Vous êtes sur la page 1sur 14

Tunnelling and

Underground Space
incorporating Trenchless
Technology Research
Tunnelling and Underground Space Technology 20 (2005) 487–500

Structural health monitoring of underground facilities

– Technological issues and challenges
S. Bhalla b, Y.W. Yang a,*
, J. Zhao a, C.K. Soh a

School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore
Indian Institute of Technology Delhi, Hauz Khas, New Delhi 110016, India

Received 12 August 2004; received in revised form 23 March 2005; accepted 25 March 2005
Available online 31 May 2005


Driven by the scarcity of land, many urban planners are seriously considering underground space to meet residential, commer-
cial, transportation, industrial and municipal needs of their cities. Besides saving land resources, the benefits offered by underground
structures include safety against earthquakes and hurricanes, and freedom from urban noise. However, owing to their unique design
and construction, they call for rigorous structural health monitoring (SHM) programmes during construction and operation, espe-
cially when important structures are located nearby on the ground surface. Their continuous monitoring can serve to mitigate poten-
tial hazards, ensure better performance and facilitate in-depth understanding of the overall structural behaviour. This paper
addresses major technological issues and challenges associated with structural monitoring of underground structures. A detailed
review of the available sensor technologies and methods for comprehensive monitoring is presented, with special emphasis on con-
ditions encountered underground. Practical benefits arising out of such monitoring are also highlighted, with the help of several real-
life case studies involving underground structures.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Structural health monitoring; Underground; Damage; Sensors; Strain

1. Introduction and loading environment, as well as critical responses

to track and evaluate any symptoms of operational inci-
Civil infrastructures constitute a major part of the dents, anomalies and deteriorations/damages, which
critical assets of any nation. Their failure to perform at might affect its smooth operation, serviceability or safety
an optimum level can severely affect the gross domestic reliability (Aktan et al., 2000). Deteriorations/damages
product of the country (Aktan et al., 1998). However, may result from changes in material properties, geomet-
in many parts of the world, continuous ageing of civil rical configuration, boundary conditions, system connec-
infrastructures is creating unforeseen maintenance prob- tivity and loading environment. Hence, comprehensive
lems for civil engineers. This has motivated active SHM calls for close monitoring of each aspect.
research in the development of real-time and automated The majority of SHM related research completed to
structural health monitoring (SHM) systems, which can date has focused exclusively on surface structures only,
facilitate continuous monitoring of critical infrastruc- such as bridges and buildings (e.g. Pandey and Biswas,
tures with minimum human involvement. SHM is 1994; Farrar and Jauregui, 1998; Aktan et al., 2000;
defined as the measurement of a structureÕs operating Moyo, 2002; Brownjohn et al., 2003a,b). Several real-life
bridges have been comprehensively instrumented and
Corresponding author. Tel.: +65 6790 4057; fax: +65 6791 0676. are currently being monitored (e.g. Moyo, 2002).
E-mail address: cywyang@ntu.edu.sg (Y.W. Yang). However, insufficient effort has been made to develop

0886-7798/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
488 S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500

monitoring systems and techniques for underground constructions. This becomes more relevant in circum-
structures, which have recently attracted the attention stances where unfamiliar construction technologies are
of urban developers worldwide for a number of reasons. used or when a known technology is extended beyond
One of the strongest arguments in favour of the con- the normal range of application. This fact is highlighted
struction of underground structures is that they release by a recent accident in Singapore, on April 20, 2004,
pressure on scarce land resources, especially in metro- during the construction of the mass rapid transit
politan cities. The usable surface space saved by under- (MRT) circle line at Nicoll Highway. In this construc-
ground construction can be alternatively utilised for tion, a temporary wall, which supported a 33 m deep
social endeavours or for creation/preservation of eco- excavation adjacent to Nicoll Highway, collapsed
parks. In addition, underground structures are consid- without warning, leading to widespread cave-in across
ered less prone to damage from extreme events such as six lanes of the Highway. The accident site, shown in
earthquakes, storms/hurricanes and as compared to sur- Fig. 2, covered an area of the size of two basketball
face structures, are less susceptible to noise pollution courts. The accident resulted in the loss of four lives
(Goel, 2001). They are also much safer than surface and injuries to many more, besides causing the closure
structures for the storage of inflammable hydrocarbons of a vital transportation link. Fortunately, no motorists
and other hazardous chemicals. Underground space is were travelling along the usually congested highway at
also free from induced electro-magnetic waves, making that time, otherwise the casualties would have been
it ideal for electro-magnetic related research/manufac- much higher.
turing. The trend of going underground is well reflected Although the incident is still under investigation,
by the proposed Underground Science City in Singapore interim report by the Committee of Inquiry (Magnus
(Zhao et al., 1999). Fig. 1 shows the conceptual design of et al., 2005) indirectly pointed finger at the design and
the proposed Science City, which will house prominent construction of jet grout piles that formed a component
scientific research laboratories (Zhao, 2003). This con- of the soil retaining system. The committee pointed out
struction does not affect in any way the current land- that being unfamiliar construction technique, more
scaping of the area nor compromises any emphasis should have been laid on periodic monitoring
environmental issue. The deep noise free environment of jet grout piles. Comparison between predicted and
will be ideal for research laboratories.
However, unlike surface structures, design and con-
struction of underground structures is far more complex
and expensive. They often encounter unprecedented
problems, both during construction and operation.
Quite often, many of the complex geotechnical and envi-
ronmental parameters cannot be accurately considered
at the design stage. Hence, many of them might be con-
servatively or erroneously estimated, resulting in uneco-
nomical or unsafe designs. Unfortunately, after
construction, the design assumptions are often not vali-
dated. Therefore, a comprehensive instrumentation of
underground structures during construction can pave
way for long-term monitoring of external loads, stress
distributions, deflections and occurrence of damages in
a continuous manner, thereby ensuring a high level of
safety. At the same time, it can serve as means for design
validation as well as a database for economizing future Fig. 2. Nicoll Highway after collapse.

