Scholarly article on topic 'An Integrated Risk Sensing System for Geostructural Safety'

An Integrated Risk Sensing System for Geostructural Safety Academic research paper on "Materials engineering"

Share paper
{"Risk control" / "Wireless sensor network (WSN)" / Geo-structure / Resilience / Robustness}

Abstract of research paper on Materials engineering, author of scientific article — H.W. Huang, D.M. Zhang, B.M. Ayyub

Abstract Over the last decades, geo-structures are experiencing a rapid development in China. The potential risks inherent in the huge amount of construction and asset operation projects in China were well managed in the major project, i.e. the project of Shanghai Yangtze tunnel in 2002. Since then, risk assessment of geo-structures has been gradually developed from a qualitative manner to a quantitative manner. However, the current practices of risk management have been paid considerable attention to the assessment, but little on risk control. As a result, the responses to risks occurrences after a comprehensive assessment are basically too late. In this paper, a smart system for risk sensing incorporating the wireless sensor network (WSN) on-site visualization techniques and the resilience-based repair strategy was proposed. The merit of this system is the real-time monitoring for geo-structural performance and dynamic pre-warning for safety of on-site workers. The sectional convergence, joint opening, and seepage of segmental lining of shield tunnel were monitored by the micro-electro-mechanical systems (MEMS) based sensors. The light emitting diode (LED) coupling with the above WSN system was used to indicate different risk levels on site. By sensing the risks and telling the risks in real time, the geo-risks could be controlled and the safety of geo-structures could be assured to a certain degree. Finally, a resilience-based analysis model was proposed for designing the repair strategy by using the measured data from the WSN system. The application and efficiency of this system have been validated by two cases including Shanghai metro tunnel and underwater road tunnel.

Academic research paper on topic "An Integrated Risk Sensing System for Geostructural Safety"

Accepted Manuscript

An Integrated Risk Sensing System for Geostructural Safety H.W. Huang, D.M. Zhang, B.M. Ayyub

PII: S1674-7755(16)30269-4

DOI: 10.1016/j.jrmge.2016.09.005

Reference: JRMGE 309

To appear in: Journal of Rock Mechanics and Geotechnical Engineering

Received Date: 13 June 2016 Revised Date: 13 September 2016 Accepted Date: 18 September 2016

Please cite this article as: Huang HW, Zhang DM, Ayyub BM, An Integrated Risk Sensing System for Geostructural Safety, Journal of Rock Mechanics and Geotechnical Engineering (2017), doi: 10.1016/ j.jrmge.2016.09.005.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Rock Mechanics and



—- —

2 An Integrated Risk Sensing System for Geostructural Safety

3 H.W. Huang1, D.M. Zhang2* & B.M. Ayyub3

4 1:Key Laboratory of Geotechnical and Underground Engineering, Department of Geotechnical

5 Engineering, Tongji University, Shanghai, China, Email:

6 2:Key Laboratory of Geotechnical and Underground Engineering, Department of Geotechnical

7 Engineering, Tongji University, Shanghai, China, Email: (Corresponding Author)

8 3: Center for Technology and Systems Management, Department of Civil and Environmental Engineering,

9 University of Maryland, College Park, MD, USA. Email:

11 ABSTRACT: In recent decades, the geo-structures of China are experiencing a rapid develop-

12 ment. The potential risks inherent in the huge amount of construction and asset operation

13 works in China were well managed, for the first time, in the mega project: The Shanghai Yang-

14 tze Tunnel in 2002. Since then, risk assessment of geo-structures has been gradually developed

15 from a qualitative manner to a quantitative manner. However, the current practice of risk

16 management has focused considerable attention on the assessment, but little on risk control.

17 As a result, the responses to risks after a comprehensive assessment are generally too late for

18 controlling the safety of geo-structures and geo-workers. In this paper, a smart system for risk

19 sensing incorporating the wireless sensoring network (WSN), on-site visualization techniques

20 and the resilience-based repair strategy is proposed. The merit of this system is the real-time

21 monitoring for geo-structural performance and dynamic pre-warning for safety of on-site

22 workers. The sectional convergence, joint opening, and seepages of segmental lining of shield

23 tunnel are monitored by the Micro-electromechanical-Systems (MEMS) based sensors. The

24 light-emitting-diode (LED) coupling with the above WSN system is used to indicate different

25 level of risk on site. By sensing the risk and telling the risk in real time, the geo-risks could be

26 controlled and the safety of geo-structures could be assured to a certain degree. Finally, a re-

27 silience-based analysis model was proposed for designing the repair strategy by using the

28 measured data from the WSN system. The application of this system has been validated by

29 two real cases including Shanghai metro tunnel and underwater road tunnel, respectively. The

30 efficiency and workability of this system has been verified using this field case studies.

31 KEYWORDS: risk control, wireless sensoring network, geo-structure, resilience, robust.


34 Accompanying with the fast development of infrastructure construction and asset operation in

35 China, there are huge amount of risks inherent in these geo-engineering works. It might poten-

36 tially cause an unsafe geo-structures and impact on environments. For example, in August 2014,

37 a boring hole for a site investigation in Guangzhou of China was driven through and completely

38 broke one segment lining of an operational metro tunnel buried directly under the drilling loca-

39 tion. The accident caused the operation of this metro line to stop. In February 2015, serious

40 leakage has occurred in a shield machine due to poor ground condition in Wuhan of China, and

subsequently the piping and sand boiling came out causing severe ground loss and large cracks of pavement of the road.

