Proceedings of the Institution of

Civil Engineers
Geotechnical Engineering 163
June 2010 Issue GE3
Pages 119–130
doi: 10.1680/geng.2010.163.3.119

Paper 900053
Received 22/06/2009 Loretta Cheung Kenichi Soga Peter J. Bennett Yusuke Kobayashi Binod Amatya Peter Wright
Accepted 17/02/2010 Geotechnical Engineer, Professor of Civil Engineering, Formerly Senior Research Senior Researcher Japan Formerly Research Associate, Senior Tunnel
Keywords: electrical engineering & Geotechnical Engineering Department of Engineering, Associate, Department of Railway Technical Research Department of Engineering, Engineer, Tube
distribution/field testing & Office, Hong Kong; formerly University of Cambridge, UK Engineering, University of Institute, Tokyo, Japan; University of Cambridge, UK Lines Ltd,
monitoring/tunnels & tunnelling Graduate Student, Cambridge, UK formerly Visiting Researcher, London, UK
Department of Engineering, Department of Engineering,
University of Cambridge, UK University of Cambridge, UK

Optical fibre strain measurement for tunnel lining monitoring

L. L. K. Cheung MSc, MPhil, CEng, MHKIE, K. Soga PhD, FICE, P. J. Bennett PhD, CPhys, Y. Kobayashi PhD,
B. Amatya PhD and P. Wright MSc, CEng, MICE

A trial strain-monitoring system using Brillouin optical communications for carrying information, and is one of the
time-domain reflectometry (BOTDR) technology was set primary candidates for resident sensing systems. In this study,
up to monitor joint movements in the concrete tunnel a unique distributed strain measurement technology called
lining in an existing London Underground tunnel. The Brillouin optical time-domain reflectometry (BOTDR) is used.
BOTDR strain sensor system allows the measurement of Brillouin scattering is generated by interaction of the incident
strain distribution along an optical fibre using the light travelling along a standard optical fibre with the acoustic
reflective technique, requiring access to only one end of phonons, and the frequency of the backscattered light changes
the fibre. Measurements were obtained by a strain- in proportion to the strain and temperature along the fibre.
sensing optical fibre installed along the tunnel lining. The Monitoring the frequency shift of the backscattered light
joint movements were captured by measuring the strain coming from various locations of an optical fibre allows
along the fibre across the segment joints. The results continuous measurement of the strain along the fibre. The
show that there is good agreement between the joint novel aspect of this new technology lies in the fact that tens of
movements evaluated by the BOTDR strain sensor kilometres of fibre can be sensed at once for continuous
system and those by conventional vibrating-wire strain distributed strain measurement, providing a relatively cheap
gauges. Whereas conventional strain measurement but highly effective monitoring system. Further details of this
gauges monitor the strain variations at discrete technology can be found in Horiguchi et al. (1989), Oka et al.
locations, a BOTDR strain sensor can provide a (1999), Ohno et al. (2001a, 2001b) and Klar et al. (2006).
continuous strain distribution of the tunnel lining. The
results demonstrate the practicality of using the BOTDR The number of applications of this technology for monitoring
strain-sensing system to monitor the movement of geotechnical infrastructure has grown in recent years. They
tunnel linings. include Fujihashi et al. (2005) and Mohamad et al. (2007a,
2010) on tunnel monitoring during nearby underground
1. INTRODUCTION construction, Naruse et al. (2000) and Kihara et al. (2002) on
Deterioration of ageing civil engineering infrastructure and the monitoring the movement of a large-scale river levee model by
associated increase in the proportion of budgets spent on increased seepage force, Kojima et al. (2003) and Janmonta et
maintenance present significant challenges to our society. As al. (2008) on slope movement monitoring of transportation
the mechanisms of ageing can be complex, dense spatial and embankments, Ohno et al. (2002) and Klar et al. (2006) on
temporal information, integrated with appropriate data piled foundation performance monitoring, Naruse et al. (2007)
analytical tools, is required to assess and reduce the likelihood on ground movement monitoring associated with mining
of, or improve the efficient response to, failures of key activities, Mohamad et al. (2007b), Mohamad (2008) on lateral
elements of critical infrastructure resulting from degradation movement monitoring of secant piled walls during deep
and overload. The need to maintain and refurbish ageing excavation, and Vorster et al. (2006) on monitoring of buried
infrastructure and operate it safely under severe financial pipeline behaviour during adjacent tunnelling. The above
constraints forces civil engineers to strive for technological monitoring works and the associated experiments indicated
advances that will allow them to sense, monitor and better that the BOTDR technique is capable of detecting the
understand the behaviour of their engineering systems under movement of geotechnical structures by capturing the strain
both normal and extreme operating conditions. At present, variation.
network-wide monitoring is prohibitively expensive, and it is
difficult to obtain the necessary data for quick assessment. This paper describes the results of a field trial conducted to
examine the feasibility of BOTDR for monitoring the joint
Structural integration of optical fibre sensing systems movements of concrete tunnel linings in an existing London
represents a new branch of engineering, which involves the Underground tunnel. Recent tunnel inspection records at the
unique marriage of optical fibres, optoelectronics and location of interest revealed that the concrete segments
composite material science. The transmission of light down an forming the running tunnel lining were spalling next to the
optical fibre is an established technique in optical joints. Because of the large number of segments to be

Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al. 119
 monitored along the tunnel, the distributed nature of the
BOTDR system makes it an attractive option. In this study, Drift deposits
optical fibre was installed along the tunnel lining, and the Approximately 25 m
movements were captured by measuring the strain along the
London Clay
fibre attached across the segment joints. At selected locations Areas of spalling found in the tunnel
Optical fibre installation
joint movement was also measured by vibrating-wire strain
Jubilee Line southbound tunnel Upper mottled
gauges (VWSGs), and the two sets of data were compared to clay
evaluate the accuracy of the BOTDR monitoring system in Upnor Formation
relation to that of the conventional strain-monitoring device.
Chalk