Science Park I

Science Park II

~400 m Shaft
40 m


Caverns for the proposed

Underground Science City

Fig. 1. Conceptual design of the proposed Underground Science City in Singapore (after Zhao, 2003).
S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500 489

actual values, and regular monitoring of critical junc- ciples of various sensing systems, which could possibly
tions could have allowed earlier detection of adverse be deployed for monitoring underground structures.
trends. In particular, the report noted, ‘‘There is need Their pros and cons are highlighted with regard to the
to integrate information from various instruments and conditions encountered in underground structures.
to relate the crucial information to what is happening
at the worksite, as well as the quality of each of the ele- 2.1. Strain gauges
ments in construction.’’ The Nicoll Highway collapse
strongly highlights the need for SHM during construc- Strain gauges are the most widely used sensors for
tion, especially in urban centres where critical structures structural behaviour monitoring. On a structural sur-
are likely to be situated on the ground surface above the face, strains are caused by member deformations result-
excavation/construction sites. SHM can provide suffi- ing from bending, torsion, shearing and elongation/
cient warning before serious damage or failure of the contraction. Hence, strain measurements can capture
soil supporting structures. an elementÕs behaviour quite well (Sanayei and Saletnik,
Although the importance of SHM is widely recogni- 1996). For efficient performance, it is desirable that the
sed, it has still not developed to the stage of being recog- gauge should be stable with respect to both time and
nised as a crucial element of the overall underground temperature. In addition, it should have minimum
asset management system. The main reason for this is dimensions and inertia, and should exhibit linear re-
that SHM has not yet been a part of the relevant codes sponse over the strain range of interest (Dally et al.,
of practice. Another reason is that so far, detailed ben- 1984). Commercial strain gauges are available in various
efit cost analysis have not been carried out with respect types, based on mechanical, electrical or optical princi-
to underground structures. At the same time, no means ple. Described below are the prominent types of strain
of certification is available on a broader basis. gauges available commercially.
This paper addresses major issues related to the SHM
of underground structures. A detailed description of 2.1.1. Vibrating wire strain gauges (VWSG)
various available sensing technologies is covered, high- Fig. 3 shows the fabrication details of a typical
lighting the special problems encountered in subsurface VWSG. A VWSG essentially consists of a pretensioned
environment. Practical advantages of SHM are demon- stainless steel wire whose ends are fixed to lugs that are
strated by means of several case studies, including those spot-welded to the monitored component. A sensor coil,
carried out by the authors. positioned above the wire, when energised, plucks the
wire and measures the frequency of the resulting vibra-
tions. From the theory of vibrations, the natural fre-
2. Sensor systems for SHM quency of vibration, f, is related to the tension F in
the wire by
Comprehensive structural monitoring can be realised rffiffiffiffi
1 F
only if the structure to be monitored is instrumented f ¼ ; ð1Þ
with arrays of sensor systems at all critical locations. 2l m
At the same time, it is implausible to expect that any where l denotes the length of the wire and m its mass per
one type of sensors alone would be able to track down unit length. Any change in the strain, De, causes a
the complete structural behaviour as well as to detect change in tension by DF, thereby altering the natural fre-
all possible structural abnormalities. Hence, comprehen- quency by Df. For small strains, using HookeÕs law, the
sive monitoring calls for deploying complementary sen- change in strain can be expressed as (Batten et al., 1999)
sor systems with sufficient redundancy, so that a few of  m
the sensors could be permitted to fail without triggering De ¼ 4l2 ðDf Þ2 ; ð2Þ
total collapse of the monitoring system (Boller, 2002). In
where Y is YoungÕs modulus of the wire and A its cross-
addition, the sensors and the associated data retrieval
sectional area.
systems should be capable of withstanding the harsh
conditions encountered underground during construc-
tion and operation.
In general, SHM sensors can be classified as surface- Vibration sensor
bonded type and embedded type. The surface-bonded Vibrating wire
sensors can be replaced if they develop fault at any Spot weld
stage. However, there is very limited possibility of repair
or replacement for the embedded sensors. Hence, Host structure
embedded sensors, if employed in underground struc-
tures, should be exceptionally robust and durable. The
following parts of this section cover the operating prin- Fig. 3. A vibrating wire strain gauge.
490 S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500

If VWSGs are employed for long-term monitoring of underground structures, primarily based on VWSGs,
structural components which could to be subjected to spanning five years. This clearly demonstrates the
temperature fluctuations, proper care should be exer- robustness and longevity of VWSGs.
cised in interpreting their readings. If the structural com- The main drawback of the VWSGs is that they are
ponent and the VWSG have the same coefficient of only suitable for measuring static strains since they re-
thermal expansion, a, and if the monitored component quire plucking of wire. They are also susceptible to
is unrestrained, each will undergo a free thermal expan- extraneous noise in the form of ambient vibrations. If
sion strain aDT for a temperature change of DT. Thus, installed externally (such as on steel struts used for sup-
no additional stress would occur either in the monitored porting temporary retaining walls in deep excavations),
component or the VWSG. Consequently, the reading of special protection is required to prevent damage from
the VWSG will not undergo any change. If now an exter- routine construction activities.
nal load is applied on the component, strain will develop
due to the applied load. Thus, in real situations where the 2.1.2. Electrical strain gauges (ESG)
two effects (thermal expansion and external load) are ESGs are based on the principle that under a
superimposed, the VWSG will only capture the strain mechanical stress, the electrical resistance of a conduc-
in the member resulting from external loads. Hence, no tor varies in proportion to the load induced strain. An
correction is necessary in the strain measured by a ESG essentially consists of thin metallic foil grids,
VWSG if the coefficient of thermal expansion of the bonded to a thin, tough and flexible polyimide plastic
monitored component is the same as that of the VWSG. film, which can be adhesively bonded to the surface of
However, if the coefficient of thermal expansion of the the monitored component, as shown in Fig. 4(a). The
monitored component is different, say b, and if the com- polyimide film provides electrical insulation between
ponent is unrestrained, a strain equal to (b  a)DT will the gauge and the monitored component. When the
develop in the VWSG (and will be indicated by it). This structural component is loaded, its strain is transferred
is spurious, since it does not cause any stress in the mon- to the foil grid and its resistance changes accordingly.
itored component. Hence, in general, when load (or con- The relative change in resistance, DR/R of the foil, is re-
straints) and temperature fluctuation occur concurrently, lated to the strain e by (Dally et al., 1984)
the measured strain em should be corrected as
ecorr ¼ em  ðb  aÞDT . ð3Þ ¼ S g e; ð4Þ
VWSGs are structurally strong and quite robust for where Sg is called the gauge factor or the calibration
use in underground structures. They are especially suit- constant of the ESG. It varies between 2 and 4 for most
able for long-term monitoring since vibration wires do alloys used in strain gauge fabrication, such as constan-
not undergo any decay with time (Oosterhout, 2003). tan, karma and nichrome. The output DR/R of the strain
They can be easily spot welded to the reinforcement bars gauge is converted into a voltage signal by means of a
of the reinforced concrete structural members. This Wheatstone bridge circuit, as shown in Fig. 4(b). For
instrumentation technique was adopted in the bridge this circuit, the output voltage Vo is given by
linking Singapore and Malaysia, which was extensively  
1 DR 1
instrumented with VWSGs during construction in 1997 Vo ¼ Vi ¼ V i S g e. ð5Þ
4 R 4
(Moyo, 2002). The VWSGs installed on the bridge are
still operational, even after eight years. Oosterhout The input voltage, Vi, is governed by the power that can
(2003) reported a similar monitoring programme for be dissipated by the gauge, which in turn depends on the