Reports show that a deadly accident approximately occurs every ten minutes in civil engineering construction site including geotechnical works (ILO, 2003). The geo-structural risk can be managed, minimized, shared, transferred or accepted, but it cannot be ignored (Latham, 1994). Hence, a rational risk management approach during the life-time of geo-structures should be of great necessity and is mostly concerned by the engineers and local governors. By definition, risk is a combination of frequency of occurrence of a defined hazard and the consequences of its occurrence (ITA, 2002). Casagrande (1965) classified risk into two major types. One is the engineering related, and the other is human related. Currently, the former type is emphasized and sub-divided into unknown risk and calculated risk. So far there has been many researches on the topic of risk assessment, no matter with qualitative perspective or quantitative perspective. Some selected seminal works on this topic include Ang and Tang (1975), Whitman (1984, 2000) and Lacasse (2015). However, the numerical value of risk produced by a comprehensive assessment can tell limited information about risk to the workers on site. It is better that reliability-based methods should be supported by large amounts of measured data from monitoring which captures the real performance of geo-structures, e.g., displacement, stress, forces and etc. (Langford, et al., 2016; Vardanega and Bolton, 2016).

Hence, the risk is better to be sensed quickly and translated easily so that human life and economic cost involved in the risk of geo-structures could be saved (Huang and Zhang, 2015). Structural Health Monitoring (SHM) is helpful to understand the safety state of operated geo-structures, in particular for large-scaled distributed geo-systems such as metro tunnel networks (O'Rourke, 2010). But it also has to be recognized that the application of sensing and monitoring would benefit little to owners when brittle response of geo-structural during construction or operation is encountered, such as collapse. Because the design of this type of geo-structure would have no time to be altered (Peck, 1969). Hence, it is accepted that sensing and monitoring systems could support engineering judgement and even complement it especially for the case of a ductile performance response, but will not attempt to replace it. If the monitoring for structural system is too late for recovery of structure, then, the real time pre-warning should be helpful to protect the on-site workers for their safeties. In recent years, micro-electro-mechanical systems (MEMS) and wire-less sensor network (WSN) are popular for SHM in civil engineering for a variety of situations (Hoult, et al., 2009; Bennett, et al., 2010; Huang, et al., 2013; He, et al., 2014). As the technology evolving, MEMS sensors become more integrated and less power-consuming which is ideal for its combination with WSN. In addition to the WSN, the Light-Emitting-Diode (LED) has been used for the risk pre-warning systems. Different signals with colors represent different levels of risk (Akutagawa, et al., 2011; Liu, et al., 2011).

The timing for repair measures on disrupted geo-structures after risk event occurs is often vague to engineers, i.e., either too late or too early but without a full estimation of the effect of these measures. It thus may result in a high cost but small effect on the performance recovery (Chang, et al., 2001; Ni and Cheng, 2012). In view of this, engineers need to understand the recovery ability of the geo-structures in terms of the degree and rapidity before a specific repair measure is implemented. Few studies have been carried out on the performance recovery of geo-structures at this moment. But even when the measured data are available, the understanding of recovery ability is not straightforward due to the absence of a rational model to assess the recovery efficiency. To be more specific, there are no criteria to guide engineers to evaluate the effectiveness of repair measures and their associated time cost (Doherty, et al., 2012; Titi and Biondini, 2013).

In view of this condition, the aim of this paper is to present an integrated system to sense the risk in real time, monitor the risk dynamically and decide the repair strategy rationally after the occurrence of risk event. It should be noted that part of this work, i.e., resilience analysis, has recently been presented at the International Conference on Smart Infrastructure and Construction (Zhang, et al., 2016). Presented here in this paper is an integral system including risk assessment, evaluation, monitoring, visualization and control. To reach this goal, a brief introduction of current practice of risk management and the associated limitations are presented. Then, a smart system for sensing the risk of geo-structures by using the WSN is proposed. The MEMS-based tilter sensor, crack sensor and seepage sensor have been compiled in this WSN system. Based on the wireless network, the risk can be sensed by those sensors in real time. With these real-time monitoring data, the risk could be transformed into the recognizable colors by using the LED system. However, there is no risk-free geo-structures that can be constructed. Hence, this paper also presents a resilience-based analysis model for the repair strategy after a risk event occurs. The effectiveness of WSN system on the resilient ability of geo-structures has been discussed. Finally, this integrated risk sensing system for geo-structures has been applied into two real tunnel case in Shanghai. One is the Shanghai metro tunnel, which was disrupted by an extreme surface surcharge. The other is an underwater road tunnel in a normal condition through Shanghai Huangpu River. The application cases validate that this risk sensing system is quite helpful to the dynamic risk control both for the decision makers and the on-site workers. 2 RISK MANAGEMENT IN CURRENT PRACTICE

A systematic risk management procedure includes risk delineation, risk identification, risk estimation, risk evaluation and risk control, each carried out in sequence. It is clear from the definition of risk management that the procedure is dynamic during the lifetime of geo-structures. However, currently, risk management is carried out largely based on single stage that is not sustainable. The safety of geo-structures contains large uncertainty since potential high risk might

be ignored due to the independent management at different stages. To ensure the sustainability of the management, quantitative and dynamic procedure of risk identification, estimation and evaluation is needed, which will be elaborated as below: 2.1 Quantitative Risk Assessment (QRA)

Quantitative risk assessment is a method of quantifying the degree of risk through a systematic examination of the hazard that threatens the geo-structural safety. In general, risk is quantitatively estimated in terms of the annual exceedance rate of a particular consequence level (Ayyub 2014). For small annual rates, annual probabilities can be used instead as an estimates of these rates. Additionally, it is often to approximate risk by the multiplication of the annual probability of the occurrence of the hazard and the respective consequence due to the occurrence hazard event, and is expressed as follows:

R = P(A) x C (A) (1)

where A stands for the specific hazard or risk event, P(A) is the probability of the occurrence of event A and C(A) is the corresponding consequence. To be more specific, the consequence could be sub-divided into the degree of system performance loss, i.e., vulnerability V and its corresponding cost E. Eq. 1 can be expressed in detail as below:

R( A) = P( A) xV ( A) x E

Vulnerability V(A) here is defined as a function of the hazard intensity (I) associated with exposed elements at risk and the resistance ability (RE) of the elements to withstand a threat (Uzielli, et al., 2008). It can be mathematically expressed by Eq. 3 (Li, et al., 2010).

