VWSG
VWSG

25 25 25 25 25 25
−500 0 500 −1000 0 1000 −1000 −500 0 −1 0 1 −2 0 2 −4 −2 0
(a) Axial Strain [με] Axial Strain [με] Axial Strain [με] (b) Axial Force [MN] Axial Force [MN] Axial Force [MN]

P=720kN P=1080kN P=1985kN

0 0 0
FO
FE

5 5 5
Depth from pile top [m]

Depth from pile top [m]

Depth from pile top [m]

10 10 10

15 15 15

20 20 20

25 25 25
0.000 0.005 0.010 0.000 0.010 0.020 0.000 0.020 0.040
(c) Vertical Displacement [m]

Fig. 6. Monitored data profiles for Case 1—Broadgate: (a) axial strain; (b) axial force; (c) vertical displacement

© ASCE 04017122-6 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 P=720kN P=1440kN P=1985kN
0 0 0 200 10
FE
Layer 1
Layer 2
5 5 5 Layer 3
Layer 4

Shaft Friction, SF [kPa]

Depth from pile top [m]

Depth from pile top [m]

Depth from pile top [m]

Base pressure, qb [MPa]

Base

10 10 10

100 5

15 15 15

20 20 20
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

25 25 25 0 0
0 50 0 50 100 0 100 200 0 0.5 1 1.5 2
(a) Shaft Friction [kPa] (b) Applied load, P [MN]

200 10 3
API-Layer 1
Layer 1 API-Layer 2
Layer 2 API-Layer 3
Layer 3 API-Layer 4
API-Base
Shaft Friction, SF [kPa]

Layer 4 150
Base pressure, q [MPa]

Shaft Friction, SF [kPa]

Base pressure, q b [MPa]

Base b 2

100 5 100

1
50

0 0 0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035
(c) Displacement, uz [m] (d) Displacement, u z [m]

Fig. 7. Calculated pile shaft friction from FE analysis for Case 1—Broadgate: (a) shaft friction profiles; (b) shaft friction development with applied
load; (c) shaft friction development with vertical displacement; (d) relevant API t-z and q-z curves

was obtained between the field data (from FO and VWSGs) and the variation of FO strain data led to a large noise:signal ratio, and
FE back-calculations. therefore the evaluation of SF (i.e., determination of the actual
Fig. 7(a) shows the shaft friction (SF) profiles for the three peak slope of the strain profiles) values may become cumbersome.
values of the three cycles calculated from the FE analysis. Because Furthermore, Figs. 7(b and c) show the evolution of SF with the
the FO data exhibit some (inherent) undulations, deriving SF values applied load, P, and the local vertical displacement, u, at various
from the slope of the axial force might be cumbersome. Therefore depths (according to the local soil stratigraphy) along the pile
this study followed a synthetic approach, in which a numerical and the pile base pressure, qb . The SF was mobilized early in
model was established that reproduced accurately the monitored the test, whereas the pile base pressure was mobilized at later stages
axial strain and vertical displacements from FOs (Fig. 6) and then for higher loads (Fig. 7). As expected, the SF development curves
SF profiles were obtained from the FE analysis of the model. This show an initial stiffness that decreased with the displacement, due
study followed this numerical analysis approach because of the in- to the plasticity of the soil close to the pile shaft. The first layer
ability to obtain SF values directly from the wavy FO strains. In (0–6 m), which is covered by the pile casing, did not show signifi-
fact, direct estimation of SF requires differentiation of axial strains, cant development of strains and reached an ultimate value of SF of
which in the case of wavy strain profiles leads to unrealistically approximately 20 kPa. In addition, although the three layers con-
large fluctuations of SF values with the depth of the pile. Generally sidered within the London Clay showed variable SF development,
larger SF values were obtained within the London Clay stratum it is accepted that the majority of the London Clay reaches SF of
(i.e., at z < −4 m) compared with the SF observed at the top soil approximately 70–100 kPa, whereas the bottom of the London
layers (i.e. at z > −4 m). However, at the bottom of the pile, very Clay shows minimal development of SF. However, this likely was
small SF values were mobilized, perhaps due to the small strains due to the small layer thickness considered in the data analysis—
experienced by the pile. Because a numerical optimization pro- the FO data existed in Layer 4 between y ¼ 19–22.5 m.
cedure was followed to obtain the SF, the small values of strains Nevertheless, in general, the evolution of shaft friction with the
experienced at the bottom of the pile compared with the usual vertical displacements seemed to reach (roughly) a plateau for

© ASCE 04017122-7 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 displacements of approximately 0.01–0.03 m, which is slightly less
than 10% of the pile diameter.
Finally, Fig. 7(d) shows that the relevant design t-z and q-z
curves following the API (2002) methodology (Appendix II).
Although there are some differences between the observed [Fig. 7(c)]
and design [Fig. 7(d)] curves, in general they seem to agree quite
well, yielding comparable values of ultimate pile shaft and base
resistance.

Remarks
A typical interpretation of the geotechnical data would consider
Eqs. (9) and (10) to calculate the ultimate shaft capacity, qs ,
and Eqs. (11) and (12) for the base capacity, qb of the pile
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