Fig. 4. (a) An electrical strain gauge foil; (b) wheatstone bridge circuit.
S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500 491

gauge length (generally between 0.2 and 100 mm) and where they are likely to encounter excessive dampness.
the initial resistance of the gauge. The output voltage In general, ESGs tend to be less stable as compared to
usually ranges between 1 and 10 lV per microunit of VWSGs over long periods of time. Because of this rea-
strain. son, majority of the long-term studies reported in the lit-
Like VWSGs, the strain measured by an ESG also erature have employed VWSGs rather than ESGs.
needs to be compensated against temperature. A tem- However, ESGs are preferred over VWSGs for short-
perature change causes change in Sg as well as a change term measurements due to their low cost.
in the original gauge resistance (R), in addition to caus-
ing the spurious strain equal to (b  a)DT. The temper- 2.1.3. Optical fibre bragg grating (FBG) based strain
ature related changes in Sg are very small, usually less gauges
than 1% for a temperature variation of 100 C. Hence, Optical fibres, which are thin fibres (few lm to few
they can be neglected in routine stress analysis, unless hundred lm in diameter) of glass and silica, utilise fibre
temperature variations of several hundred degrees are properties to generate optoelectronic signals indicative
expected. Other effects, if combined, lead to a spurious of the external physical parameters to be measured.
change in the resistance of the gauge, given by Though originally developed for telecommunication
  purposes, they have found tremendous recognition as
¼ ðb  aÞS g DT þ cDT ; ð6Þ sensors since the 1990s. Of the various available fibre
R spurious
optic technologies, FBG has proven to be most versatile
where c is the temperature coefficient of resistivity of the (Tjin et al., 2002). A Bragg grating is a periodic struc-
gauge alloy. A temperature compensated gauge can be ture, fabricated by exposing a photosensitised fibre to
obtained if both the terms in Eq. (6) are zero or cancel an ultraviolet light. Fig. 5 illustrates the working princi-
each other. However, this rarely occurs, and that too ple of the FBG based strain gauge. As shown in this fig-
is over narrow temperature ranges only. Hence, in gen- ure, when light from a broad band source interacts with
eral, corrections should be made by subtracting Eq. (6) the grating, a single wavelength, known as Bragg wave-
from the measured DR/R (Eq. (5)). length, is reflected back while the rest of the signal is
ESGs demand considerable care during installation transmitted. The Bragg wavelength, kB, is related to
due to their fragile nature. Further, electrical noise is the grating pitch K and the effective refractive index neff
very frequently associated with ESGs since the output of the grating by
voltage from a Wheatstone bridge is of the order of kB ¼ 2Kneff . ð7Þ
few millivolts only. Fortunately, electrical noise can be
minimised to a permissible level by employing twisted An external mechanical strain in the fibre shifts the
leads with a properly grounded shield (Dally et al., Bragg wavelength through expansion/contraction of
1984). In addition, ESGs are very prone to deterioration the grating periodicity and the photo elastic effect. Sim-
by water. This problem was encountered during the ilarly, temperature variation causes thermal expansion/
monitoring of rock bolts in the underground caverns contraction of the grating periodicity and also the
in Singapore (Zhao et al., 2002). Hence, ESGs must be refractive index. These effects provide the means of
properly sealed if used in the underground structures, employing the FBG written fibres as the sensor elements

UV light

Phase mask
Reflected signal
Input signal

Optical fibre

Grating pitch
Transmitted signal

Fig. 5. Fabrication and principle of FBG based sensors.

492 S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500

for measuring strains and temperatures. The change in reason, efforts to install FBG sensors on civil structures
Bragg wavelength due to an external disturbance, com- often result in high rate of sensor failure due to the
prising a strain change De, a temperature change DT and presence of harsh environment (Storoy et al., 1997).
a pressure change DP, can be quantitatively expressed as They are yet to fully mature into standard field proven
DkB devices like the ESGs and the VWSGs. In addition,
¼ K e De þ K T DT þ K P DP ; ð8Þ the measurement system and the sensors themselves
are relatively expensive as compared to the conventional
where Ke, KP and KT are the wavelength sensitivity coef- sensor systems.
ficients for strain (e), temperature (T) and pressure (P)
respectively. Hence, in order to be used as strain sensor 2.2. Other commonly used sensor systems for SHM
(in the absence of pressure), the measured strain needs
to be corrected for temperature. Usually, this is achieved Table 1 summarises the salient features of extensom-
by installing an additional FBG close to the strain sens- eters, accelerometers, pressure transducers and tempera-
ing FBG, but not bonded to the structure. This mea- ture sensors, which are also commonly used for SHM.
sures the wavelength shift on account of temperature The next subsection describes the Ôpiezo-impedance
change alone, which can then be compensated. In addi- transducersÕ which have recently emerged as universal
tion, as in the case of VWSGs and ESGs, the spurious damage detection sensors.
strain (b  a)DT, due to differential thermal expansion
of the fibre and structure, also need to be eliminated 2.3. Piezo-impedance transducers
from the measured strain.
A comprehensive research programme is underway at Unlike the transducers described above, piezo-imped-
Nanyang Technological University (NTU), Singapore, ance transducers are relatively new type of sensors,
for the development of cost-effective and durable FBG barely 10 years old. They do not measure any direct
based sensors for civil-structural applications (Ng physical parameter like stresses, strains or temperatures.
et al., 1998; Tjin et al., 2001, 2002; Brownjohn et al., Rather, they extract a signature of the host structure to
2003a). The strain sensor developed at NTU consists analyse for the presence of any structural damage. The
of an FBG, about 10 mm long, sandwiched between car- technique utilizing the piezo-impedance transducers is
bon composite layers (50 mm long and 0.5 mm thick), as commonly known as the electro-mechanical impedance
shown in Fig. 6. Experimental evaluation of this FBG (EMI) technique, and is capable of detecting even the
sensor on lab sized structures by Moyo (2002), through incipient damages.
static and dynamic tests, demonstrated good possibility Piezo-impedance transducers are made up of piezo-
of its widespread use on civil structures. Several other electric materials such as lead zirconate titanate (acrony-
applications of FBG sensors on ground based structures med PZT), and are referred to as piezoelectric–ceramic
have also been reported (Storoy et al., 1997; Tjin et al., patches or PZT patches. Fig. 7(a) shows the details of a
2002). Liu et al. (2002) recently demonstrated feasibility typical commercially available PZT transducer, suitable
of their application in monitoring underground for use as a piezo-impedance transducer. The PZT
structures. patches exhibit two special effects due to the phenomenon
FBG sensors are small, lightweight, corrosion resis- of piezoelectricity. They generate surface charges in re-
tant and durable. VWSGs and ESGs require cables for sponse to mechanical stresses applied in the plane of the
recording data, which, for long distance monitoring, patch, which is called the Ôdirect effectÕ. Conversely, they
suffer from electro-magnetic interference and electrical undergo mechanical deformations in response to an elec-
noise. FBG sensors, on the other hand, are immune to tric field applied across their thickness, which is known as
the electro-magnetic interference and can be multi- the Ôconverse effectÕ. The EMI technique utilises both the
plexed, thus eliminating long cables. However, they are effects concurrently in total synergy.
very fragile as compared to VWSGs and ESGs. For this In the EMI technique, the PZT patch is bonded to the
surface of the monitored structure using a high-strength
epoxy adhesive and electrically excited by means of an
impedance analyser. In this configuration, the patch
(length 2l, width w and thickness h) behaves as a thin
bar undergoing axial vibration, as shown in Fig. 7(b).
An electro-mechanical model of the system is shown in
Fig. 7(c), where the structure has been replaced by two
equal mechanical impedances Z. The complex electro-
mechanical admittance Y (inverse of electrical imped-
ance) of the coupled system shown in this figure can
Fig. 6. FBG based strain sensor developed at NTU (Moyo, 2002). be derived as (Liang et al., 1994)
S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500 493