V = f (I, RE)--

1.0 -1.0

2(RE -1)2 RE2

-£ 0.5

0.5 < — £ 1.0 RE

-> 1.0

The system vulnerability varies with the intensity and resistance non-linearly. The characterization of hazard intensity I and the system resistance R could be different from case to case. The probability P(A) could be obtained from the reliability analysis under a specific limit function determined by the failure criteria of risk event A. Then, the only thing left for quantitative evaluation is the consequence or economic cost E. It should be realized that the consequence depends on the exposure place and the exposure time to the risk event. Besides the vulnerability and the cost of the loss, the time and space dependency should be included in a detailed quantitative consequence analysis. Eq. 2 is thus revised by a refined equation below,

Re ( A) = P( A) x P (r|A)x £ ( P ( S\A)xV ( A|S )x E )

RH ( A) = P ( A) x P (T\A )x X ( P ( S\A )x F ( A|S ))

20 21 22 23

where Eq. 4a is referred to the economic loss and Eq. 4b is referred to the human loss. P(T\ A) is the conditional probability of the hazard A happened in the time interval T, and P(S \ A) is the conditional probability of the hazard A happened in the space area S. E stands for the value of the economic loss. The above complex analysis of the consequence in terms of the summarization of all the conditions can be visually explained by the event tree, as shown in Fig. 1 (Li, et al., 2014). The PR represents the probability of occurrence of risk event. The T stands for the different time intervals when risk event occurs. The P(n) means the probability of n workers on site for the time interval T. The p(i) stands for probability of scenario i occurs. The expectation of the consequence of the events in last column of the tree is essentially expressed by Eq. 4 mentioned above.

Whether it occures? Pr

Scenarios and

their probabilities

Casualty index Vfis recommended vulnerability

value of death and missing Vs is recommended vulnerability

value of light injuries Vl is recommended vulnerability value of serious injuries

n foo. P(D / p(2)

P(m) P(3)

p(m+2) \ P(C+2>

- Ox Vf+ 0x Vs +mx F

- Ox Vf+lx Vs + (m-1) xVl

- 0xVf +2x Vs + (m-2) X v

- mx V+ Ox Vs + Ox V

Fig. 1 An example of event tree analysis (ETA) in QRA for shallow tunnels 2.2 Dynamic Risk Assessment (DRA)

As mentioned in Eq. 4, risk is regarded to be closely related to the time when the hazard happens. Hence, it should be a dynamic process for a detailed risk assessment in geotechnical engineering. Monitoring data directly indicate the safety and health state of structures for risk early-warning strategies. Hence, the data-based dynamic risk assessment should be most helpful to capture the real state of the geo-structures. The monitoring data based DRA consists of three major parts, including project monitoring, design of the risk warning index and subsequent dynamic risk assessment. The risk warning index is determined by the design requirement for the interested performance and the risk correction factor. The former one is calculated through the mechanical analysis under the dynamic construction conditions and the latter one is obtained by

10 11 12

20 21 22

analyzing the corresponding performance of the structure apart from mechanical perspective. The flowchart for monitoring data based DRA is shown as Fig. 2.

Project information

Construction features

Project parameters

Risk accidents

Design of monitor value

Design Phase

Moniter Construction

parameter conditions Loads Risk property Loss types

Coefficient of risk correction

Monitoring data

Risk warning index

Evaluation criterion

Dynamic risk assessment

Construction Phase

Dynamic Assessment Phase

Fig. 2 Flowchart of the monitoring data based dynamic risk assessment 2.3 Current limitation in risk management

It has to be noted that the above procedure for quantitative risk management mainly focuses on the assessment which includes the identification, estimation and evaluation. However, the risk control after the comprehensive and sophisticated assessments has not been paid much attention on. As for the on-site workers, the numerical number or discrete level of risk R(A) from the assessment can be hardly understood. On the other hand, for traditional monitoring systems, the measured data for a severe disruption accident are generally too late to be recognized and reported. Therefore risks cannot be managed with traditional monitoring approaches. In a word, there is an urgent necessity to develop a smart system incorporating the abilities of fast risk sensing, efficient risk assessing and in-situ visualization. By doing so, the risk of geo-structures can be monitored in real-time and the workers can be informed simultaneously by the visualized risk signals.


An integrated WSN system includes the sensor elements, wireless network and the transformation model from sensor signal to the performance indicator for geo-structures. In this paper, a systematic wireless sensoring network integrated by the authors will be introduced in the three major parts as below. It should be noted that the WSN system introduced here is mainly applied for segmental lining structure as a kind of typical geo-structures encountered these days in urban area. Detailed description of the technical specifications for all the sensors discussed in this paper, e.g., data-sheet or manuals of all sensors, could be referred to our webpage (WISEN, 2016).

1 3.1 MEMS-based Sensors

2 3.1.1 Tilt sensor

3 In order to measure the deformational performance of shield tunnel, TJ-UWIS, i.e., a series of

4 MEMS inclinometers, were developed. The chip selection, schematic design, Printed Circuit

5 Board (PCB) layout and parameter optimization in the MEMS inclinometers were all invented

6 with independent proprietary. For various measurement objectives, three different specifications

7 of TJ-UWIS sensors are proposed, namely dual axis analog inclinometer (V1.1), single axis dif-

8 ferential inclinometer (V1.2), and single axis analog inclinometer (V1.3). A photo of the dual

9 axis analog inclinometer is shown in Fig. 3 with a comparison of commercial inclinometer in

10 size. The TJ-UWIS is compatible with our own wireless smart sensor network (details are

11 shown later). The MEMS based inclinometer has the functionality that the inclination of the

12 sensor could be measured because of the change of the direction of sensors' gravity. In order to

13 validate the MEMS sensor's adaptability to various environments, temperature drift test, power

14 consumption test, and calibration are conducted. All the sensors are tested under the same con-

15 dition. Due to the page limitation, details of the performance tests are not provided here, which

16 could be referred to the paper by Huang et al. (2013) for more information. Only the perfor-

17 mance indexes of TJ-UWIS from the various performance tests are compared and shown in Ta-

18 ble 1.