(Salgado 2008; Knappett and Craig 2012; Tomlinson and

Woodward 2014)
qsðcohesiveÞ ¼ α · Su ð9Þ

0 0
qsðnon-cohesiveÞ ¼ β · σvo ¼ K o · tan δ · σvo ð10Þ
(a)
qbðcohesiveÞ ¼ N c · Su ð11Þ
East Village pile test
25
qbðnon-cohesiveÞ ¼ N q · 0
σvo ð12Þ Applied load

where α = empirical shaft coefficient, usually approximately 0.5 for

20
London Clay (Tomlinson 1997); K o = earth pressure at rest; δ =
pile–soil interface friction angle, usually approximately 0.75ϕ
(Stas and Kulhawy 1984); and N q and N c = base bearing capacity
Force, P [MN]

coefficients, usually approximately 50 (Berezantzev et al. 1961; 15

Knappett and Craig 2012) and 9 (Kulhawy and Prakoso 1999),
respectively.
Using the preceding equations and the geotechnical data in Fig. 5 10
obtained an ultimate value of shaft capacity of 8 kPa for the first
layer using Eq. (10) and approximately 32–120 kPa for the second
layer using Eq. (9). These values compare well with the calculated 5
values from FO in Fig. 7, which suggested approximately 20 kPa
for the first layer and approximately 10–120 kPa for the second
layer. Back-calculating the values of α and β, the first layer yielded 0
a value of β ¼ 0.5, whereas Eq. (10) yielded β ¼ 0.2; Layers 2–4 0 10 20 30 40 50 60 70 80
yielded values of α ¼ 0.9, 0.89, 0.1, respectively, whereas the (b) Time, t [hours]
common assumption is 0.5 (Tomlinson 1997).
Fig. 8. Description of Case 2—East Village pile load test case: (a) pile
Similarly, calculating the ultimate base capacity obtain approx-
geometry and soil stratigraphy; (b) test schedule
imately 2 MPa using Eq. (11) (i.e., based on Su ) and approxi-
mately 25 MPa using Eq. (12) (i.e., based on c and ϕ). These
values are different from and, respectively, below and well above
the (linear) 6 MPa observed from the FOs during this test. This
was unexpected and could be due to a number of possible reasons, Case Study 2: Pile Load Test at East Village,
e.g., it may suggest that the relation usually used for the pile base London
bearing capacity [Eq. (11)] might be significantly unconservative,
or it may suggest that the material parameter values used to Description of Pile Test
calculate σv00 were too small. Nevertheless, Eq. (11) (i.e., S )
u
provides a better estimate. The second case study considered a pile test at East Village (former
The ability to fit a numerical model to the monitoring data (in Athletes Village) in Stratford, London. The examined pile is 32 m
particular, the continuous vertical displacement profile) to further long with a 900-mm diameter (930 mm at the top 14 m). The local
understand the behavior of piles is a great advantage. More con- soil stratigraphy consists of made ground, alluvium, and river ter-
fidence in the results of the back-analyzed model is built when a race deposit finishing at approximately 14 m deep and along which
continuous strain profile is available which can show the full pic- the pile is covered by a 15-mm-thick steel casing. These layers are
ture of the strains over the whole length of the pile and can by followed by two thick layers of Lambeth Group and Thanet Sand
direct integration give reliable estimates of pile displacements. that interface at a depth of 23 m. Fig. 8(a) gives information about
Finally, the benefits of obtaining such a relevant numerical pile geometry, soil stratigraphy and some basic soil properties. The
model can include the development of load transfer curves derived pile test consisted of a static maintained load applied at the top of
from the calculated shaft friction with respect to the vertical the pile following two cycles of loading-unloading until the pile
displacement. failed. Fig. 8(b) show details of the pile test sequence. Similar

© ASCE 04017122-8 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 P=6MN P=12.25MN P=17.25MN P=6MN P=12.25MN P=17.25MN
0 0 0 0 0 0

5 5 5 5 5 5

10 10 10 10 10 10
Depth from pile top [m]

Depth from pile top [m]

15 15 15 15 15 15

20 20 20 20 20 20

25 25 25 25 25 25
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

FE
30 30 30 30 FE 30 30
FO
FO
VWSG
VWSG
35 35 35 35 35 35
−400 −200 0 −1000 −500 0 −1000 −500 0 −10 −5 0 −20 −10 0 −40 −20 0
(a) Axial Strain [με] (b) Axial Force [MN]

P=6MN P=12.25MN P=17.25MN

0 0 0

5 5 5

10 10 10
Depth from pile top [m]

15 15 15

20 20 20

25 25 25

30 30 30
FE
FO
35 35 35
0 0.005 0.01 0 0.02 0.04 0 0.05
(c) Vertical Displacement [m]

Fig. 9. Monitored data profiles for Case 2—East Village: (a) axial strain; (b) axial force; (c) vertical displacement

to the previous case, the pile was instrumented with distributed hence of the axial stiffness EA. This analysis was conducted to
FO cables and discrete VWSGs; the latter were installed at various match the observed axial strains and vertical displacements in
locations along the pile depth. Figs. 9(a and c). The latter figure shows that the vertical displace-
ments obtained by the direct integration of the observed axial
strains matched the displacements resulting from the FE model that
Data Interpretation reproduced the axial strains.
Figs. 9(a and b) show the monitored axial strains and the calculated Fig. 10(a) shows the calculated shaft friction profiles for the
axial force in the pile for three selected load stages from both the three selected load cases determined from the FE analysis. Again,
FOs and the VWSGs. Similar to the previous case, although the these were obtained from the FE model that was calibrated to re-
FOs showed some scatter in the data, good agreement was obtained produce accurately the monitored axial strain and vertical displace-
between the two sensors for both strains and forces. Observed ments from FOs (Fig. 9). Furthermore, Figs. 10(b and c) show the
strains and forces were roughly constant for the first 14 m, which evolution of SF with the applied load, P, and the local vertical dis-
suggests that minor shaft friction develops over that depth. This placement, u, at three selected depths along the pile, according to
was expected because the pile is surrounded by a steel casing at the local soil stratigraphy, i.e., in the shallow layers (covered with
the top 14 m. Moreover, at depths below 14 m, the axial strains pile casing), Lambeth Group and Thanet Sand. The first layer,
and forces decreased, which was due to the interaction with the which is covered by the pile casing, did not show significant devel-
surrounding soil and the developed soil–pile interface friction. opment of strains and reached an ultimate value of SF of approx-
Additionally, Figs. 9(a and b) include the results of a simple FE imately 40 kPa. In contrast, Lambeth Group and Thanet Sand
analysis similar to that used in the first case (Appendix I); the strain exhibited a larger development of SF that reached approximately
step in the first figure is due to the change of pile diameter and 200 and 110 kPa, respectively. This difference was expected