2. SITE DESCRIPTION Baker Street station ⬃1000 m Bond Street station ⬃600 m
The site is located at the southbound tunnel section between
Baker Street station and Bond Street station of the Jubilee Line, Figure 1. Monitoring location and site geology (not to scale)
London Underground. The tunnel was constructed from 1973
to 1979 as part of the Stage 1 Fleet Line construction, as a
solution to relieve the overcrowded Bakerloo Line at that time. 3850 mm internal diameter or concrete segments of 3810 mm
The Fleet Line was subsequently renamed the Jubilee Line in internal diameter. About 60% of the southbound tunnel and
1977, Queen Elizabeth II’s Silver Jubilee year. 80% of the northbound tunnel are lined with concrete
segments, and they were used when ‘the ground condition was
The geology consists of 5–11 m of drift deposits (mainly sand good’ (Jobling, 1980). The concrete lining ring, which is
and gravel) underlain by 6–25 m of London Clay. According to instrumented in this study, consists of 20 segments plus two
the as-built records, the tunnel between the two stations is at a wedge blocks at knee level, as shown in Figure 2. The
depth of between 23 m and 37 m below the ground surface, as manufacturing tolerances of the concrete segments were of the
shown in Figure 1. It was driven through London Clay for the order of
1 mm on the main dimension, and
5 mm on the
first 500 m section from the northern end (i.e. Baker Street convex surface on the radial joint (Jobling, 1980). The 168 mm
station), and then through the Lambeth Group strata for the thick concrete segments were held in position by pushing two
remaining section towards Bond Street station to the south. wedge-key segments at knee level, and locking them in
About two-thirds of the tunnel is in the top portion of the compression against the ground and each other.
Lambeth Group stratum, which exists below London Clay and
consists of interlayered sand, silt and clay. The tunnel section For the construction of the running tunnel, conventional
at the trial site lies at the top of the upper mottled clay and red Greathead tunnelling shields were adopted (Lyons, 1979). Most
mottled layer of the Lambeth Group. Beneath the Lambeth of the excavation was carried out manually, and by mechanical
Group stratum is the highly permeable Thanet Sand and chalk. excavators in some cases. When building expanded concrete
linings, the Greathead shield operated without a tailskin, and
The running tunnel is lined with either cast iron segments of the concrete segments were erected directly against the smooth

F G
H 19 000 rad.
E J

D
K
115
C L
Inner diameter: 3810 1 in 10
Outer diameter: 4146
100 220
B M
Concrete segments
1 in 10
A N 520
Wedge
segment
Wedge
segment
19 000 rad.

600

Figure 2. Details of the concrete segments (dimensions in mm)

120 Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al.
clay trimmed by the shield. However, block failures above ovalisation refers to squatting deformation, that is, squeezing
shoulder level were experienced where the segments were of tunnel section along a horizontal axis. Ovalisation of both
being erected. The miners were subsequently protected from the southbound and northbound running tunnels maintained
injury by the installation of trailing steel headboards behind an almost constant level of 0.5% for the first 600 m section
the shield. The steel headboards attached to the rear of the from Baker Street station. The ovalisation at the southbound
shield were initially trapped by the expanded ring, and hence tunnel becomes larger and more variable thereafter, between
the headboards were given a slight overcut to relieve this rings 1423 and 1781, which are in the Lambeth Group. The
construction problem. The voids formed by the headboards distortion reaches a maximum of 2% near southbound ring
behind the crown segments were apparently grouted 1620.
subsequently, even though there is little evidence to confirm
that this grouting was actually carried out. Because of the construction difficulties described above, it is
likely that the ring sections were not constructed within
There was approximately 5 mm of surface settlement observed tolerance, resulting in ovalisation during the placement of the
during the construction of the running tunnels, 7 mm at the linings. The large distortion at the concrete lining section may
cross-passages, and 13 mm near the working shafts. In 2006 also be due to faulting in the area and the consequently mixed
Tube Lines carried out circularity measurements of the tunnels ground conditions that were encountered. The large
between Baker Street and Bond Street stations using a trolley- manufacturing tolerances (
5 mm) of the concrete segments at
mounted laser device (Wright, 2010). Measurements were taken the radial joint as reported in Jobling (1980) are another
at one in every ten tunnel rings, and where a change in tunnel probable reason for the poorly built tolerance of the tunnel
lining type or diameter was encountered. Best-fit centre and ring.
radius were evaluated for the surveyed rings. The radial
distances from the measured points to the best-fit circle were 3. INSTALLATION OF BOTDR MONITORING
calculated, and the difference was output as a percentage of the SYSTEM
tunnel radius. The maximum deviation of the surveyed point to A trial BOTDR strain monitoring system was installed from ring
the best-fit circle was defined as ovalisation. Figure 3 shows a 1652 to ring 1680 in the southbound tunnel, as shown in
plot of the ovalisation against ring number. Positive Figures 3 and 4. The location where concrete spalling was

Monitoring location
Southbound ring number
2580
2540
2500
2460
2420
2380
2340
2300
2260
2220
2180
2140
2100
2060
2020
1980
1940
1900
1860
1820
1780
1740
1700
1660
1620
1580
1540
1500
1460
1420
1380
1340
1300
2·0 2·0
Northbound
Ovalisation: % (squat ⫹ve)