20 Table 1 Performance index for different version of invented tilt sensors (Huang, et al., 2013)

Type Axis Preci-sion(°) ^ Resolu-tion(°) Range(°) Input volt-age(Vo) Output Volt- age(Vo) Working tempera-ture(°C)

TJ UWIS1.1 2 0.01 0.0025 90~90 12 0~5 -40~125

TJ UWIS1.2 1 0.01 0.0025 30~30 12 0~5 -40~125

TJ UWIS1.3 1 0.01 0.0013 30~30 12 0~5 -40~125

Fig. 3 TJ-UWIS prototype (left) and commercial inclinometer (right)

10 11 12 13

20 21 22

3.1.2 Crack sensor

In order to detect the concrete cracks in shield tunnel, the crack sensor element is also developed. It is essentially developed based on the traditional displacement transducer coupled with the wireless signal module. In general, the frame of sensor mainly consists of power module, displacement transducer, microcontroller, AD convertor and temperature sensor (see in Fig. 4). LPDT or LVDT which has small range and high precision, is selected for displacement transducer. Temperature compensation is considered by temperature sensor. The microcontroller is mainly composed of two parts, i.e., one is a wireless transceiver (based on Zigbee protocol), and the other is an industry-standard enhanced micro control unit. It has various operating modes, including the work pattern of ultralow power consumption, making it consume lower power but with a high data rate. Figure 4b has shown a photo for the prototype of this crack sensors. In addition, the stability of data acquisition also has been verified by conducting an indoor stability tests which could be referred to the work done by Wang and Huang (2013) for details.

Temperature sensor


Displacement transducer


AD convertor Microcontroller

ADS1115 CC2530

High-precision voltage

IB 1205

(a) layout of crack sensor (b) photo of the prototype of the crack sensor

Fig. 4 Wireless crack sensor

3.1.3 Seepage sensor

The seepage sensor was developed with the principle of electrical conductivity or temperature gradient changes due to the seeped water on the concrete segments. Since the stability of the sensors play a much important role during the working condition, the seepage sensor here is developed based on the electrical conductivity method. Figure 5a shows a prototype photo for this seepage sensor, which consists of electrodes (see in Fig. 5b), ammeter, microcontroller, AD convertor and power module (see in Fig. 5c). The dry concrete is almost nonconductive. However its conductivity increases rapidly and resistance reduces greatly when concrete is soaked by water. Based on this characteristic, the tunnel leakage can be detected. The length (for one dimensional) or the area (for two dimensional) of soaked concrete could be read off from the current-time curves if the relationship between the current and the flow-rate of soaked concrete has been calibrated. Hence, a series of indoor calibration tests have been conducted to characterize the relationship between current and flow, as shown in Fig. 6. In other words, if the sensor cap-

tures the current signal, it could be transformed into a stable flow level using the current-flow curve shown in Fig. 6.

(a) (c)

Fig. 5 Functionality of electrical conductivity based seepage sensor: a) General frame of seepage sensor; b) Electricity conductivity sensor node; and c) Controller

Current I (mA)

Fig. 6 Calibration of current-flow curve from indoor test 3.2 Sensoring Network

The frame of wireless sensoring network consists of three basic parts: structure of network, communication protocols of network, and wireless transmission and compression methods for measured data. In terms of the network structure, wireless sensoring network adopts an expansible two-layered network topology. The network communication protocols is composed of the ZigBee and WiFi communication protocols. The WSN structure, based on wide area wireless network, wireless LAN and wireless area network, is constructed, mainly including Zigbee-3G, Zigbee-WiFi, Zigbee-Bluetooth. With an advanced methods of wireless transmission and compression for huge amount of data (He, et al., 2014), the capacity of the transmitted data is in-

creased. Furthermore, it can reduce energy consumption and extend the network life. Figure 7 shows a layout of the network in a typical tunnel associated with the sensor nodes deployed.

# Sensor Node A Gateway

__!___—J-——|— ' —I— _ ' 4— '

• ••••• ••••••

Fig. 7 WSN deployment in a shield tunnel 3.3 Performance Transformation Model

A simplified model is developed for calculating the horizontal convergence with the assumption of a rigid concrete segment, that is, tunnel deformation mainly depends on the joints rotation. Figure 8 is the schematic for this deformation mechanism. The horizontal convergence of tunnel can be calculated by Eq. 4:

AD = L(AÔrAÔ2)cosa

where AD is the change of horizontal convergence, L is the distance from the base point to inner surface of segment at the tunnel center point level, A9X is change of the tilt sensor on segment Bl, A 62 is change of the tilt sensor on segment B2, a is the angle shown in the Figure 8.

Fig. 8 Transformation model for horizontal convergence of shield tunnel The above transmission function is purely based on the assumption of ignoring the deflection or distortion of concrete segment per se. Hence, the calculation bias by applying this assumption is also analyzed. The horizontal convergence is calculated under the same load conditions by

two methods, i.e., one is from Eq. 4 and the other is calculated by the model proposed by Lee et al. (2001). Then, the error is defined as the ratio of the convergence from Eq. 4 over the results from Lee's model. Figure 9 plots the calculated error along different position angle at the tunnel perimeter for different joint stiffness. The joint stiffness is characterized by a coefficient X defined as a ratio of joint rotational stiffness kg over the lining bending stiffness EI. Result indicates that the best installation point for tilt sensor is 120° (setting the crown of tunnel as 0°) for different level of relative joint stiffness coefficient.