© ASCE 04017122-9 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 P=6MN P=12.25MN P=17.25MN
0 0 0 200 20
FE
Layer 1
5 5 5 Layer 2
Layer 3

Shaft Friction, SF [kPa]

Base
Depth from pile top [m]

Base pressure, q [MPa]

10 10 10

b
15 15 15
100 10
20 20 20

25 25 25

30 30 30
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

35 35 35 0 0
0 50 100 0 100 200 0 100 200 0 2 4 6 8 10 12 14 16 18
(a) Shaft Friction [kPa] (b) Applied load, P [MN]

200 20 30
Layer 1 API-Layer 1
Layer 2 API-Layer 2
API-Layer 3
Layer 3
API-Base
Base
Shaft Friction, SF [kPa]

150
Base pressure, q [MPa]

Base pressure, q b [MPa]

Shaft Friction, SF [kPa]
20
b

100 10 100

10

50

0 0 0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
(c) Displacement, u [m] (d) Displacement, u z [m]
z

Fig. 10. Calculated pile shaft friction from FE analysis for Case 2—East Village: (a) shaft friction profiles; (b) shaft friction development with applied
load; (c) shaft friction development with vertical displacement; (d) relevant API t-z and q-z curves

because the pile in the latter two layers was not covered with a steel Back-calculating the values of α and β, the first layer yielded a
casing and therefore pile–soil interaction friction developed, value of β ¼ 0.21, whereas Eq. (10) yielded approximately
resisting the pile movement. In general, as expected, the SF devel- β ¼ 0.2; Layer 2 yielded a value of α ¼ 0.8, whereas the common
opment curves show an initial stiffness that decreased with the dis- assumption is 0.5 (Tomlinson 1997); and the third layer yielded a
placement, due to the plastic deformation of the soil close to the pile value of β ¼ 0.2, in agreement with Eq. (10), which yielded ap-
shaft. Finally, SF was mobilized early in the test, whereas the pile proximately β ¼ 0.2). Therefore the β-method seemed to work
base pressure was mobilized at later stages for higher loads. well, whereas the appropriate value for α was slightly larger than
Fig. 10(d) shows the relevant design t-z and q-z curves follow- that commonly used (0.5).
ing the API (2002) methodology. Although there are some differ- Similarly, calculating the ultimate base capacity obtained ap-
ences between the observed [Fig. 10(c)] and design [Fig. 10(d)] proximately 27 MPa using Eq. (12) (i.e., based on c and ϕ), which
curves, in general they seem to agree quite well, yielding compa- is well above the (roughly linear) 12 MPa observed from the FOs
rable values of ultimate pile shaft resistance. during this test. Again, the relation for the base capacity seemed to
significantly overestimate the observed pile base capacity.

Remarks
Case Study 3: Osterberg Cell Pile Test at Francis
Using Eqs. (9)–(12) and the geotechnical data in Fig. 8 obtained an Crick Institute, London
ultimate value of shaft capacity of 29 kPa for the first layer using
Eq. (10), 32–219 kPa for the second layer using Eq. (9), and ap-
proximately 116 kPa for the third layer using Eq. (10). These values Description of Pile Test
compare very well with the observed values from FO in Fig. 10, This particular case study focused on the behavior of a 31.5-m-
which suggest approximately 30, 200, and 110 kPa for the three long, 1,500-mm-diameter bored pile during a preliminary load test
layers. at the Francis Crick Institute. This is a biomedical research centre