Southbound
1·0 N/B moving average 1·0
S/B moving average

0 0

⫺1·0 ⫺1·0

⫺2·0 ⫺2·0
0
40
80
120
160
200
240
280
320
360
400
440
480
520
560
600
640
680
720
760
800
840
880
920
960
1000
1040
1080
1120
1160
1200
1240
1280

Northbound ring number

Southbound ring number

1280
1240
1200
1160
1120
1080
1040
1000
960
920
880
840
800
760
720
680
640
600
560
520
480
440
400
360
320
280
240
200
160
120
80
40
0

2·0 2·0
Ovalisation: % (squat ⫹ve)

1·0 1·0

0 0

⫺1·0 ⫺1·0

⫺2·0 ⫺2·0
1300
1340
1380
1420
1460
1500
1540
1580
1620
1660
1700
1740
1780
1820
1860
1900
1940
1980
2020
2060
2100
2140
2180
2220
2260
2300
2340
2380
2420
2460
2500
2540
2580

Northbound ring number

Figure 3. Ovalisation ratio against ring number and the monitoring location

Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al. 121
 Tunnel shaft at Glass core
Park Square Gardens Ring 1680
Ring 1653
Glass cladding

Plastic coating, 900 µm

Northbound (a)
Southbound tunnel
tunnel

Figure 4. Schematic diagram of monitoring location 8 optical fibres, 250 µm coating

Gel-filled tube
Central strengthening member

extensive is about 600 m from Bond Street station. The optical Dummy tube
fibre system was installed over 12 working days in February Kevlar yarns
2008 during the engineering hours of the London Underground
Cable jacket
(i.e. between 1 am and 4 am). To monitor the fibre, an N8510
BOTDR analyser by Advantest Corporation was placed at a (b)
tunnel ventilation shaft located about 500 m away from the
monitoring location at Park Square Gardens (see Figure 4). The Figure 6. Cross-sections of the optical fibres used for the trial:
analyser was connected to the monitoring location via the (a) 900
m fibre; (b) connecting cable
connecting cable.

Figure 5 shows the layout of the strain-monitoring system.

Two types of optical fibre cable were used in the system: (a) The optical fibre was fixed to the lining by a disc system, as
900
m fibre for strain sensing; and (b) general-purpose loose- shown in Figure 7. The attachment of the fibre to the 50 mm
tube cable for connecting the strain sensing fibre to the diameter disc is intended to ensure that there are no sharp
analyser. Figure 6 shows the cross-sections of the cables. The bends along the fibre, and therefore minimal loss of signal at
900
m fibre contains one single-mode optical fibre. The fibre the bends. Figure 7 shows the locations where the discs were
is protected by a plastic coating, which adheres to the fibre so fixed on the concrete segment. The discs were attached to the
that strain transfer is possible. The general-purpose loose-tube concrete segments by drilling holes into the concrete and
connecting cable contains two fibre bunches, each of them fixing them with screws and plugs. A total of 57 discs were
containing eight optical fibres. The fibre bunches are contained installed in about two and a half working nights. The optical
in gel-filled tubes. The tubes are surrounded by a thick jacket fibre cable was then mounted under tension and glued to the
coating with a central strengthening member and three dummy fixed discs. The installation enables the detection of joint
tubes, making the cable more robust and fit for connecting opening at the inner surface of the lining by an increase in
purposes between the sensing cable and the analyser, which tensile strain along the cable, while joint closing is indicated
was located at the ventilation shaft. by a decrease in strain, as shown in Figure 8.

The ring consists of 20 segments of unbolted precast concrete

segments, as shown in Figure 2. In most cases the 900
m
cable was fixed under tension at the tunnel crown on concrete 618 mm
segments F and G of 27 rings (ring 1653 to ring 1680), as
shown in Figures 2 and 5. At three rings (1675, 1676 and 1677)
the fibre was also installed along some other segments (D to J)
(see Figure 5).
Holes on 602 mm
concrete panel
50 mm
diameter disc
Ring numbers
1672

1662

1652
1680

1674

1670

1664

1660

1654
1678
1676

1668
1666

1658
1656

N
M 50 mm
L
K VWSGs diameter disc Screw
Segments

J
H
G
F
E
D
C
B Junction box
A

Figure 5. Optical fibre installation layout and vibrating-wire

strain gauge (VWSG) locations Figure 7. Optical fibre fixing using the disc system

122 Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al.
 Concrete segment

Cable

50 mm diameter disc
(a)

(c)

(b)

Figure 8. Potential movements of segments in the transverse direction: (a) zero movement; (b) tensile movement (positive strain);
(c) compressive movement (negative strain)

For the subsequent discussion, the layout of the strain

monitoring system is divided into six sections, as shown in
Figure 9 90° section
(rings 1656 to 1665,
cable length approx.
(a) the 908 section 15 m)
(b) the diagonal section
(c) the extended section Diagonal section
(rings 1669 to 1673,
(d ) the first tension section cable length approx.
(e) the second tension section 2 m)
(f) the slack section.
Extended section
(rings 1675 to 1677,
The 908 section and the extended section capture the cable length approx.
8 m)
movement of each single joint, whereas the diagonal section is
designed to investigate the difference in signal loss by reducing First tension section
the number of bends (i.e. the number of discs) along the cable. (rings 1680 to 1663,
cable length approx.
However, in contrast to the 908 and extended sections, the 16 m)
strain variation in the diagonal section is affected by the
movements of four adjacent joints. The first and second tension Second tension
section
sections and the slack section are designed for the purpose of (rings 1663 to 1680,
temperature compensation. cable length approx.
16 m)

Slack section
4. DATA ANALYSIS (rings 1680 to 1663,
cable length approx.
16 m)
4.1. Power loss
Upon completion of system installation, the strain monitoring
data were taken about once every week from 21 February 2008 Figure 9. Location of the six monitoring sections
to 4 June 2008 (except between 17 March to 14 April, when

Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al. 123
 the cable was damaged and subsequently repaired, and hence

Ring 1656

Ring 1665

Ring 1669
Ring 1673
Ring 1675

Ring 1677
⫺15·5
no readings were possible). Nine datasets were collected within
⫺15·7
this monitoring period. For each dataset five or six readings
⫺15·9
were taken, which took about 20–30 min each. Although the
⫺16·1
analyser can take readings more frequently (e.g. daily) by

Power: dB
⫺16·3
programming, this was not possible in this trial, because the ⫺16·5
power supply at the ventilation shaft was not available ⫺16·7
continuously during the monitoring period. ⫺16·9
Diagonal Extended
⫺17·1 Connecting cable 90° section section section
Figure 10 shows the initial BOTDR strain data obtained on 21 ⫺17·3
February 2008. As shown, the connecting cable from the ⫺17·5
495 500 505 510 515 520 525 530 535 540 545 550
analyser at the tunnel shaft to the monitoring location is about Distance: m
500 m long. Distinct pre-tensioned strain sections are observed,
referring to the different strained cable sections as defined in Figure 11. Power loss of the strain-monitoring system with
Figure 9. Although it was intended to have a constant pre- distance from the analyser
tension along each section, some sections, such as the 908 and
diagonal sections, had variable strain profiles because of
friction between the cable and the discs. However, this was not 4.2. Digital filtering and temperature compensation
an issue, as the change in BOTDR strain from the base reading The BOTDR data are affected by temperature. The connecting
was the main point of interest. cable section and the slack section will have no additional
mechanical strain from the initial BOTDR strain data. Hence the
The power loss of the strain-monitoring system is shown in strain differences at these two sections reflect the temperature
Figure 11, and is summarised in Table 1. The power loss is variations, and these sections can be used for temperature
about 0.014 dB per bend at the 908 bend section, whereas it is compensation. Furthermore, the reading fluctuates within a
about 0.013 dB per bend at the diagonal section. Therefore strain level of about 50 microstrain. In order to obtain the
there is insignificant difference in power loss between the two actual mechanical strain change, data filtering was conducted
types of bend. In the diagonal section the movement of each on the raw BOTDR data before temperature compensation.
joint cannot be distinguished, as the strain variation of the
cable is affected by four adjacent joints. Therefore it was As the BOTDR analyser measures a weighted average of strain
concluded that the 908 section provides more precise data on over 1 m, it is considered that a rolling filter would be suitable
joint movement than the diagonal section. for the noise reduction. In this study a Savitzky–Golay filter
was used. The method minimises the least-squares error in
fitting a polynomial to each frame of the data. By adopting a
rolling-frame approach the heights and widths of the curves
can be well preserved. There is no guideline for choosing the
0·8
window size and parameter for the filter, and so they are
0·7 chosen on a trial-and-error basis until the specifications are
0·6 satisfied (Shenoi, 2006). With the aim of reducing noise, a
second-order polynomial fitting and a frame size of 9 was
0·5
Strain: %

chosen, as this setting shrinks the noise range significantly

0·4 without distorting the trend of the data. As the data are
Connecting cable (to 0 m)
0·3 recorded at 50 mm intervals, a frame size of 9 refers to
90° section performing the filtering with reference to the data points
0·2 Diagonal section and extended section
within
200 mm. As the span length of the trial monitoring
First tension section
0·1 Second tension section system is 600 mm, the 400 mm (200 mm + 200 mm) frame size
0
Slack section lies well within every single span, and therefore was considered
400 450 500 550 600 to be suitable.
Distance: m

To ensure that appropriate sections are selected for temperature

Figure 10. Initial BOTDR strain profile along the installed
optical fibre compensation, the change in the strain level from the initial
data (i.e. data taken on 21 February 2008) is examined.

Cable section Power loss: dB Type and No. of joint Loss per bend: dB

908 section 0.28 908 bend 3 20 ,0.014

Diagonal section 0.04 908 bend 3 3 ,0.013
Extended section 0.08 908 bend 3 6 ,0.013

Table 1. Power loss of the cable and disc system

124 Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al.
Without the influence of mechanical strain (i.e. due to the temperature rise from February 2008 to June 2008 was
structural movement), the strain difference due solely to approximately 4–5 Celsius degrees (unfortunately, no
temperature variation should be consistent over the cable independent measurement was made to confirm this). The
section. By subtracting the initial data taken on 21 February values of temperature change evaluated for the two strained
2008, the strain differences with the initial data for each sections (the first and second tension sections) are smaller than
dataset are obtained, and are shown in Figure 12 for the slack the value for the unstrained section (the slack cable section).
section where no mechanical strain is applied. Within this The difference is due to the mechanical strain captured by the
section, the fluctuation in the data is due to resolution of the two sections of cable to which tension was applied. The figure
BOTDR readings (approximately 50 microstrain), which seems also shows the temperature change evaluated using the
to be slightly location dependent. To find the location that had connecting cable that was placed from the monitoring location
the smallest fluctuation, the standard deviation of the strain to the ventilation shaft. The data show similar trends to those
difference was calculated for every 2 m length, and a minimum of the other sections, which suggests that the temperature is
was found between 598 m and 600 m. This section was relatively constant throughout the tunnel.
therefore considered to be suitable for temperature
compensation. Similar exercises were conducted for the first 4.3. Gauge length and resolution
and second tension sections. At the 908 section, the BOTDR system measures the relative
movement between the two discs, which has a 600 mm span
By adjusting the strain value with respect to the temperature length. However, the strain reading given by BOTDR is a
coefficient of the optical fibre cable, the temperature difference weighted average of the strain over 1 m, which comes from the
from the initial reading was obtained. Laboratory experiments length of the light pulse generated by the analyser (Advantest,
gave the temperature coefficients of the connecting cable and 2006). This gauge length issue is an inherent problem for most
the 900
m cable to be 23.6 and 23.8 microstrain/8C of the BOTDR analysers currently available on the market. To
respectively (Mohamad, 2008). Using these values, the overcome the influence of this gauge length problem, there is a
temperature changes from the initial reading in February were need to evaluate the true strain between two discs based on the
calculated for three sections (first tension section, second 1 m resolution BOTDR data, as shown in Figure 14.
tension section and slack cable section), and the results are
shown in Figure 13. The three sections show a similar trend: Klar et al. (2006) report that the strain distribution profile can
be expressed by the integration of local true strain multiplied
by the Gaussian distribution, as shown by the equations
0·025 10 March 2008 21 April 2008 30 April 2008
7 May 2008 15 May 2008 22 May 2008
28 May 2008 21 April 2008 ð1
0·020
1 åA ð x Þ ¼ åT ðôÞ gð x  ôÞdô
1
BOTDR strain: %