Table 1
Salient features of some common sensors used in SHM
S.no. Device Purpose Remarks
1 Extensometers
(a) Tape extensometer Measurement of convergence/divergence Both extensometers entail manual
(relative displacement between two points), recording, which proves tedious Slight
e.g. in a rock caverns after construction loosening of convergence pins could
(b) Borehole extensometer Measurement of relative displacements severely affect measurement accuracy
between several points. Can provide
displacement distribution in large rock volumes
2 Accelerometers Measurement of dynamic response, Drawbacks include bulkiness, small
either harmonic (e.g. vibration tests) bandwidth, high cost and susceptibility to
or transient (e.g. earthquakes) mechanical and electrical noise
3 Pressure transducers Measurement of pressure When used in concrete, shrinkage of
(a) Diaphragm type Measurement range between 200 and 700 MPa, concrete often causes problems
up to 10 kHz frequencies
(b) Quartz based Measurement range 700 MPa,
up to 200 kHz frequencies
4 Temperature sensors Measurement of temperature
(a) Expansion type Bulky, generally provide visual readouts
only, not suitable for automated
(b) Resistance temperature detectors Miniaturised, temperature measurement
(RTD) can be automated; but tend to be unstable
near upper limit
(c) Thermistors More miniaturised, sensitive and stable as
compared to RTDs
(d) Thermocouples Demand maintaining constant
temperature at one terminal

3 (z)
1 (x) Alternating electric 3
10mm 2
field source E3
PZT Patch 1
l l PZT patch

Z w Z
Top electrode Host structure
film Structural
Bottom electrode film
l l Impedance
wrapped to top surface
(a) (b) (c)

Fig. 7. (a) A commercial PZT patch; (b) principle of EMI technique; (c) equivalent model of system.

wl T Za 2 E tan jl 2 E host structure is a function of the structural parameters,
Y ¼ 2xj e þ d Y  d 31 Y ;
h 33 Z þ Z a 31 jl i.e., the stiffness, the damping and the mass. Any dam-
ð9Þ age to the structure will cause these parameters to
change, and hence changes the drive point mechanical
where d31 is the piezoelectric strain coefficient of the impedance Z. Consequently, as can be seen from
PZT material, Y E the complex YoungÕs modulus under Eq. (9), the electromechanical admittance, Y , will under-
constant electric field, eT33 the complex electric permittiv- go change, and this serves as an indicator of the state of
ity at constant stress, Za the mechanical impedance of health of the structure. Due to high frequency of excita-
the PZT patch, x the angular frequency, and j is the tion, the EMI has very high sensitivity to damage, typi-
wave number. cally of the order of the ultrasonic techniques (Park
The electromechanical coupling represented by Eq. et al., 2003). Typically, the technique can detect flexural
(9) is utilised in damage detection in the EMI technique. and shear crack before they could be visible to the naked
In this equation, the mechanical impedance Z of the eyes (Bhalla and Soh, 2004a). Fig. 8 illustrates how
494 S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500

PZT patch 10x10x0.3mm gated without necessitating the removal of finishes. No

complex data processing or any expensive hardware is
warranted. By means of an array of such transducers,
damage location can be easily identified (Soh et al.,
Hole of 5mm
2000). The technique has far greater sensitivity to dam-
age than the global methods. Several proof of concept
10mm 48mm non-destructive evaluation (NDE) applications of the
EMI technique have been reported in the literature
(Sun et al., 1995; Ayres et al., 1998; Soh et al., 2000;
0.005 Bhalla and Soh, 2004a).
Conductance (S)