© ASCE 04017122-10 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122


Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.
(a)
Francis Crick pile test
9 high values of strain at a depth of approximately 18–23 m. This
8 Applied load
7
6
Force, P [MN]
5 mottled beds in the Lambeth Group at approximately 18–19 m
4 records, the presence of sandy glauconitic clay may have caused
3 may have caused necking of the pile (smaller cross section).
2
1
0 Computing the axial force by multiplying the axial strains with
0 2 4 6 8 10
(b) Time, t [hours]
Fig. 11. Description of Case 3—Francis Crick pile load test case:
(a) pile geometry and soil stratigraphy; (b) test schedule
situated next to St. Pancras International train station in the London
Borough of Camden. One of the key differences from the previous
case studies was the loading mechanism. Bidirectional Osterberg
cells (O-cells) (Osterberg and Pepper 1984) were used to apply load
from the bottom of the pile. Fig. 11(a) shows the geometry of the
pile and the local stratigraphy; the ground consists of two thick
layers of London Clay and Lambeth Group, with varying undrained
strength, overlying Thanet Sand.
Similar to the previous case study, Fujikura reinforced ribbon
cable (JBT-03813) and eight-core single-mode fiber (205-301
Excel OS1 8C 9/125 Loose Tube LSOH Black) were used to mea-
sure strain and temperature, respectively. The installation process
was identical to Case Study 1; both fiber optic cables were routed
along opposite sides of the reinforcement cage from the pile head to
the top of O-cell where a 10-m-long reference loop was located. A
prestrain of 2,000 με was induced in the strain cable during instal-
lation. Anchorage was provided by IC-ROC clamps manufactured
by Fujikura. To serve as a comparison, VWSG was installed at five
levels along the pile depth. The pile test consisted of a single load
© ASCE
J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122
cycle reaching a maximum of 8.33 MN after seven loading steps
and then unloading to zero after three steps [Fig. 11(b)].
Data Interpretation
Fig. 12(a) shows the measured axial strain profiles of the pile for
three selected load stages from FO and VWSGs. Considering first
the VWSGs only, as expected, large values of strain occurred at the
bottom of the pile (close to the O-cell) and smaller values occurred
at the top. Interestingly, at a depth of approximately 19 m there was
a significantly higher value of VWSG strain which, in practice,
could be considered as not representative of the actual strains in
the pile and therefore ignored and discarded by the design engi-
neers. Eliminating outliers that do not conform to the expected
ranges is common in data interpretation because instrument mal-
functions do occur occasionally. Signs of VWSG malfunction
may not always be clear, and these anomalies can be caused by
a number of scenarios, such as cable damage. In some cases,
the recorded data are in fact a true representation which can be
attributed to changes in ground conditions and construction quality.
However, the fiber optic cable produced some unexpectedly
is unexpected when the pile diameter is uniform at that depth,
and therefore no step was expected in the axial strains. The data
indicated that the pile sustained high localized strains in that region.
Similarities in trend for both systems triggered further investigation
into the soil strata, where the nearest borehole log (BH04) (distance
∼10 m) recorded a change in soil layers from lignite beds and lower
deep. Although such a scenario was not reported in the construction
a localized collapse during the construction of the pile, which
Subsequently, the cross-sectional area, A, as well as the integrity
of the concrete (e.g., Young’s modulus, E) at this location would
have been compromised. Therefore a much higher strain reading
would be very likely (εa ¼ Fa =EA).
12 a constant axial stiffness, EA, would therefore be unrealistic.
This study used a FE model in which the axial rigidity, EA, of
the pile was kept constant along the pile depth except at a depth
of 18–23 m, at which it was reduced. A parametric study showed
that when EA at that location was reduced to 35% of the initial EA
(E ¼ 30,000 MPa and A ¼ 0.25πd2 , where d is the design diam-
eter shown in Fig. 11), a good match was obtained between the
axial strains [Fig. 12(a)] and the vertical displacements [Fig. 12(c)].
This apparent reduction in EA could be due to some pile necking
(smaller A) or some mixing of the pile concrete with adjacent
ground materials (smaller E). Then, using the results of the FE
model, the axial force profiles in the pile were calculated by multi-
plying the axial strains by EA everywhere except at depth 18–23 m,
where 0.35EA was used. Fig. 11(b) shows the latter profiles along
with the axial force from the FE model. Good comparison was ob-
tained between the two monitoring instruments and the relevant
numerical analysis. The axial force profiles with the nonuniform
EA varied smoothly with the depth (in contrast to the axial strain
profiles); this was expected because of force equilibrium (because
the soil spring stiffness values did not change). The use of 0.35EA
for the pile analysis was not ideal and it was literally obtained from
a back-analysis matching the observed strain profiles. Another
option would be to conduct a series of solid FE analyses—e.g.,
two-dimensional (2D) axisymmetric analyses—which consider
different values of reduced E (of blended concrete and soil) and
reduced A (i.e., reduced d2 ).
04017122-11 J. Geotech. Geoenviron. Eng.
 P=4.75MN P=7.14MN P=8.33MN P=4.75MN P=7.14MN P=8.33MN
0 0 0 0 0 0

FE FE
5 5FO 5 5 5
FO 5
VWSG VWSG

10 10 10 10 10 10
Depth, z [m]

Depth, z [m]
15 15 15 15 15 15

20 20 20 20 20 20
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

25 25 25 25 25 25

30 30 30 30 30 30
−500 0 500 −500 0 500 −500 0 500 −10 0 10 −20 0 20 −20 0 20
(a) Axial strain, ε [με] (b) Axial force, F [MN]
a a

P=4.75MN P=7.14MN P=8.33MN

0 0 0

FE
5 5 5
FO
Depth from pile top [m]

10 10 10

15 15 15

20 20 20

25 25 25

30 30 30
0 0.005 0.01 0 0.01 0.02 0 0.01 0.02
(c) Vertical Displacement, uz [m]

Fig. 12. Monitored data profiles for Case 3—FrancisCrick: (a) axial strain; (b) axial force; (c) vertical displacement

Fig. 13(a) shows the calculated shaft friction profiles for the Remarks
three chosen values of applied load from the FE analysis. Again,
The monitoring data from the FO cables agreed very well with the
these were obtained from the FE model that reproduced accurately
monitoring data from the VWSG. Moreover, the continuity of the
the monitored axial strain and vertical displacements from FOs
FO data highlighted a region of localized high strain development
(Fig. 12). Furthermore, Figs. 13(b and c) show the evolution of
which spread across 6–8 m in the pile shaft. A high value of strain
SF with the applied load, P, and the local vertical displacement,
u, at two selected depths along the pile, according to the local soil was also captured by the VWSG sensors at the same depth, but
stratigraphy, i.e. the London Clay and the Lambeth Group. The because this was only a single value it could easily have been
Lambeth Group, which is deeper and closer to the O-cell, exhibited ignored and its significant difference from the other data points
early development of shaft friction with the applied load, P, and had could have been erroneously attributed to instrument malfunction.
a stiffer response than the upper London Clay, which seemed to However, the presence of continuous FO data was able to support
reach a SF plateau of approximately 35 kPa at approximately the localized high values of strain, which might be due to some low-
0.01 m displacement. In contrast, Lambeth Group showed an in- quality concrete material of the pile or some mixing of ground
creasing development of SF which did not reach an ultimate value material with pile concrete.
in this test. Moreover, the availability of these monitoring data allowed the
Finally, Fig. 13(d) shows the relevant design t-z curves follow- derivation of shaft friction development curves with the applied
ing the API (2002) methodology. Once again, although there are load or vertical displacement. These curves show that the devel-
some differences between the observed [Fig. 13(c)] and design oped shaft friction in the deeper soil layers (e.g., Lambeth Group),
[Fig. 13(d)] curves, in general they seem to agree quite well, pro- i.e., closer to the O-cell was, as expected, higher than the corre-
viding similar values of ultimate pile shaft resistance. sponding friction at the top of the pile, close to the ground surface.