70 80 90 100 110 120 130 140 Angle of position with vertical axis (0 )

Fig. 9 Calculation error at different measurement location 4 VISUALIZED RISK PRE-WARNING

Currently, the traditional procedure of the risk pre-warning is that 1) firstly, the monitoring data are collected manually on site; 2) then the collected data is back analyzed offsite and the risk is assessed based on these data; and 3) finally, the risk pre-warning is sent out if the result of analysis is beyond the design criteria. Quite often, the time cost for this procedure is so significant that usually loses the merit of the "pre-" warning. The undefined measurement frequency could lead to the lack of adequate detection of anomalies and trends, accidents, higher costs for geo-structures (ITA, 2014). In view of this circumstance, a real-time risk pre-warning system for geotechnical construction should be necessary to retain the feature of the response speed. In other words, the real time pre-warning system could make the risk visualized. It has to be noted that the visualized risk pre-warning system is a safety system to protect on-site workers (not the asset per se) as opposed to a threshold checking system (Webb, et al., 2015) to assess infrastructure condition.

4.1 Principle of Light Emitting Diode (LED) Based Visualization

The first visualization technique is developed based on the Light Emitting Diode (LED). The signal to capture the structural performance, the risk assessment based on the captured performances and the risk transformation from the assessed level of the risk to the visualized optical signal are all compiled in a microprocessor using the internal program. Finally the risk level of the construction could be reflected directly by the change of the colors of the LEDs on site.

The whole process of risk visualization is controlled automatically by the computer that enables the risk pre-warning system to be rational, real-time and visible. An example of difference geo-structural behavior with the corresponding color of LED is shown in Table 2, including the tunnel convergence and the strut axial forces in deep excavation, respectively. Table 2 Risk levels and color of LED for geo-structure behaviors (DGJ08-10, 2010)

Risk Level Safe Warning Danger

Color of LEDs Green Yellow Red

Tunnel convergence to diameter ratio AD/D (%%) 0-5 5-10 >10

Strut axial forces for deep excavation (kN) 0-4500 4500-5500 >5500

Different kinds of sensors could be integrated in this LEDs aided visualization system. The specific choice of the sensors depends on the type of structural performance that the engineers are interested in, e.g., the WSN based real-time monitoring system. It is until the threshold for each level of risk has been set that the system is activated to work. Once the measured data exceed the pre-set threshold, the system will then change the corresponding LEDs color and flash the LEDs to make an on-site warning automatically. For some important geo-projects, the wireless transmission technology is used to connect the microprocessors and the remote output terminal. In this way, the remote risk pre-warning is achieved besides the on-site risk pre-warning. The memory chips can store the real time measured data for later check and analysis. An example of the whole module of this system is illustrated in Fig. 10, which is installed for the pre-warning of excessive displacement of a spring.

Fig. 10 Schematic of LEDs aided risk visualization system: a) LEDs system; b) sensor; c) circuit board; and

d) LED

4.2 LED Visualization on Tunnels

Behaviors of tunnel lining structures, e.g., displacement, convergence, stress and settlement, could be transformed into electronic signals. All these signals will be analyzed and output by the microprocessor. When the convergence of tunnel lining is interested, specific risk levels of convergence could be linked to corresponding colors of LEDs. The LED could receive those signals from microprocessor to show the right color for visualization (see in Table 2). Figure 11

has shown a full-scaled in-situ test on Shanghai metro tunnel by using the LED based risk pre-warning system presented in this paper.

(a) Green Level (Safe)

(b) Yellow Level (Warning)

(c) Red Level (Danger) Fig. 11 In-situ Test on the LED based risk pre-warning system for metro tunnels

4.3 LED Visualization on Deep Excavations

By using a similar technique described above, Fig. 12 shows a full-scaled in-situ test of the system on deep excavation in Shanghai with the risk pre-warning level shown in Table 2. It indicates that the monitoring and risk pre-warning by this LEDs aided risk visualization system is also reasonable and feasible for the deep excavation system.

(a) Green level

10 11 12

20 21 22

(b) Yellow level

(c) Red Level

Fig. 12 Application of risk visualization system into a deep excavation in Shanghai 5 RESILIENT GEOSTRUCTURES

It has to be realized that there are no risk free projects in geotechnical engineering. No risk sensing or risk visualization system can eliminate all risk. If a disaster occurs to the geo-structure, rapidity of repair works to an acceptable level for the basic functionality and the corresponding cost and time consuming are probably the most crucial factors in decision-making. This is in particular critical for the geo-structure in operation, such as the case that disruption of a running metro tunnel will stop operation of the metro system. However, applying the above risk assessment, risk sensoring and visualization are not sufficient to support the decision making. A rational analytical-based design of repair strategy with optimal effectiveness of repair is thus of great necessity. Hence, a resilience-based analysis model is elaborated here for the use of establishing a reasonable strategy for repair. 5.1 Resilience Model

In a life-cycle context, geo-structures may be subjected to unexpected natural and human-related hazards, which can harm serviceability or even safety of infrastructure systems. In view of this circumstance, the resilience concept is gaining more and more attentions for research on disaster relief. The resilience is defined as the ability of a system to absorb the disruption caused by hazards and the ability to recover to an acceptable performance level. Figure 13 has illustrated the schematic for resilience analysis. It is based on the general model proposed by Ayyub (2014) but has taken the decision making stage into account in overall time period of risk event.

High-impact low-chance hazard

.L . j

Recovered State

In iti al Q0

tf ts tr

Fig. 13 Schematic of performance transition in resilience analysis Due to its major status within disrupted period, a resilient geo-structure consists of three stages, namely disruption stage, decision-making (evolution) stage and recovery stage, as shown in Fig. 13. At time ti, some hazard starts to act on the structure and its performance decreases until time tf when significant disruption has been detected. From time tf to time ts, it is the decision-making period for emergency relief and selection of recovery measures. When the repair works are applied at time ts, the structure performance recovers and reaches an acceptable level at time tr. Based on this framework for resilience analysis, the metric of resilience is denoted as the index Re. The Re can be calculated by the following equation:

AP (f - L ) + As (ts - tf ) + Ar (tr - ts )

tr - tt

12 where the weight coefficients, i.e., AF, AS and AR, can be further expressed as follows:

['ff (t)dt


I ss(t)dt Jtf_

|''Q(t )dt

r(t)dt _

I Q(t)dt

Essentially, the three weight coefficients stand for the area ratios determined by the three curves of performance transition in Fig. 13, i.e., disruption curve f(t), evolution curve s(t) and recovery curve r(t), respectively. All these curves can differ from case to case due to different

1 designs and different recovery measures. Several dimensions can be covered in the above metric

2 of the resilience, including degradation, robustness, vulnerability, rapidity and recovery. Details

3 are summarized in Table 3.