After Before Although the EMI technique has been demonstrated
damage damage
to be highly sensitive to damage, technical details such
as sensor packaging, instrumentation and long-term
protection are not yet standardised. Hence, special con-
siderations need to be addressed before applying the
180 184 188 192 196 200
EMI technique on underground structures. A compre-
(b) Frequency (kHz)
hensive research programme is currently underway at
NTU for developing low-cost practical SHM systems
Fig. 8. An application of EMI technique: (a) test structure; (b) effect of based on the EMI technique. This includes the develop-
damage (two holes) on conductance signature.
ment of realistic PZT-structure phenomenological mod-
els (Bhalla and Soh, 2003, 2004b,c; Yang et al., 2005) as
damage can be interpreted from the measured admit- well as the development of damage localisation and
tance signature by means of a practical example, an alu- quantification algorithms (Xu et al., 2004; Naidu and
minium block of size 50 · 48 · 10 mm, instrumented Soh, 2004). Research groups in USA are carrying out
with a PZT patch 10 · 10 · 0.3 mm in size (see Fig. active research for the development of low-cost and por-
8(a)). Damage was experimentally induced in the struc- table signature acquisition systems and wireless moni-
ture by drilling two 5 mm diameter holes through the toring (Park et al., 2003; Peairs et al., 2004) that will
thickness of the specimen. Fig. 8(b) shows the effect of render the technique more economical and standardised
the induced damage on the conductance (real part of in the near future.
admittance) signature in 180–200 kHz frequency range.
A major resonant peak can be observed to have shifted
to the left on account of this damage. A simple but crude 3. Practical benefits from SHM of underground
way to quantify damage is to use statistical quantifiers structures: case studies
such as root mean square deviation or correlation coef-
ficient between the signatures corresponding to the This section presents a glimpse of the results and the
undamaged and damaged states (Park et al., 2000; Bhal- experience arising from the instrumentation and moni-
la and Soh, 2004a). However, very recently, Bhalla and toring of underground facilities in Singapore and other
Soh (2004b,c) formulated a parametric approach by parts of the world. Specifically, practical benefits arising
identifying structural parameters such as stiffness and from such instrumentation are highlighted.
damping inherent in the host structure. Using this ap-
proach, the identified stiffness of the aluminium block 3.1. Monitoring deep strutted excavations using VWSGs
was found to reduce from 5.18 · 107 N/m in the pristine
state to 4.55 · 107 after the damage. By calibrating such Temporary walls, supported laterally by props, are
structural parameters with damage severity, as demon- often used for retaining the sides of an excavation for
strated by Soh and Bhalla (in press) for concrete, mean- constructing basements. Coutts et al. (2001) reported
ingful interpretation can be easily made from the monitoring earth pressures in such excavations, made
measured electro-mechanical admittance signatures. for constructing two mass rapid transit (MRT) stations
Due to high sensitivity, the EMI technique has at- in Singapore. The excavations were temporarily retained
tracted intensive research during the last eight years. It by driven soldier piles and timber lagging, supported lat-
also has several other advantages over the conventional erally by multi level double I-beams. The maximum
monitoring techniques which rely on the measurement depth of the excavation was 25 m in the Serangoon sta-
of loads, stresses or strains since it does not warrant tion and 23.4 m in the Woodleigh station. In order to
any complex analytical or numerical modelling of the monitor the load in an I-beam, two VWSGs were in-
monitored structure. It employs low-cost and low-power stalled on the web, 100 mm from the top and bottom
demanding PZT patches, which can be non-intrusively flanges. Averaging of measurements from the two
bonded to the monitored structure and can be interro- gauges ensured elimination of bending strain if any.
S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500 495

The gauges were instrumented 3 m from the ends of the bending moments. For the configuration shown in
beams in order to circumvent possible end effects. No Fig. 9, the bending moments about the xx and yy axes
temperature corrections were performed, assuming al- at the instrumented section were determined as
most equal thermal expansion coefficients for the gauges
YI ðetop  ebottom Þ
and the struts. M xx ¼ ;
y 2
By monitoring the strut strains, the investigators ð10Þ
found that the measured earth pressures differed signifi- YI ðeleft  eright Þ
M yy ¼ ;
cantly from those predicted theoretically. At the Seran- x 2
goon station, the measured earth pressures were much where Y denotes YoungÕs modulus of elasticity of the
lower than the theoretical predictions, whereas at the prop, I the moment of inertia, e the measured corrected
Woodleigh station, it was the other way round. It was strain, and x and y the coordinates of the strain gauges.
also found that at several strut locations, the compres- The axial load was determined as
sion forces exceeded their design values by as much as
51%. Using the measured earth pressures, Couts and P¼ ðetop þ ebottom þ eleft þ eright Þ; ð11Þ
co-workers were able to estimate the correct soil param- 4
eters, which were subsequently utilised for more accu- where A denotes the cross-sectional area of the prop.
rate earth pressure calculations as well as for ensuring Batten and co-workers recommended installing at least
a safer and more economical design of the temporary four gauges at a section, especially if the instrumenta-
supports at the other stations. tion is desired near the ends of the prop from practical
Batten et al. (1999) similarly reported monitoring point of view. Averaging eliminates the errors due to
strains and temperatures on circular steel props, about non-uniform stress distributions, even if very severe.
1 m in diameter, supporting temporary retaining walls Batten and co-workers also strongly recommended rep-
of the underground excavations. At any monitored sec- etition of readings to eliminate effects of extraneous
tion of the prop, four special VWSGs, which had built- noise, especially due to ambient vibrations.
in thermistors (for simultaneous recording of both strain Monitoring prop loads can help in economizing the
and temperature) were installed. Fig. 9 shows the sche- design of temporary support systems. This can also en-
matic arrangement of the VWSGs at any monitored sec- sure a safer work environment for the workers as well
tion. Each prop was instrumented at both ends and the as the general public in the areas surrounding the con-
data was recorded at an interval of 2 h. The presence of struction sites by preventing accidents during under-
thermistors facilitated temperature measurement, en- ground excavations.
abling accurate temperature compensation of the mea-
sured strain, as well as estimation of the load 3.2. Monitoring deflections of large underground cavity
variations on account of temperature changes. The prop
load on account of temperature rise was found by the Simple instrumentation can immensely help engineers
measurements to be typically about 52–63% of that for in making decisions pertaining to the safety, optimisa-
fixed-end conditions. Batten et al. (1999) therefore rec- tion and upgrading of cavern supports, thereby prevent-
ommended frequent monitoring of the strains for accu- ing unnecessary wastage of resources. This fact can be
rate recording of maximum prop strains. It was also better appreciated in light of the monitoring programme
noted that the top of the prop was about 12 C hotter of an underground cavern reported by Goel (2001). The
than the bottom during summers, leading to additional monitored underground structure was a large rock cav-
ern in India, 23 m wide and 57 m high, as shown in
Fig. 10. After completing the excavation work, the roof
displacement was monitored for a period of 30 months.
(VWSG)top A slow but continuous displacement of 0.024 mm/
month was observed (see Fig. 10) from these measure-
Prop (1m in diameter)
ments. This small but continuous movement was a mat-
ter of concern for engineers, although no adverse
symptoms were observed visually. To correct the prob-
x lem, additional supports were installed on the roof in
(VWSG)left (VWSG)right the form of longer rock bolts. In order to verify the effi-
cacy of the upgradation, displacement monitoring was
continued for another three years. It was found that
(VWSG)bottom the displacement was very effectively arrested by the
longer rock bolts, as evident from Fig. 10. This case
Fig. 9. Instrumentation of a section of prop with VWSG (after Batten strongly supports the benefits of monitoring excavations
et al., 1999). and other structures after construction.
496 S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500

Fig. 10. Monitoring roof displacement of a large underground cavity (Goel, 2001).