© ASCE 04017122-12 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 P=4.75MN P=7.14MN P=8.33MN
0 0 0 150
FE
Layer 1
5 5 5 Layer 2

Shaft Friction, SF [kPa]

Depth from pile top [m]

10 10 10 100

15 15 15

20 20 20 50
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

25 25 25

30 30 30 0
0 50 100 0 100 200 0 100 200 0 1 2 3 4 5 6 7 8 9
(a) Shaft Friction, SF [kPa] (b) Applied load, P [MN]

150 120
API-Layer 1
Layer 1 API-Layer 2
Layer 2 100
Shaft Friction, SF [kPa]

Shaft Friction, SF [kPa]

100 80

60

50 40

20

0 0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0 0.005 0.01 0.015
(c) Displacement, u [m]
z (d) Displacement, u z [m]

Fig. 13. Calculated pile shaft friction from FE analysis for Case 3—Francis Crick: (a) shaft friction profiles; (b) shaft friction development with
applied load; (c) shaft friction development with vertical displacement; (d) relevant API t-z curves

Using Eq. (9) (i.e., based on Su ) and the geotechnical data • The BOTDR distributed monitoring system is able to provide a
in Fig. 11 obtained an ultimate value of shaft capacity of continuous profile of the induced strain within piles, and this
23–98 kPa for the first layer and approximately 100–121 kPa offers more confidence in determining the developed shaft fric-
for the second layer. These values compare well with the observed tion profiles along the pile. The availability of continuous strain
values from FO in Fig. 13, which suggest average values of approx- measurements offers a clear view of the condition of the entire
imately 30 and 150 kPa for the two layers. The shaft friction values pile and hence provides an indication of any localized regions of
interpreted from the observed FO data are very close to the ex- weakness, shaft area inhomogeneity, or strain concentration.
pected design based on Eq. (9). Finally, back-calculating the values This is clearly a limitation of discrete monitoring systems such
of α in Eq. (9) for the two layers yielded values of α ¼ 0.25 as VWSG, which do not provide adequate information for the
and 0.68, respectively, whereas the common assumption is 0.5 whole length of the pile.
(Tomlinson 1997). • The distributed FO data provided reliable information about
vertical pile displacements by direct integration of the spatially
continuous strain data. The calculated displacements from the
Conclusions FO strains were verified against the displacements obtained
from a relevant FE model. Such vertical displacement profiles
This paper presented the application of distributed fiber optic strain are very useful in calibrating the model parameters of a
measurement technology for monitoring the actual field behavior of FE model.
axially loaded piles. The fiber optic data from three representative • An available and reliable set of monitoring data over the whole
case studies of pile load tests conducted recently in London were length of the pile allows an estimation of the shaft friction de-
analyzed and compared with spatially discrete point VWSGs and velopment curves with the applied load or vertical displacement
relevant simple finite-element analyses. The main findings of this (load-transfer) which may be used in future design of piles in a
study are the following: similar geographical region and soil stratigraphy.

© ASCE 04017122-13 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 P P
(y)

y
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

(a) (b) (c) (d) (e)

Fig. 14. Numerical analysis model of pile-soil interaction: (a) pile; (b) axial strain distribution; (c) top load-displacement; (d) numerical beam-spring
model; (e) load-transfer curve

• The obtained values of shaft friction and base resistance were Table 1. Parameters of the Numerical FE Beam-Spring Model for All
compared with expected values from existing methods of geo- Cases Considered
technical design (e.g., α and β methods) and were generally Depth km tm
found to be in good agreement. The observed values of α Case Layer (m) (MN=m3 ) (MN=m2 ) d h
and β were back-analyzed and were also found to be, in general,
Case 1— 1 0–6 8 0.011 2 0.8
in good agreement with the suggested values from the literature. Broadgate 2 6–12 14 0.157 0.9 1.5
• The obtained load-transfer (t-z and q-z) curves were compared pile 3 12–19 16 0.136 2.5 1
with design curves from the literature (API 2002). Although no- 4 19–25 2 0.008 1.2 1
table differences were observed regarding the pile base curves, Base 25 459 65,573 1 1
the pile shaft curves were generally in good agreement. Case 2—East 1 0–14 14 0.053 1 1
Village pile 2 14–23 37 0.223 1 1
3 23–32 24 0.195 1.6 1
Base 32 513 17.113 1 1
Appendix I. Finite-Element Model Formulation Case 3—Francis 1 0–21 21 0.036 3 1
Crick pile 2 1–25 57 0.117 3 0.7
Fig. 14 describes the numerical finite-element analysis used in this
paper. A vertical axially loaded pile was modeled with a series of
linear-elastic two-noded beam elements with vertical displacement
degrees-of-freedom only and a series of nonlinear springs, repre- with those observed in the field [Figs. 6(a and c), 9(a and c),
senting the surrounding soil, attached to each node. and 12(a and c)].
The behavior of the soil spring is governed by a nonlinear load- The equations satisfying the global equilibrium of the pile–soil
transfer curve that follows the degradation and hardening hyper- problem follow a standard static finite-element formulation (Bathe
bolic model (DHHM) of (Pelecanos and Soga 2017) described by 1996)
½K p þ K s  · fug ¼ fFg ð14Þ
km z
t ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ð13Þ where ½K p  and ½K s  = global pile and soil stiffness matrices respec-
d
ð1 þ ðktmm zÞhd Þ tively, which contain information about the geometry and the
material properties of the pile and soil, respectively; fug = vector
of the displacement degrees-of-freedom; and fFg = vector of the
where km = maximum stiffness for displacement; z ¼ 0 (units: externally applied forces.
force=length3 ); tm = maximum value of shear stress, t (maximum Boundary conditions applied consist only of the applied load,
only in the case of no hardening/softening, i.e., h ¼ 0) (units: which is specified as a known value in the fFg vector, at the first
force=length2 ); d ¼ unitless degradation parameter that governs node for a top-loaded pile or at the last node for a bottom-loaded
the degradation of subgrade modulus, k, with displacement, z;
O-cell test. Table 1 lists the numerical model parameters adopted
and h ¼ unitless hardening parameter that mostly governs the
for the analyses of the case studies presented in this paper, which,
model behavior at large displacements, z. Some t-z curves (see sec-
as explained previously, were obtained by matching the observed
tion “Introduction”) include the effect of the pile diameter as well,
pile responses.
but in the considered cases that mostly involved London Clay and
large pile diameters (i.e., no significant arching) it was expected
that the diameter did not affect the obtained t-z curves. Appendix II. API Load-Transfer Curves
The values of the four parameters of the model (km , tm , d, and h)
were obtained by matching the axial strain, εa ðzÞ, and vertical Table 2 lists the data used for the API (2002) curves in Figs. 7, 10,
displacement, u(z), profiles resulting from the numerical model and 13. These curves depend only on the soil material properties