4 Table 3. Resilience dimension and its property



Degradation Robustness Vulnerability Rapidity Recovery

fi fd frfrfd AT=tr-ts

Degraded performance Q(t) at t, Residual performance at tf Performance loss at tf Speed of recovery Recovered performance

6 5.2 Effect of Smart Sensoring on Resilience

7 The rapidity is a crucial dimension in assessing the structural resilience. The smart sensoring

8 could increase the response time of disruption of geo-structures and further increase the rapidity

9 of recovery. If there is an ideal geo-structure that its performance do not degrade with time t as

10 shown in Fig. 14. In other words, the performance Q is always equal to unit. Once the structure

11 is unfortunately disrupted by extreme hazard at time t, the performance has been reduced to fd

12 (/d<1) through a period of time ATi. After implementing the repair measures, the performance

13 has been recovered to normal state (i.e., Q=1) through a period time of AT2. By applying the

14 above mentioned resilience metric, the calculated resilience index Re1 could be expressed as be-

15 low:

1 + fd

20 21 22 23

tif tf!

Smart monitoring TJaditional monitoring n: Rel ative r esp onse ti me

tr1 time

Figure 14 Difference of performance curves between smart sensoring and traditional monitoring In this benchmark problem, if the smart sensoring technique could be used before the disruption happens, the reduction of performance could be captured once it is being reduced. Thus, suppose the time period for geo-structure response time in this case of applying smart sensoring is n times faster than the time traditional monitoring needs, i.e., equal to AT1/n, as shown in blue arrow line in Fig. 14. For example, if the monitoring frequency for a smart sensoring is once per

day, while the frequency for traditional monitoring is once per month. Then, the parameter n should be equal to 30 in this case. By applying the same repair measures, the recovery duration could also be n times faster than the original time AT2 on the basis of geometric laws. In other words, by applying the same repair measures, the rapidity of recovery by using smart sensoring could be n times faster than the traditional monitoring system. It could be derived further that the area of performance loss in Fig. 14 for the smart sensoring case could be n2 times smaller than that for the traditional sensoring case. The resilience index Re2 for the smart sensoring case could be generally n2 times larger than the Re1 for traditional sensoring case and is represented as below:

Re2 = 1 "A (1 - f) (7b)

Relative response time (n)

Fig. 15 Effect of rapidity of response time by using smart sensoring on the geostructural resilience Given the robustness performance fd after the disruption at the level of 0.8, 0.5, 0.3 and 0.1, by applying Eq. 7b, the calculated index Re2 for smart sensoring case could be plotted against the relative response time coefficient n in Fig. 15. It is clear that the coefficient n could greatly affect the results of Re2. If n is larger than 5, the resilience could be incredibly high and almost equal to unit. That is to say, the geo-structure performance could be strongly resilient, regardless of the vulnerability under disruption. This is the reason that the smart sensoring could improve the geostructural resilience even with the same repair or rehabilitation techniques.

Table 4 Comparison of resilience between smart and traditional sensoring in this benchmark problem

Dimension Effect of smart sensoring

Response duration 1/n

Vulnerability 1/n

Robustness A - B/n

Rapidity 1/n

Resilience C - D/n2

Resilience loss 1/n2

Note: Parameter A, B, C, and D is constants when the case is specific and could be calculated by Eq. 6.

Table 4 has summarized the overall effects of smart sensoring on geo-structure resilience expressed by using relative response time coefficient n. Since the loss of resilience is related to se-

1 cond order of n, i.e., Q(n~). it thus could clearly indicate the great effect of smart sensoring on

2 the structural resilience.


4 6.1 Case 1: Metro tunnel in Shanghai

5 Because of extreme surcharge loading on the ground surface above the metro tunnel, several se-

6 rious defects including large convergence, crack, and seepage in the metro shield tunnel lining

7 were observed. In order to ensure the safety of tunnel operation, the convergence performance is

8 recovered by soil grouting at two sides of the tunnel lining. WSN is employed for the behavior

9 study of shield tunnel structure during the soil grouting. Applications of convergence monitorio ing by MEMS Tilt sensors are illustrated here. Details of this case could be found in the paper

11 (Huang, etal., 2016).

12 6.1.1 Applying WSN during the recovery

13 Two monitoring sections, i.e., ring No. 411 and No. 433, were selected to measure the conver-

14 gence development during the repair work for soil grouting. The segmental lining for this metro

15 tunnel has an outer diameter D of 6.2 m and a wall thickness of 0.35 m, as shown in Figure 8.

16 The tilt sensor was installed on the inner surface of lining LI, Bl, L2 and B2. Total of four

17 MEMS tilt sensors were installed within one ring. In the meanwhile, one laser distance meter

18 sensor was installed on B1 at the same level of tunnel center. Target of laser distance meter was

19 installed on B2 at the same level.

20 Tilt sensors measured the rotation of segmental lining during the soil grouting. Figure 16

21 shows the tilt data of Ring 433 on 18th June, 2014. Figure 17 shows the tilt change and its direc-

22 tion of segmental lining after one night soil grouting. It indicates that the horizontal conver-

23 gence become smaller during the grouting.