3.3. Monitoring tunnel deformations due to adjacent


Massive underground tunnelling works were carried

out in the central business district of Singapore during
the mid eighties, as a part of the MRT project. Most
tunnels were constructed using tunnel boring machines
and only a few were constructed using conventional
mining. The MRT CorporationÕs Code of Practice does
not allow any construction in close proximity of the tun-
nels that might displace the tunnels by more than 15 mm
(Sharma et al., 2001). However, the scarcity of land in
the central business district necessitated a few construc-
tions in close proximity of the MRT tunnels. One such
project was the redevelopment of Tan Tock Seng Hospi-
tal, described in detail by Sharma et al. (2001). The pro-
ject involved the construction of a 15-storey hospital
building, with two levels of basement. A large excava-
tion, about 200 m long, 140 m wide and 15 m deep
was necessary. Fig. 11 shows the layout and cross-sec-
tion of the planned excavation work. As shown in the
figure, two MRT tunnels, 6 m in diameter, criss–crossed
the proposed construction site. Significant displace-
ments were expected to occur in the tunnel lining due
to stress relief in the ground caused by the excavation. Fig. 11. (a) Layout of site. (b) Cross-section along G–G (Sharma
Finite element analysis revealed that the tunnel dis- et al., 2001).
placements were likely to be within the permissible limits.
However, due to the critical nature of the tunnels, exten- ground instrumentation consisted of water standpipes,
sive field instrumentation was additionally carried out in piezometers, inclinometers (indicated by I1, I2 and I3 in
order to ensure compliance of codal requirements. The Fig. 11(a)) and settlement markers. In addition, two
S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500 497

automatic tunnel displacement monitoring systems, each 3.5. Dynamic strain measurement using FBG system
consisting of a motorised total station, were installed.
These systems measured the tunnelÕs displacements with Liu et al. (2002) developed a FBG based dynamic
respect to a few stable reference points located suffi- strain measurement system, capable of detecting dy-
ciently far away from the excavation site (therefore unli- namic strain signals of up to 5 kHz frequency in the
kely to be affected by the excavation). With these underground structures. The researchers demonstrated,
systems, the displacements of the tunnel lining were con- by means of a laboratory experiment, that rock bolts
tinuously monitored during the excavation and construc- supporting excavations in hard rocks can be employed
tion. The maximum horizontal displacement recorded by as sensor heads by bonding a FBG at the anchor head.
the instruments was 6.0 mm, against 8.0 mm predicted Fig. 12 shows the scheme of the setup employed by the
by the finite element method (FEM). Similarly, the max- researchers. In this system, a 3 dB m distributed feed-
imum vertical displacement was measured to be 3.8 mm, back laser, in 1548.75–1551.25 mm wavelength range,
against 3.0 mm predicted by FEM. Thus, the instrumen- sent a signal at an optimised wavelength through a fibre
tation provided a direct evidence that construction at the optical circulator. Part of the signal was reflected back
site did not violate the codal requirements and the MRT from the grating and converted into an electrical signal.
tunnels were cleared as safe for operation. The intensity change of the optical signal provided a
measure of the strain in the rock bolt. The system was
3.4. Fibre optic based systems for monitoring underground found capable of measuring strains as low as 109 at a
structures frequency of 1.7 kHz.
However, the system was reportedly prone to noise
Inaudi et al. (1994, 1995) reported the development due to laser instability, disturbance from optical cables
and application of a fibre optic low-coherence interfer- and connectors, and the harsh conditions found under-
ometry technique for measuring displacements in under- ground. It was demonstrated that a powerful laser
ground structures. The system could measure source, coupled with an adaptive filtering algorithm,
displacements with an accuracy of 10 lm, over a base could effectively reduce the noise levels, besides enhanc-
ranging from 1 to 50 m. Thus, it measured overall rather ing the sensitivity of the system.
than local strains.
The technique consisted of employing two single 3.6. Structural monitoring of underground tunnels
mode communication fibres, one in mechanical contact
with the structure and the other loosely in a neighbour- Bakker (2000) reported comprehensive investigation
ing pipe. A low-coherence double Michelson interferom- involving monitoring of the Second Heinenoord and
eter detected the difference in length between the two Botlek railway tunnels in the Netherlands, during and
fibres, through the formation of interference fringes. after construction. Additional results were published
Since the displacement information was encoded in by Oosterhout (2003). The overall approach adopted
coherence properties of light and not its intensity, the by these investigators consisted of three aspects – predic-
technique was immune against connector losses or tion, monitoring and evaluation.
changes in the fibre. It was also capable of taking mea- Prediction was carried out using 1D analytical and
surements at multiple points along the fibre. Being por- 3D FEM models, with the inherent assumption that
table, waterproof and battery powered, the system was assembly of the lining was perfect, i.e. free from misa-
robust for the underground structures. lignments and imperfections. In addition, a homoge-
Inaudi and co-workers reported experimental success neous stress distribution was assumed across the
of the technique in laboratory tests involving a concrete tunnel cross-section.
beam, 1 · 0.5 · 5 m in size, under tension, and in moni- For the purpose of monitoring, two rings at different
toring the drying shrinkage and thermal deformations of locations along the tunnel, were instrumented with
a large slab, 20 · 5 · 0.5 m in size. They also reported strain gauges, pressure transducers and linear variable
monitoring several real-life underground structures such differential transducers (LVDTs). For strain monitoring,
as anchorage cables supporting a rock wall (fibre em- a total of five pairs of VWSGs, with a resolution of
ployed to measure change in length between anchor 1 lm/m, were instrumented in each segment, which mea-
head and grouted zone) and monitoring strains along sured two axial and three tangential strain components.
the length of cast-in-situ concrete piles (fibres were at- VWSG principle was adopted due to its negligible time
tached to the reinforcement bars). decay. In addition, two pressure cells (consisting of mer-
With the recent advent of FBG, which is technologi- cury filled reservoir, with 0–3500 kPa range and 1 kPa
cally more advanced than the low-coherence interferom- resolution) were instrumented. The LVDTs, which had
etery reported by Inaudi and co-workers, there should a range of ±5 mm and a resolution of 25 lm, were in-
be much better prospects of monitoring underground stalled to measure the axial displacements between two
structures using fibre optic technology. adjacent rings and the tangential displacements between
498 S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500

Fig. 12. Measurement system of Liu et al. (2002).