© ASCE 04017122-14 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 Table 2. Data Used for the API (2002) Curves vb0 = central Brillouin frequency at zero strain and temperature
q-z for sand difference;
t-z for sand t-z for clay and clay vα = acoustic velocity in the optical fiber;
z (in.) t=tult z=D t=tult z=D q=q ult
y = depth;
z = local vertical displacement;
0 0 0 0 0 0
0.1 1 0.0016 0.3 0.002 0.25 α = adhesion factor;
0.4 1 0.0031 0.5 0.13 0.5 γ = soil unit weight;
0.0057 0.75 0.042 0.75 ΔT = temperature change;
0.008 0.9 0.073 0.9 ΔvbS = Brillouin frequency change reading from strain cable;
0.01 1 0.1 1
ΔvbT = Brillouin frequency change reading from temperature
0.02 0.9 0.2 1
0.03 0.9
cable;
εa = axial pile strain;
εmech = mechanical strain;
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.

εreal = real (observed) strain;

and the geometry (diameter, D) of the pile. The values for tult were εtemp = thermal expansion strain;
obtained by using Eqs. (9) and (10) for clay and sand, respectively, λl = wavelength of the input light; and
whereas those for qult were obtained by using Eqs. (11) and (12) for ν = Poisson’s ratio.
clay and sand, respectively.

Acknowledgments References

This research was conducted within the Centre for Smart Infra-
structure and Construction (CSIC) of the University of Cambridge,
funded by EPSRC and Innovate U.K. Their financial assistance is
gratefully acknowledged. The assistance of the CSIC team is ac-
knowledged, including Professor Lord Robert Mair, Dr. Jennifer
Schooling, Peter Knott, Jason Shardelow, and Jules Birks. Finally,
the authors acknowledge the contribution of the numerous CSIC
Industry partners, especially ARUP (Duncan Nicholson, Paul
Morrison, and Stuart Pennington), Cementation Skanska (Andrew
Bell, Martin Pedley, and Rab Fernie), and Laing O’Rourke.
Notation
The following symbols are used in this paper:
A = pile cross-sectional area;
CE = optical fiber parameter;
CT = optical fiber parameter;
CTT = optical fiber parameter;
D = pile diameter;
d = nonlinear model degradation parameter;
E = pile Young’s modulus;
Fa = axial pile force;
h = nonlinear model hardening parameter;
km = nonlinear model maximum subgrade modulus
parameter;
L = total length of pile;
M = optical fiber strain coefficient;
N = optical fiber temperature coefficient;
N c = pile end bearing capacity factor;
nf = fiber core refractive index;
P = top load value;
qb = pile base pressure;
r = pile radius;
SF = shaft friction;
Su = undrained soil shear strength;
t = nonlinear model shear stress parameter;
tm = nonlinear model maximum shear stress parameter;
u = vertical displacement;
vb = central Brillouin frequency;
Abchir, Z., Burlon, S., Frank, R., Habert, J., and Legrand, S. (2016). “t-z
curves for piles from pressuremeter test results.” Géotechnique, 66(2),
137–148.
Acikgoz, M. S., Pelecanos, L., Giardina, G., Aitken, J., and Soga, K.
(2017). “Distributed sensing of a masonry vault during nearby piling.”
Struct. Control Health Monitor., 24(3), e1872.
Acikgoz, M. S., Pelecanos, L., Giardina, G., and Soga, K. (2016). “Field
monitoring of piling effects on a nearby masonry vault using distributed
sensing.” Int. Conf. of Smart Infrastructure and Construction, ICE
Publishing, Cambridge, U.K.
API (American Petroleum Institute). (2002). “Planning, designing and
constructing fixed offshore platforms—Working stress design.” API RP
2A-WSD, Washington, DC.
Bathe, K. J. (1996). Finite element procedures, 1st Ed., Prentice Hall,
Englewood Cliffs, NJ.
Berezantzev, V. G., Khristoforov, V. S., and Golubkov, V. N. (1961). “Load
bearing capacity and deformation of piled foundations.” Proc., 5th Int.
Conf. on Soil Mechanics and Foundation Engineering, Paris.
British Standards Institution. (1995). “Part 1: General rules (together with
United Kingdom national application document).” Eurocode 7, London.
Bustamante, M., and Gianeselli, L. (1982). “Pile bearing capacity predic-
tions by means of static.” Proc., 2nd European Symp. on Penetration
Testing, Amsterdam, Netherlands.
Butterfield, R., and Banergee, P. K. (1971). “The elastic analysis of com-
pressible piles and pile groups.” Geotechnique, 21(1), 43–60.
Cheung, L., Soga, K., Bennett, P. J., Kobayashi, Y., Amatya, B., and
Wright, P. (2010). “Optical fibre strain measurement for tunnel lining
monitoring.” Proc. ICE Geotech. Eng., 163(3), 119–130.
Comodromos, E. M., and Bareka, S. V. (2009). “Response evaluation of
axially loaded fixed-head pile groups in clayey soils.” Int. J. Numer.
Anal. Methods Geomech., 33(17), 1839–1865.
de Battista, N., Elshafie, M. Z. E. B., Soga, K., Williamson, M., Hazelden,
G., and Hsu, Y. S. (2015). “Strain monitoring using embedded distrib-
uted fibre optic sensors in a sprayed concrete tunnel lining during the
excavation of cross-passages.” 7th Int. Conf. on Structural Health
Monitoring and Intelligent Infrastructure, London.
de Battista, N., Kechavarzi, C., Seo, H., Soga, K., and Pennington, S.
(2016). “Distributed fibre optic sensors for measuring strain and tem-
perature of cast-in-situ concrete test piles.” Proc., Int. Conf. on Smart
Infrastructure and Construction, Thomas Telford, Cambridge, U.K.
Di Murro, V., Pelecanos, L., Soga, K., Kechavarzi, C., Morton, R. F., and
Scibile, L. (2016). “Distributed fibre optic long-term monitoring of
concrete-lined tunnel section TT10 at CERN.” Int. Conf. of Smart Infra-
structure and Construction, ICE Publishing, Cambridge, U.K.
Doherty, P., et al. (2015). “Field validation of fibre Bragg grating sensors for
measuring strain on driven steel piles.” Géotechnique Lett., 5(2), 74–79.