0·015

0·010

0·005  
1 ÷2
2 gð÷Þ ¼ pffiffiffiffiffiffi exp  2
0 w 2ð 2w
⫺0·005

⫺0·010
587 589 591 593 595 597 599 601 603 where åA (x) is the averaged strain obtained at point x along the
Distance: m fibre; åT (x) is the true strain; ô is the variable of integration;
g(÷) is the weighing function, which takes the form of a
Figure 12. Temperature effect on BOTDR strain readings Gaussian distribution; and w is the width of the Gaussian
distribution. Results of laboratory experiments using the
BOTDR analyser used in this study gave the value of w as
5·5 ranging from 0.3 m to 0.45 m; the magnitude of w decreased
Connecting gel cable
5·0 First tension section with increasing strain (Advantest, personal communication,
4·5
Temperature difference: °C

Second tension section 2009; Cheung, 2008).

4·0 Slack cable section
3·5 Average of four sections
3·0 Using Equations 1 and 2, the BOTDR reading from the stepwise
2·5
2·0
distribution of true strains (åT ) will be (see Figure 14)
1·5
1·0
0·5 X
þ1

0 3 åAi ¼ åT n G i, n
⫺0·5 n¼1
⫺1·0
21 Feb 2008
28 Feb 2008
6 Mar 2008
13 Mar 2008
20 Mar 2008
23 Mar 2008
3 Apr 2008
10 Apr 2008
17 Apr 2008
24 Apr 2008
1 May 2008
8 May 2008
15 May 2008
22 May 2008
29 May 2008
5 Jun 2008

ð x nþ1
4 G i, n ¼ gð x i  ôÞdô
xn
Date

Figure 13. Temperature change during the monitoring period where åTn is the actual strain of the nth span, and x n is the
distance at the beginning of the nth span.

Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al. 125
 0 1
å0
nth span B C
εTn B å1 C
εAi ⫽ .... ⫹ εn⫺1 ⫹ εnGi,n B C
⫹ εn⫹1Gi,n⫹1 ⫹ ... B C
ε *Ai B . C
Strain

∆e True strain B .. C ¼
εAi B C
B C
B C
εTn⫹1 Strain reading B ån C
n ⫹ 1th B C
span
@ A
Gi,n Gaussian distribution ..
.
xn xi xn⫹1 xn⫹2 Distance 0X 11
X X
G i,0 G i,0 G i,1 G i,0    G i, m G i,0   
B C
Figure 14. BOTDR strain, true strain and Gaussian B i i i C
B C
distributions BX X X C
B G i,0 G i,1 G i,1 G i,1    G i, m G i,0    C
B C
B i C
B i i C
B C
To evaluate the true strain, deconvolution of the BOTDR data B . . . C
B . . . C
B . . . C
was done by the least-squares method (LSM) approach, which 7 B C
BX X X C
minimises the error ˜e between the strain åAi measured by the B
B G i,0 G i, n G i,1 G i, n G i, m G i, n
C
C
analyser and the calculated strain åA i . A partial differential of B i i i C
B C
the sum of error squares is calculated by the equation @ A
.. .. ..
. . .
0X 1
X 2 åAiG
 å 
i,0
B C
@ åAi Ai B i C
B C
i BX C
B åAi G i,1 C

@ån B C
" # B i C
B C
@ X  2 X

X B C
5 ¼ åAi  2 åAi åAi þ ðåAi Þ 2
3BB .
.
C
C
@ån i i i B . C
B C
! B X C
X X X B 
åAi G i, n C
B C
¼ 2 åAi Gi,n þ 2 åm Gi,m Gi,n B i C
B C
i i m @ A
..
¼0 .