two adjacent segments within the ring. A total of 200 de- (v) Since stress paths may criss–cross in the ring joints,
vices, consisting of 146 VWSGs, 28 pressure cells and 18 Bakker and co-workers recommended that at least
LVDTs were installed on the two rings, out of which two or three adjacent rings should be monitored. In
only five were found to be spoilt during construction. addition, monitoring should preferably start right
Before starting the field monitoring, the strains were cal- from the assembling stage itself so that the assem-
ibrated against axial force, tangential force and tangen- bly stresses could also be carefully monitored.
tial bending moment of the lining through laboratory
tests. Monitoring was continuously performed between Bakker and co-workers implemented some of these
1996 and 1999. During this period, the researchers re- lessons during monitoring of the Botek Railway tunnel.
ported two major problems – (i) damage to the data From this, they concluded that the process of assem-
cables during construction; and (ii) accumulation of bling the lining and the subsequent 24 h are the most sig-
water behind the tail of the tunnel boring machine nificant ones for the development of strains. However,
(TBM), causing damage to several connectors. they noted that strain measurements provided only local
Following major observations made by the research- and not global information such as bending moments. It
ers after evaluating the measurements are worth noting: was recommended that laser based measurement tech-
niques could provide a more direct assessment of the
(i) The assembly process itself led to local stresses in global response. In addition, wireless data retrieval
the tunnel lining, of magnitude comparable to could circumvent some of the problems faced by the
those induced by the external loads. In addition, cable based data retrieval system. Hence, there is a need
the stress distribution was highly non-uniform. to seriously consider the recommendations of these
Hence, both the assumptions made for prediction researchers in future experiments involving tunnels,
proved incorrect. which can ensure more useful information.
(ii) The predictions made by the FEM and the analyt-
ical models proved to underestimate the lining
stresses. In addition, the calibration of strain 4. Conclusions
against bending moment and tunnel force was
invalid due to the reasons stated in (i). Once constructed, underground structures, such as
(iii) Large displacements were observed at the segment transportation tunnels, caverns and retaining walls are
joints. The number of LVDTs instrumented was rarely replaceable. Usually, critical regions in these
therefore insufficient. The researchers recom- structures are inaccessible for visual inspections. In
mended that all the joint segments should be addition, there are numerous considerations specific to
instrumented with LVDTs. In addition, due to each underground structure, such as earth pressures,
non-uniform stress distribution, strain gauges also ground movements due to construction, underground
proved insufficient in number. formations and ground water fluctuations, quite often
(iv) Due to different thermal expansion coefficients of difficult to predict accurately at the design stage and
mercury and concrete, strong temperature depen- very rarely monitored after the structure has been con-
dence of the measured pressure was observed. This structed. This paper presented a detailed review of the
produced a fake load effect. The VWSG data, on current status and the importance of SHM for under-
the other hand, were corrected against temperature. ground structures, during construction as well as opera-
Still, mild temperature dependence was observed tion. An in-depth review of the various available sensor
due to partially restraining end conditions. technologies, and their pros and cons with regard to
S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500 499

underground structures are also covered. The critical Bhalla, S., Soh, C.K., 2004c. Structural health monitoring by piezo-
nature of the underground structures as well as their impedance transducers. Part II: Applications. Journal of Aerospace
Engineering, ASCE 17, 166–175.
complex interactions and influences on the nearby Boller, C., 2002. Structural health management of ageing aircraft and
ground structures make a strong case for pursuing active other infrastructure. Monograph on Structural Health Monitoring,
research and development in the area of SHM. Long- Institute of Smart Structures and Systems (ISSS), Bangalore, pp. 1–
term monitoring of the underground structures can go 59.
a long way in ensuring greater safety as well as design Brownjohn, J.M.W., Moyo, P., Rizos, C., Chuan, T.S., 2003a.
Practical issues in using novel sensors in SHM of civil infrastruc-
validation and optimisation, thereby ensuring a high ture: problems and solutions in implementation of GPS and fibre
benefit–cost ratio. optics. In: Proceedings, 4th International Workshop on Structural
At the present moment, SHM is not a codal require- Health Monitoring, Stanford University, CA, September 15–17,
ment for underground structures. The problem is further pp. 499–506.
aggravated by the fact that there is no unique solution in Brownjohn, J.M.W., Moyo, P., Omenzetter, P., Lu, Y., 2003b.
Assessment of highway bridge upgrading by dynamic testing and
terms of instrumentation technology and monitoring finite-element model updating. Journal of Bridge Engineering,
algorithms since each underground structure is unique ASCE 8, 499–506.
in itself. The harsh and rough environment of the under- Coutts, D.R., Wang, J., Cai, J.G., 2001. Monitoring and analysis of
ground structures necessitates active research in the results for two strutted deep excavations using vibrating wire strain
development of more durable and robust sensors and gauges. Tunnelling and Underground Space Technology 16, 87–92.
Dally, J.W., Riley, W.F., McConnell, K.G., 1984. Instrumentation for
data retrieval systems. In particular, the emphasis Engineering Measurements. Wiley, New York.
should be focused on the new generation ÔsmartÕ tech- Farrar, C.R., Jauregui, D.A., 1998. Comparative study of damage
nology based sensor systems, such as the piezo-imped- identification algorithms applied to a bridge: I. Experiment. Smart
ance transducers, the FBG and the MEMS based Materials and Structures 7, 704–719.
technologies. Development of wireless technology based Goel, R.K., 2001. Status of tunnelling and underground construction
activities and technologies in India. Tunnelling and Underground
remote data retrieval systems should also be pursued. Space Technology 16, 63–75.
In spite of so many challenges, any instrumentation of Inaudi, D., Elamari, A., Pflug, L., Gisin, N., Breguet, J., Vurpillot, S.,
the structure, however elementary, promises significant 1994. Low-coherence deformation sensors for the monitoring of
information that could help better understand its behav- civil engineering structures. Sensors and Actuators 44, 125–130.
iour as well as play an important role in the validation of Inaudi, D., Vulliet, L., Pflug, L., Vurpillot, S., Wyser, A., 1995. Low-
coherence sensors for the monitoring of underground works. In:
key design assumptions. Thus, the underground struc- Proceeding, North American Conference on Smart Structures and
tures such as tunnels, caverns, mines, piles and anchored Materials, San Diego, vol. 2444, pp. 171–178.
walls are suitable candidates for channelling research Liang, C., Sun, F.P., Rogers, C.A., 1994. Coupled electro-mechanical
funds for developing appropriate monitoring systems. analysis of adaptive material systems – determination of the
actuator power consumption and system energy transfer. Journal
of Intelligent Material Systems and Structures 5, 12–20.
Liu, J.-G., Hattenberger, C.S., Borm, G., 2002. Dynamic strain
References measurement with a fibre Bragg grating sensor system. Measure-
ment 32, 151–161.
Aktan, A.E., Helmicki, A.J., Hunt, V.J., 1998. Issues in health Magnus, R.R., Teh, C.I., Lau, J.M., 2005. Interim report of the
monitoring for intelligent infrastructure. Smart Materials and Committee of Inquiry into the incident at the MRT Circle Line
Structures 7, 674–692. Worksite that led to the collapse of Nicoll Highway on 20 April
Aktan, A.E., Catbas, F.N., Grimmelsman, K.A., Tsikos, C.J., 2000. 2004, Singapore. Available from: <http://www.mom.gov.sg/MOM/
Issues in infrastructure health monitoring for management. Journal CDA>.
of Engineering Mechanics, ASCE 126, 711–724. Moyo, P., 2002. Structural performance monitoring and health
Ayres, J.W., Lalande, F., Chaudhry, Z., Rogers, C.A., 1998. Quali- assessment of highway bridges. Ph.D. Thesis, Nanyang Techno-
tative impedance-based health monitoring of civil infrastructures. logical University, Singapore.
Smart Materials and Structures 7, 599–605. Naidu, A.S.K., Soh, C.K., 2004. Damage severity and propagation
Bakker, K.J., 2000. Soil retaining structures: development of models characterization with admittance signatures of piezo transducers.
for structural analysis. Ph.D. thesis, Delfts University of Technol- Smart Materials and Structures 13, 393–403.
ogy, Netherlands. Ng, S.L., Tjin, S.C., Xie, T.Q., Soo, K.T., 1998. Side-projected fiber-
Batten, M., Boorman, R., Leiper, Q., 1999. Use of vibrating wire strain optic velocity sensors. IEEE Photonics Technology Letters 10, 249–
gauges to measure loads in tubular steel props supporting deep 251.
retaining walls. Proceeding, Institution of Civil Engineers, Geo- van Oosterhout, G.P.C., 2003. Recent Dutch experiences in developing
technical Engineering 137, 3–13. structural monitoring systems for shield driven tunnels. HERON
Bhalla, S., Soh, C.K., 2003. Structural impedance based damage 48, 65–78.
diagnosis by piezo-transducers. Earthquake Engineering and Pandey, A.K., Biswas, M., 1994. Damage detection in structures using
Structural Dynamics 32, 1897–1916. changes in flexibility. Journal of Sound and Vibration 169, 3–17.
Bhalla, S., Soh, C.K., 2004a. High frequency piezoelectric signatures Park, G., Cudney, H.H., Inman, D.J., 2000. Impedance-based health
for diagnosis of seismic/blast induced structural damages. NDT&E monitoring of civil structural components. Journal of Infrastruc-
International 37, 23–33. ture Systems, ASCE 6, 153–160.
Bhalla, S., Soh, C.K., 2004b. Structural health monitoring by piezo- Park, G., Sohn, H., Farrar, C.R., Inman, D.J., 2003. Overview of
impedance transducers. Part I: Modeling. Journal of Aerospace piezoelectric impedance-based health monitoring and path for-
Engineering, ASCE 17, 154–165. ward. The Shock and Vibration Digest 35, 451–463.
500 S. Bhalla et al. / Tunnelling and Underground Space Technology 20 (2005) 487–500