© ASCE 04017122-15 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122

 Eslami, A., and Fellenius, B. H. (1997). “Pile capacity by direct CPT and
CPTU methods applied to 102 case histories.” Can. Geotech. J., 34(6),
886–904.
Fellenius, B. H. (1989). “Prediction of pile capacity.” ASCE Symp. on
Predicted and Observed Behavior of Piles, London.
Frank, R., and Zhao, S. R. (1982). “Estimating the settlement of axially
loaded bored piles in fine sand by PMT data.” Bull. Liaison LPC, 119.
Goldfeld, Y., and Klar, A. (2013). “Damage identification in reinforced
concrete beams using spatially distributed strain measurements.” J.
Struct. Eng., 10.1061/(ASCE)ST.1943-541X.0000795, 04013013-11.
Hauswirth, D., Puzrin, A. M., Carrera, A., Standing, J. R., and Wan, M. S. P.
(2014). “Use of fibre-optic sensors for simple assessment of ground sur-
face displacements during tunnelling.” Geotechnique, 64(10), 837–842.
Horiguchi, T., Shimizu, K., Kurashima, T., Tateda, M., and Koyamada, Y.
(1995). “Development of a distributed sensing technique using
Downloaded from ascelibrary.org by University of Exeter on 12/29/17. Copyright ASCE. For personal use only; all rights reserved.
Brillouin scattering.” J. Light-Wave Technol., 13(7), 1296–1302.
Iten, M. (2011). “Novel applications of distributed fiber-optic sensing in
geotechnical engineering.” Ph.D. thesis, ETH, Zurich, Switzerland.
Jardine, R. J., and Chow, F. C. (1996). New design methods for offshore
piles, Marine Technology Directorate, London.
Kechavarzi, C., Soga, K., de Battista, N., Pelecanos, L., Elshafie, M. Z. E. B.,
and Mair, R. J. (2016). Distributed fibre optic strain sensing for monitor-
ing civil infrastructure, Thomas Telford, London.
Kersey, A. D., and Morey, W. W. (1993). “Multiplexed Bragg grating fibre-
laser strain-sensor system with mode-locked interrogation.” Electron.
Lett, 29(1), 112–114.
Kitiyodom, P., and Matsumoto, T. (2003). “A simplified analysis method
for piled raft foundations in non-homogeneous soils.” Int. J. Numer.
Anal. Methods Geomech., 27(2), 85–109.
Klar, A., et al. (2006). “Distributed strain measurement for pile founda-
tions.” Proc. ICE Geotech. Eng., 159(3), 135–144.
Klar, A., Dromy, I., and Linker, R. (2014). “Monitoring tunneling induced
ground displacements using distributed fiber-optic sensing.” Tunnell.
Underground Space Technol., 40, 141–150.
Knappett, J. A., and Craig, R. F. (2012). Craig’s soil mechanics, 8th Ed.,
CRC Press, London.
Kraft, M. L., Ray, R. P., and Kakaaki, T. (1981). “Theoretical t-z curves.”
J. Geotech. Eng. Div., 107(11), 1543–1561.
Kulhawy, F. H., and Prakoso, W. A. (1999). “Discussion of end bearing
capacity of drilled shafts in rock.” J. Geotech. Eng., 10.1061/(ASCE)
1090-0241(1999)125:12(1106), 1106–1110.
Kurashima, T., Horiguchi, T., Izumita, H., and Tateda, M. (1993). “Bril-
louin optical-fiber time domain reflectometry.” IEICE Trans. Cummun.,
E76-B(4), 382–390.
Lee, C. Y. (1993). “Pile group settlement analysis by hybrid layer ap-
proach.” J. Geotech. Eng. Div., 10.1061/(ASCE)0733-9410(1993)
119:6(984), 984–997.
Lee, W., Lee, W., Lee, S., and Salgado, R. (2004). “Measurement of pile
load transfer using the fiber Bragg grating sensor system.” Can. Geo-
tech. J., 41(6), 1222–1232.
Lehane, B. M., Jardine, R. J., Bond, A. J., and Frank, R. (1993). “Mech-
anisms of shaft friction in sand from instrumented pile tests.” J. Geo-
tech. Eng., 10.1061/(ASCE)0733-9410(1993)119:1(19), 19–35.
Levenberg, K. (1944). “A method for the solution of certain non-linear
problems in least squares.” Q. Appl. Math., 2(2), 164–168.
Linker, R., and Klar, A. (2017). “Detection of sinkhole formation by strain
profile measurements using BOTDR: Simulation study.” J. Eng. Mech.,
10.1061/(ASCE)EM.1943-7889.0000963, B4015002.
Liu, J., and Zhang, M. (2012). “Measurement of residual force locked in
open-ended pipe pile using FBG-based sensors.” Electron. J. Geotech.
Eng., 17, 2145–2154.
Marduardt, D. (1963). “An algorithm for least-squares estimation of non-
linear parameters.” SIAM J. Appl. Math., 11(2), 431–441.
McCabe, B. A., and Lehane, B. M. (2006). “Behavior of axially loaded pile
groups driven in clayey silt.” J. Geotech. Geoenviron. Eng., 10.1061