-2.24 -2.25 -2.26 -2.27 -2.28 -2.29 -2.3 -2.31

-1.98 -1.99 -2

-2.01 -2.02 -2.03 -2.04 -2.05

23:00:00 0:00:00 1:00:00 2:00:00 3:00:00 4:00:00

£ 0.72 0.71 0.38 ID0 A 0.37 ■tluaiiMu *

G O 0.7 r^^ Tn/flft 036

a e 0.69 0.35

"o c 0.68 0.67 \ 0.34 0,3

0.66 -1-'-'-'-L 0.32

23:00:00 0:00:00 1:00:00 2:00:00 3:00:00 4:00:00

23:00:00 0:00:00 1:00:00 2:00:00 3:00:00 4:00:00 23:00:00 0:00:00 1:00:00 2:00:00 3:00:00 4:00:00

time time

Fig 16. Tilt data of Ring 433 on 18th June

10 11 12

20 21 22

Fig 17. Schematic of segment rotation of Ring 433

The calculated change of horizontal convergence of Ring 433 due to one night grouting is -3.057 mm on 18th June. Laser distance meter and total station were conducted to verify the results. The changes of horizontal convergence measured by laser distance meter and total station are -3 mm and -3.1 mm, respectively. For Ring 411, the calculated change of horizontal convergence is -2.735mm on 18th June, and the measured data by laser and total station is -2 mm and -3.7 mm, respectively. Hence, it is clearly verified that the WSN system works quite well compared to those traditional monitoring system. But the real-time sensing of the tunnel performance could be archived via the WSN system rather than those laser distance meter or total station by man-power. In addition, the cost for monitoring by traditional monitoring technique using man-power nowadays is becoming more and more expansive. However, by using the WSN system, this cost will be much smaller than those for traditional technique.

6.1.2 Applying the resilience analysis

The proposed resilience analysis model has been applied to the same accident for Shanghai metro tunnel. From the case study, the effect of real-time monitoring on improving the tunnel resilience can be explicitly demonstrated from a mathematical point of view. Due to page limit, details also could be found in the paper (Huang and Zhang, 2016). It should be noted that the WSN system was applied for the disrupted tunnel only during the recovery (soil grouting) stage as mention previously. Hence, the integrated performance transition curves along with the disruption and decision making stages are estimated by the traditional monitoring system (e.g., total station). The deformational performance is discussed as the serviceability performance indicator in this paper, which can be transformed by the measured tunnel convergence using the following equation:

Q(t )■

DDo DD(t )

where Q(t) is the performance indicator, AD0 is the initial tunnel horizontal convergence and AD(t) is the convergence at time t. Figure 18 has illustrated the integrated convergence perfor-

1 mance transition of lining ring No. 433 due to the extreme large surcharge loading. A best-fitted

2 performance transition curve is obtained by using the measured data with a coefficient of deter-

3 mination (.R2) for this curve equal to 0.89 (sample size n equal to 101). Almost six years has

4 been passed since the occurrence of the disruption until the complete of the recovery. The delay

5 of the reaction has resulted in a small resilience index Re at 0.34 calculated by using Eqs. 5 and

6 6. It means that 66% of the total performance has been lost because of the extreme surcharge

7 and also because of the slow reaction. The stop of monitoring is mainly because the lining is ob-

8 served in a disrupted but stable status. The lining is regarded by maintenance engineers to be

9 safe at that time resulting in the stop of monitoring. However, almost three years later engineers

10 want to enhance the serviceability in deformation of tunnel lining. That is the reason for a grout-

11 ing treatment after almost six years, which definitely lost the rapidity of recovery.

Extreme surcharge

o Field data

Best-fitted performance curve - —— Normal degraded performance -----Constant performance

>(»00 №03




Stop period




1.760 I 2.200



13 Fig. 18 Measured performance transition for tunnel convergence

14 If the tunnels has been instrumented by the WSN system before the disruption. In that case,

15 the reduction of performance due to extreme surface loading could be captured once it starts to

16 occur. Suppose the tunnel disruption can be detected within 80 days after the surcharge loading

17 on the ground, the tunnel would have only 11% loss of the performance and could be recovered

18 fully and quickly by the soil grouting. In this artificial case, the calculated resilience index Re is

19 0.94, significantly larger than the previous value at 0.34 for the real case. This comparison is

20 visually explained by Fig. 19. Hence, a timely detection of the disruption by applying the real-

21 time monitoring, could reduce the vulnerability at the beginning of the disruption and subse-

22 quently increase the resilience with much lower time cost.

10 11 12

1.20 1.00 0.80 0.60 0.40 0.20 0.00

Extreme surcharge

-Ring No. 388 measured performance

Re=0.34 - Artificial Case II


Time (day)

Fig. 19 Comparison of the resilience between real case and artificial case 6.2 Case 2: Under water road tunnel in Shanghai

MEMS tilt sensors are installed in 2 sections of the Dalian road shield tunnel under the Huang-pu River in Shanghai with an 11m outer diameter and a 0.48m segment thickness. The tunnel ring consists of 8 concrete segments, shown in Figure 20. Six tilt sensors are installed because of the limit of working space on site. Figure 21 shows the total change of tunnel convergence for nearly one year in 2015. It is nearly about 2 mm increase during the measurement in this year. It is interesting from Fig. 21 that there is a relative large increment of tunnel convergence after June of 2015. It might be reasonably attributed to the fact that during the summer season the water level of Huangpu River had raised causing an increment of convergence induced by additional water pressure.

Figure 20. Cross section and tilt sensor installation of road tunnel

o-0 5 O

2 Figure 21. Convergence change of road tunnel during a year


4 By illustrating the current practice for risk management, the limitation of paying considerable

5 attention on risk assessment while quite little on risk control has been emphasized in this paper.

6 In view of this circumstance, the paper has proposed an integrated risk sensing system incorpo-

7 rating the wireless sensoring network (WSN) for real time monitoring, light emitting diode

8 (LED) technique for on-site risk visualization and resilience analysis model for repair strategy

9 after risk occurs.