Peairs, D.M., Park, G., Inman, D.J., 2004. Improving accessibility Tjin, S.C., Hao, J.Z., Lam, Y.Z., Ho, Y.C., Ng, B.K., 2001. A pressure
of the impedance-based structural health monitoring method. sensor using fiber Bragg grating. Fiber and Integrated Optics 20,
Journal of Intelligent Material Systems and Structures 15, 129– 59–69.
139. Tjin, S.C., Rupali, S., Moyo, P., Brownjohn, J.M.W., Ngo, N.Q.,
Sanayei, M., Saletnik, M.J., 1996. Parameter estimation of structures 2002. Application of fibre Bragg grating based sensors for civil
from static strain measurements. I: Formulation. Journal of infrastructure health monitoring. In: Proceedings, ISSS-SPIE
Structural Engineering, ASCE 122, 555–562. International Conference on Smart Materials, Structures and
Sharma, J.S., Hefny, A.M., Zhao, J., Chan, C.W., 2001. Effect of large Systems, Bangalore, 12–14 December, pp. 683–690.
excavation on deformation of adjacent MRT tunnels. Tunnelling Xu, J.F., Yang, Y.W., Soh, C.K., 2004. EM impedance-based
and Underground Space Technology 16, 93–98. structural health monitoring with evolutionary programming.
Soh, C.K., Bhalla, S., in press. Calibration of piezo-impedance Journal of Aerospace Engineering, ASCE 17 (4), 182–193.
transducers for strength prediction and damage assessment of Yang, Y.W., Xu, J.F., Soh, C.K., 2005. A generic impedance-based
concrete. Smart Materials and Structures, in press. model for structure–PZT interacting system. Journal of Aerospace
Soh, C.K., Tseng, K.K.H., Bhalla, S., Gupta, A., 2000. Performance of Engineering, ASCE 18 (2), 93–101.
smart piezoceramic patches in health monitoring of a RC Bridge. Zhao, J., 2003. The Underground Science City. International Tunnel-
Smart Materials and Structures 9, 533–542. ling Association, http://www.ita-aites.org/cms/335.html.
Storoy, H., Saether, J., Johannessen, K., 1997. Fiber optic condition Zhao, J., Cai, J.G., Tunbridge, L., Song, H.W., Zhang, X.H., Zhao,
monitoring during a full scale destructive bridge test. Journal of X.B., 2002. Results of rock cavern monitoring and assessment of
Intelligent Material Systems and Structures 8, 633–643. cavern stability. Technical Report for NTU-DSTA Joint R&D
Sun, F.P., Chaudhry, Z., Rogers, C.A., Majmundar, M., Liang, C., Project on instrumentation and monitoring of rock caverns during
1995. Automated real-time structure health monitoring via signa- construction and operation.
ture pattern recognition. In: Proceedings, SPIE Conference on Zhao, J., Liu, Q., Lee, K.W., Choa, V., Teh, C.I., 1999. Underground
Smart Structures and Materials, San Diego, CA, February 27– cavern development in the Jurong sedimentary rock formation.
March 1, vol. 2443, pp. 236–247. Tunnelling and Underground Space Technology 14, 449–459.