/(ASCE)1090-0241(2006)132:3(401), 401–410.
Ménard, L. (1963). “Calcul de la force portante des fondations à partir des
essais pressiométriques.” Sols-Soils, 6, 9–27 (in French).
© ASCE 04017122-16
J. Geotech. Geoenviron. Eng., 2018, 144(3): 04017122
Meyerhof, G. G. (1965). “Shallow foundations.” J. Soil Mech. Found. Div.,
91(2), 21–31.
Mohamad, H. (2007). “Distributed optical fibre strain sensing of geotech-
nical structures.” Ph.D. thesis, Univ. of Cambridge, Cambridge, U.K.
Mohamad, H., Bennett, P. J., Soga, K., Mair, R. J., and Bowers, K. (2010).
“Behaviour of an old masonry tunnel due to tunnelling-induced ground
settlement.” Geotechnique, 60(12), 927–938.
Mohamad, H., Soga, K., Bennett, P. J., Mair, R. J., and Lim, C. S. (2012).
“Monitoring twin tunnel interaction using distributed optical fiber strain
measurements.” J. Geotech. Geoenviron. Eng., 10.1061/(ASCE)GT
.1943-5606.0000656, 957–967.
Mohamad, H., Soga, K., Pellew, A., and Bennett, P. J. (2011). “Perfor-
mance monitoring of a secant-piled wall using distributed fiber optic
strain sensing.” J. Geotech. Geoenviron. Eng., 10.1061/(ASCE)GT
.1943-5606.0000543, 1236–1243.
Osterberg, J. O., and Pepper, S. F. (1984). “A new simplified method for
load testing drilled shafts.” Found. Drill., 23(6), 9–11.
Ouyang, Y., Broadbent, K., Bell, A., Pelecanos, L., and Soga, K. (2015).
“The use of fibre optic instrumentation to monitor the O-Cell load test
on a single working pile in London.” Proc., XVI European Conf. on Soil
Mechanics and Geotechnical Engineering, London.
Pelecanos, L., et al. (2017). “Distributed fibre-optic monitoring of an
Osterberg-cell pile test in London.” Geotech. Lett., 7(2), 1–9.
Pelecanos, L., and Soga, K. (2017). “Using distributed strain data to evalu-
ate load-transfer curves for axially loaded piles.” J. Geotech. Geoen-
viron. Eng., in press.
Pelecanos, L., Soga, K., Hardy, S., Blair, A., and Carter, K. (2016).
“Distributed fibre optic monitoring of tension piles under a basement
excavation at the V&A museum in London.” Int. Conf. of Smart Infra-
structure and Construction, ICE Publishing, Cambridge, U.K.
Poulos, H. G. (1989). “Pile behaviour—Theory and application.” Geotech-
nique, 39(3), 365–415.
Poulos, H. G., and Davis, E. H. (1974). Elastic solutions for soil and rock
mechanics, Wiley, New York.
Randolph, M. F. (2003). “Science and empiricism in pile foundation
design.” Geotechnique, 53(10), 847–875.
Salgado, R. (2008). The engineering of foundations, McGraw-Hill,
New York.
Savitzky, A., and Golay, M. J. E. (1964). “Smoothing and differentiation
of data by simplified least squares procedures.” Anal. Chem., 36(8),
1627–1639.
Schwamb, T., et al. (2014). “Fibre optic monitoring of a deep circular
excavation.” Proc. ICE Geotech. Eng., 167(2), 144–154.
Schwamb, T., and Soga, K. (2015). “Numerical modelling of a deep
circular excavation at Abbey Mills in London.” Geotechnique, 65(7),
604–619.
Seo, H.-J., Pelecanos, L., Kwon, Y.-S., and Lee, I. M. (2017). “Net load–
displacement estimation in soil-nail pullout tests.” Proc. Inst. Civ. Eng.
Geotech. Eng., 170(6), 1–14.
Soga, K. (2014). “Understanding the real performance of geotechnical
structures using an innovative fibre optic distributed strain measurement
technology.” Rivista Italiana di Geotechnica, 4, 7–48.
Soga, K., et al. (2015). “The role of distributed sensing in understanding the
engineering performance of geotechnical structures.” Proc., XVI
European Conf. on Soil Mechanics and Geotechnical Engineering,
Thomas Telford, London.
Soga, K., et al. (2017). “Distributed fibre optic strain sensing for monitoring
underground structures—Tunnels case studies.” Underground sensing,
S. Pamukcu and L. Cheng, eds., 1st Ed., Elsevier, Amsterdam,
Netherlands.
Stas, C. V., and Kulhawy, F. H. (1984). “Critical evaluation of design meth-
ods for foundation under axial uplift & compression loading.” EL-3771,
Electric Power Research Institute, Palo Alto, CA.
Tomlinson, M., and Woodward, J. (2014). Pile design and construction
practice, CRC Press, New York.
Tomlinson, M. J. (1997). “The adhesion of piles driven in clay.” Proc., 4th
Int. Conf. on Soil Mechanics and Foundation Engineering, London.
J. Geotech. Geoenviron. Eng.