A computer program was developed to estimate the true

Hence
mechanical strain values by the above least-squares method.
Figure 15 shows the data from the analyser, the estimated true
0X X X 1 strain profile, and the back-calculated estimated strain profile,
G i,0 G i,0 G i,1 G i,0    G i, m G i,0    to verify the least-squares approach.
B C
B i i i C
B C
BX X X C
B G i,0 G i,1 G i,1 G i,1    G i, m G i,1    C
The distance of the measuring points along the optical fibre
B C
B i C from the analyser should be checked to ensure that the strain
B i i C
B C value at the same location is compared from different sets of
B .. .. .. C
B C data. Hence the accuracy of the readings is also affected by the
B . . . C
B C
BX X X C distance measurement accuracy of the analysers. According to
B G G G G G G C
B i,0 i, n i,1 i, n i, m i, n C Advantest (2006), the distance accuracy is estimated by the
B i i i C
B C expression
@ A
.. .. ..
6 . . .
0X 1
åAiG
[5 3 105 3 measurement distance (m)
i,0
0 1 B B i
C
C 8
å0 B C þ 0:2 m þ 2 3 reading resolution (m)]
B C B B X C
B å1 C B
B C B åAi G i,1 C

C
B C B i C
B . C B C
B .. C ¼ B C
3B C B .. C
B C B C The measurement distance of the trial monitoring system was
B C B . C
B ån C B X C about 550 m, and the reading resolution was 0.05 m. Thus the
B C B
@ A B G C
åAi C distance measurement accuracy of the trial system is
i, n C
.. B i C
. B C
327.5 mm. As the span length between two discs is about
@ A
.. 600 mm, the distance measurement accuracy of 327.5 mm can
. be significant. It is therefore important to check the distance at
various locations of the fibre. The distances can also be
checked by making reference to different sections of cable with
The following series of true strains are obtained. different strain profiles.

126 Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al.
 Ring 1656

Ring 1665

Ring 1669

Ring 1673

Ring 1675

Ring 1677

Ring 1680
0·70

0·65

0·60

0·55

Strain: % 0·50

0·45

0·40

0·35

0·30 Data from the analyser

Estimated true strain
0·25
Back-calculated strain profile by
least-squares method for verification
0·20
511 513 515 517 519 521 523 525 527 529 531 533 535 537 539 541 543 545 547 549 551
Distance: m

Figure 15. Deconvolution of BOTDR strain distribution

5. RESULTS At rings 1660 and 1662 optical fibre was also attached between
segments F and G. The strains evaluated from the BOTDR data
5.1. Comparison with vibrating-wire strain gauge data are converted to displacements using the distance between two
In 2007 Tube Lines installed vibrating-wire strain gauges discs. Figure 16 shows a comparison of the displacements
(VWSGs) at various rings within the concrete tunnel section. obtained from the BOTDR data and the VWSG data. For the
The strain gauges were fixed across adjacent concrete BOTDR data, the average temperature of the four cable sections
segments, and the relative joint movements have been was used for temperature compensation (see Figure 13). The
monitored since the installation. In this study, joint opening at BOTDR distributed strain data were then deconvoluted using
the inner surface of the tunnel lining is indicated by tensile four Gaussian width (w) values between 0.3 and 0.45. The
strain. Assuming that the measured strain originates solely evaluated strains were converted to the relative displacements
from the movement of the joint, the displacement across the between the two fixed points. The VWSG data were also
inner surfaces of the two segments can be evaluated by converted to the relative displacement. Both measurement
multiplying the measured strain by the gauge length techniques show that the joints opened by approximately 25–
(139.7 mm). 50
m in 16 weeks, and the general trends of movement are

100 1660 vibrating-wire strain gauge 100 1660 vibrating-wire strain gauge
w ⫽ 0·450 w ⫽ 0·450
90 w ⫽ 0·375 90
w ⫽ 0·375
80 w ⫽ 0·350 80 w ⫽ 0·350
70 w ⫽ 0·30 70 w ⫽ 0·30
60 60
Movement: µm

50
Movement: µm

50
40 40
30 30
20 20
10
10
0
0
⫺10
⫺10
⫺20
⫺20
⫺30
⫺30
⫺40
⫺40
21 Feb 2008

7 Mar 2008

22 Mar 2008

6 Apr 2008

21 Apr 2008

6 May 2008

21 May 2008

5 June 2008

21 Feb 2008

7 Mar 2008

22 Mar 2008

6 Apr 2008

21 Apr 2008

6 May 2008

21 May 2008

5 June 2008

Date
(a) Date
(b)

Figure 16. Comparison of joint movements estimated from BOTDR strain data and from vibrating-wire strain gauges: (a) ring 1660;
(b) ring 1662. w is the Gaussian distribution width; joint opening is taken as positive

Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al. 127
 similar. The measured movements are small, and hence the lining behaviour, further monitoring works, which include the
difference may be due to the limited accuracy of the BOTDR monitoring of groundwater and thermal response of the
measurements. Although the deconvoluted data using w ¼ 0.3 concrete segments for example, is needed.
appear to match the VWSG data, further investigation is
needed to improve the method of deconvolution. 5.3. Longitudinal movement
Figure 19 shows the movements of the joints normal to the
5.2. Transverse movement longitudinal direction. An opening movement is recorded at
Figure 17 shows the movements of the joints perpendicular to the joints between rings 1663 and 1664, and between 1664 and
the tunnel transverse direction at different times. As the trial 1665, whereas a closing movement is recorded at the joint
strain-monitoring system is installed over concrete segments F between rings 1676 and 1677. The joint movements vary
and G for most of the rings (except rings 1675 to 1677), the between +65 and 30
m.
result shows that most tunnel sections are subjected to joint
opening between the inner surfaces of the two segments. For As the optical fibre is installed across two adjacent concrete
example, at ring 1664, joint opening is increasing gradually to segments, the relative movement between two segments is
about 76
m over the 16-week monitoring period. measured. The captured strain variations may be due to the
relative vertical or angular movement between the two
At rings 1675 to 1677 the optical fibre installation was concrete segments. However, if it is assumed that the 600 mm
extended over the adjacent concrete segments, apart from wide concrete segments are aligned horizontally (Figure 20a), a
segments F and G. At ring 1675 (D to G) very limited 10
m tensile movement captured by the optical fibre, for
movement (less than 20
m) is recorded over three joints. At example, will be interpreted as a ,3.5 mm vertical movement
ring 1676 (D to J), tensile joint movements are noted between F
and J, and the maximum movement of 80
m occurs at joint
G–H. Very small movement is recorded at joints D–E and E–F.
G
At ring 1677 (F to J) there is gradual tensile movement at F
joints F–G and G–H, but very limited movement at joint H–J.
E Ring 1675

By converting the relative joint movements to angular changes,

a joint movement profile is obtained by the relative angular D
changes. Figure 18 shows the schematic movement profiles of
rings 1675 to 1677. The movements are relative to segment G. G
F H
The angular changes shown as dotted lines in the figures are
enlarged 100 times for easy reference, as the profile using the E Ring 1676 I
calculated angular changes is too small to be presented. Since a
fixed reference point is not available, the actual movement
D
profile cannot be determined. However, the data show that the
tunnel is gradually squeezing along the vertical axis for rings
1676 and 1677, which was also shown by the VWSG data. On G
F H
the other hand, ring 1675 shows the opposite trend. Based on
the data recorded by other VWSGs along the tunnel section as Ring 1677
I
well as the BOTDR data, it is considered that complex
behaviour of the tunnel lining is observed, owing to the large Figure 18. Relative movements of rings 1675, 1676 and 1677
number of segments per ring. To facilitate the study of tunnel

Movement: µm (joint opening positive)

⫺40–20 ⫺20–0 0–20 20–40 40–60 60–80 80–100 Date

4 Jun 2008
28 May 2008
22 May 2008
15 May 2008
7 May 2008
30 Apr 2008
21 Apr 2008
4 Mar 2008
21 Feb 2008
1661

1671

1673 (G to F)

1673 (F to G)

1675 (G to F)
1675 (F to G)
1675 (E to D)
1676 (D to E)
1676 (E to F)
1676 (F to G)
1676 (G to H)
1676 (H to J)
1677 (J to H)
1677 (H to G)
1677 (G to F)
1662

1672
1657

1660

1664

1670
1656

1658
1659

1663

1669
1665

Ring number and joint location

Figure 17. Changes in joint opening with time for joints perpendicular to tunnel transverse direction

128 Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al.
 Movement: µm (joint opening positive)

⫺40–20 ⫺20–0 0–20 20–40 40–60 60–80 Date

4 Jun 2008

28 May 2008

22 May 2008

15 May 2008

7 May 2008

30 Apr 2008

21 Apr 2008

4 Mar 2008

21 Feb 2008
1660–1661

1670–1671
Nil

Nil

Nil

Nil

Nil

Nil
1661–1662

1671–1672
1656–1657

1659–1660

1663–1664

1669–1670

1673–1674

1676–1677
1657–1658

1658–1659

1662–1663

1672–1673

1675–1676
1664–1665

1674–1675
Ring number and joint location

Figure 19. Changes in joint opening with time for joints parallel to tunnel transverse direction

between the two segments, which is unrealistically large. It is movements were captured by measuring the strain along the
likely that the longitudinal segments are not aligned fibre attached across the segment joints.
horizontally (as shown in the schematic diagram in Figure
20b), because of the poorly built tolerance at the construction As the BOTDR data are affected by temperature, the
stage. It is possible that the strain variations along the optical temperature influence of the strain measurement was
fibre are due to angular joint movements caused by increased compensated for by using appropriate cable sections that do
earth pressure resulting from a change in the subsurface not develop any mechanical strain. The strain reading given by
condition (e.g. changes in the pore water pressure of the BOTDR is a weighted average of the strain over 1 m. Hence the
London Clay stratum, or changes in subsurface loading due to strain data over the 600–800 mm span length between two
nearby construction activities) acting on the protruding adjacent concrete segments were deconvoluted by the least-
concrete segments, and also from thermal expansion and squares method to evaluate the true strain between two discs. It
contraction of the concrete segments. was demonstrated that the BOTDR system captured a joint
movement trend similar to that of the VWSG data, and the
difference was approximately 20
m within the monitoring
6. CONCLUSION period of 16 weeks. the deconvoluted data and the VWSG data
A trial strain-monitoring system using Brillouin optical time- match at a specified width of Gaussian distribution of the least-
domain reflectometry (BOTDR) technology was set up to squares method. Other methods or approaches may be explored
monitor the joint movements of a tunnel section located at the to improve the deconvolution when strain data over a short
southbound tunnel between Baker Street station and Bond span length are acquired.
Street station on the London Underground Jubilee Line. Optical
fibre was installed along the concrete tunnel lining, and the The BOTDR data indicated complex behaviour of the tunnel
linings, possibly due to the large number of segments per ring
as well as the poorly built tolerance at the construction stage in
the 1970s. The data from the optical fibre placed at the joints
Tensile movement Compressive movement
in the transverse section showed that most tunnel sections are
subjected to joint opening between the inner surfaces of the
two segments, suggesting that the tunnel is gradually
(a) undergoing a squatting behaviour. However, some data
indicated the opposite trend, which requires further
investigation. Since the segments in the longitudinal direction
are not aligned horizontally, owing to the poor construction, it
is possible that the strain variations along the optical fibre
placed at the joints in the longitudinal direction are due to
angular joint movements resulting from increased earth
(b)
pressure acting on the protruding concrete segments, and also
Figure 20. Possible movements of segments in longitudinal thermal expansion and contraction of the concrete segments.
direction: (a) vertical movement; and (b) angular movement Coupled with other monitoring data, such as pore pressures and
thermal change of the concrete segments, the continuous strain

Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al. 129
 distribution provided by the BOTDR can be useful for tunneling using distributed optical fiber strain
investigating tunnel lining behaviour. measurements. Proceedings of the 7th International
Symposium on Field Measurements in Geomechanics
REFERENCES (FMGM2007), ASCE Geotechnical Special Publication No.
Advantest (2006) N8510 Optical Fibre Strain Sensing System: 175, pp. 1–13.
Operation Manual. Advantest Corporation, Tokyo. Mohamad H, Bennett PJ, Soga K, Klar A and Pellow A (2007b)
Cheung LLK (2008) Fibre Optic Strain Measurement for Distributed optical fiber strain sensing in a secant piled wall.
Monitoring Tunnel Lining Movements. MPhil thesis, Proceedings of the 7th International Symposium on Field
University of Cambridge. Measurements in Geomechanics (FMGM2007), ASCE
Fujihashi K, Kurihara K, Hirayama K and Toyoda S (2005) Geotechnical Special Publication No. 175, pp. 1–12.
Monitoring system based on optical fiber sensing technology Mohamad H, Bennett PJ, Soga K, Mair RJ and Bowers K (2010)
for tunnel structures and other infrastructure. In Sensing Behaviour of an old masonry tunnel due to tunnelling
Issues in Civil Structural Health Monitoring (Ansari F (ed.)). induced ground settlement. Géotechnique.
Springer, Dordrecht, pp. 185–195. Naruse H, Uchiyama Y, Kurashima T and Unno S (2000) River
Horiguchi T, Kurashima T and Tateda M (1989) Tensile strain levee change detection using distributed fibre optic strain
dependence of Brillouin frequency shift in silica optical sensor. IEICE Transactions on Electronics E83-C(3): 462–
fiber. IEEE Photonics Technology Letters 1(5): 107–108. 467.
Janmonta K, Uchimura T, Amatya BL, Soga K, Bennett PJ, Naruse H, Uehara H, Deguchi T et al. (2007) Application of a
Lung R and Robertson I (2008) Fibre optics monitoring of distributed fibre optic strain sensing system to monitoring
clay cuttings and embankments along London’s ring changes in the state of an underground mine. Measurement
motorway. GeoCongress 2008, Characterization, Monitoring Science and Technology 18(10): 3202–3210.
and Modeling of Geosystems, ASCE Geotechnical Special Ohno H, Naruse H, Kihara M and Shimada A (2001a) Industrial
Publication No. 179, pp. 509–516. applications of the BOTDR optical fiber strain sensor. Optical
Jobling DG (1980) Development of Lining Systems for London Fiber Technology 7(1): 45–64.
Transport Tunnels. Internal reference, Tube Lines Limited, Ohno H, Naruse H, Yasue N, Miyajima Y, Uchiyama H, Sakairi
London. Y and Zhi XL (2001b) Development of highly stable BOTDR
Kihara M, Hiramatsu K, Shima M and Ikeda S (2002) strain sensor employing microwave heterodyne detection
Distributed optical fiber strain sensor for detecting river and tunable electric oscillator. Advanced Photonic Sensors
embankment collapse. IEICE Transactions on Electronics and Applications II, Proceedings of SPIE 4596: 76–84.
E85-C(4): 952–960. Ohno H, Naruse H, Kurashima T, Nobiki A, Uchiyama Y and
Klar A, Bennett PJ, Soga K et al. (2006) The importance of Kusakabe Y (2002) Application of Brillouin scattering-based
distributed strain measurement for pile foundations. distributed optical fiber strain sensor to actual concrete piles.
Proceedings of the Institution of Civil Engineers, Geotechnical IEICE Transactions on Electronics E85(4): 945–951.
Engineering 159(3): 135–144. Oka K, Ohno H, Kurashima T et al. (1999) Fiber optic
Kojima K, Ito H, Murata O, Akiyama Y, Yamaura T and distributed sensor for structural monitoring. Proceedings of
Tanamura S (2003) A basic characteristic of optical fibre the 2nd International Workshop on Structural Health
sensor used for maintenance of soil structures. Journal of the Monitoring, Stanford, CT, 672–679.
Japan Society of Civil Engineers 750: 51–67 (in Japanese). Shenoi BA (2006) Introduction to Digital Signal Processing and
Lyons AC (1979) The Jubilee Line 2: construction from Baker Filter Design. Wiley, Hoboken, NJ.
Street to Bond Street exclusive and from Admiralty Arch to Vorster TEB, Soga K, Mair RJ, Bennett PJ, Klar K and Choy CK
Aldwych. Proceedings of the Institution of Civil Engineers, (2006) The use of fibre optic sensors to monitor pipeline
Part 1 66(3): 375–394. behaviour. Proceedings of ASCE GeoCongress 2006,
Mohamad H (2008) Distributed Optical Fibre Strain Sensing of Geotechnical Engineering in the Information Age, ASCE,
Geotechnical Structures. PhD thesis, University of CD-ROM.
Cambridge. Wright P (2010) Assessment of London Underground tube
Mohamad H, Bennett PJ, Soga K et al. (2007a) Monitoring tunnels: investigation, monitoring and analysis. Smart
tunnel deformation induced by close-proximity bored Structures and Systems 6(3).

What do you think?

To discuss this paper, please email up to 500 words to the editor at [email protected]. Your contribution will be forwarded to the
author(s) for a reply and, if considered appropriate by the editorial panel, will be published as discussion in a future issue of the
journal.
Proceedings journals rely entirely on contributions sent in by civil engineering professionals, academics and students. Papers should be
2000–5000 words long (briefing papers should be 1000–2000 words long), with adequate illustrations and references. You can
submit your paper online via www.icevirtuallibrary.com/content/journals, where you will also find detailed author guidelines.

130 Geotechnical Engineering 163 Issue GE3 Optical fibre strain measurement for tunnel lining monitoring Cheung et al.