10 A "smart" wireless sensing system coupling the MEMS-based tilt sensor, seepage sensor, and

11 crack sensor with Zigbee-based Topological network is developed. The best installation point

12 for tunnel convergence measurement at 120 degree from the tunnel crown is recommended for

13 Shanghai metro tunnel. With the help of these wireless sensors, the risk could be transformed

14 into visualized signals indicated by the LED system, given that the risk level of the pre-warning

15 is set before the risk event occurs. Once the disaster occurs, a resilience analysis model, includ-

16 ing both the vulnerability and recovery stages, is proposed to design the repair strategy by using

17 the real-time monitoring data from WSN system.

18 This integrated risk sensing system has been validated by two real case studies in Shanghai,

19 including the metro tunnel and under water road tunnel. The real-time effect of soil grouting at

20 two sides of tunnels on horizontal convergence of segmental linings has been captured by using

21 this system. The measured convergences were comparable to the data from traditional monitor-

22 ing system with a bias less than 2%. But the real time feature of this proposed WSN system is

23 appreciated. In addition, the resilient ability of this tunnel that was disrupted by surface extreme

24 surcharge was found to be very weak. It is mainly caused by two facts that: 1) the deformation

25 under extreme surcharge is so significant that the robustness of deformational performance has

26 been lost significantly; 2) the repair of soil grouting was applied four years after the occurrence

27 of extreme surcharge which lost the rapidity of recovery.


29 A Risk event

30 R/R(A) Risk of event A in construction and operation of geo-structures

1 P(A) Probability of the occurrence of risk event

2 P(T\A) Conditional probability of the hazard A happened in the time interval T

3 P(S \ A) Conditional probability of the hazard A happened in the space area S

4 C(A) The consequence of occurrence of event A

5 V(A) Vulnerability of event A

6 I Intensity of risk event A

7 RE Resistance ability of geo-structure

8 E Cost of risk event A

9 Vo Voltage in the wireless sensor

10 AD Change of lining horizontal convergence

11 D Tunnel outer diameter

12 L Distance from the base point to inner surface at tunnel center

13 AO, Change of tilt sensor on segment B,

14 a Change of angle shown in Fig. 8b of this paper

15 À Relative joint stiffness to the segmental lining

16 kg Rotational stiffness for segmental joint

17 EI Bending stiffness of segmental lining

18 Re Resilience index of geo-structure

19 ti Time when risk event starts to act on the structure

20 tA Time when significant disruption has been detected

21 ts Time when disrupted structure starts to be recovered

22 tr Time when disrupted structure has been recovered

23 Q(t) Performance of geo-structure along time t

24 At) Disruption curve of performance

25 s(t) Evolution curve of performance

26 r(t) Recovery curve of performance

27 Af Weight coefficient of disruption stage

28 AS Weight coefficient of evolution stage

29 AR Weight coefficient of recovery stage

30 A Degraded performance Q at t,

31 A* Residual performance Q at tf

32 Ai Performance loss at tf

33 AT Speed of recovery (tr - ts)

34 Ar Recovered performance

35 n Time reduction factor of applying smart sensoring networks compared to tradition

36 al monitoring system


2 This work was supported by the National Natural Science Foundation of China (51278381,

3 51538009, 51608380), the International Research Cooperation Project of Shanghai Science and

4 Technology Committee (15220721600) and the Peak Discipline Construction on Civil Engi-

5 neering of Shanghai Project.


7 Akutagawa, Nomura, M., Yamada, H., Abe, R., Izumi, C., and Kusui, A. "On site visualization

8 as a new scheme for monitoring and visualization of safety conditions in geotechnical

9 engineering projects." Proc., International Symposium on Advances in Ground Technology

10 and Geo-Information, 365-372.

11 Ang, A. H.-S., and Tang, W. H. (1975). Probability concepts in engineering planning and

12 design, Wiley.

13 Ayyub, B. M. (2014). "Systems resilience for multihazard environments: definition, metrics,

14 and valuation for decision making." Risk Analysis, 34(2), 340-355.

15 Bennett, P. J., Kobayashi, Y., Soga, K., and Wright, P. (2010). "Wireless sensor network for

16 monitoring transport tunnels." Proc. Inst. Civil Eng.-Geotech. Eng., 163(3), 147-156.

17 Casagrande, A. (1965). "Role of the" calculated risk" in earthwork and foundation engineering-

18 The second Terzaghi Lecture." Journal of the Soil Mechanics and Foundations Division,

19 91(SM4), 1-40.

20 Chang, C. T., Wang, M. J., and Sun, C. W. (2001). "Repair of displaced shield tunnel of the

21 Taipei rapid transit system." Tunn. Undergr. Space Technol., 16(3), 167-173.

22 DGJ08-10 (2010). "Shanghai foundation design code (DGJ08-11-10)." Shanghai Construction

23 Committee, Shanghai.

24 Doherty, A., Dora, J., and Newsome, C. (2012). "Enhancing resilience in Britain's railway

25 infrastructure." Proc. Inst. Civil Eng.-Civil Eng., 165(6), 20-26.

26 He, B., Li, Y., Huang, H., and Tang, H. (2014). "Spatial-temporal compression and recovery in

27 a wireless sensor network in an underground tunnel environment." Knowledge and

28 Information Systems, 41(2), 449-465.

29 Hoult, N., Bennett, P. J., Stoianov, I., Fidler, P., Maksimovic, C., Middleton, C., Graham, N.,

30 and Soga, K. (2009). "Wireless sensor networks: creating 'smart infrastructure'." Proc. Inst.

31 Civil Eng.-Civil Eng., 162(3), 136-143.

32 Huang, H.-w., and Zhang, D.-m. (2016). "Resilience analysis of shield tunnel lining under

33 extreme surcharge: Characterization and field application." Tunnel. Under. Space. Tech., 51,

34 301-312.

35 Huang, H., Xu, R., and Zhang, W. (2013). "Comparative Performance Test of an Inclinometer

36 Wireless Smart Sensor Prototype for Subway Tunnel." International Journal of Architecture,

37 Engineering and Construction, 2(1), 25-34.

38 Huang, H. W., Shao, H., Zhang, D. M., and Wang, F. (2016). "Deformational responses of

39 operated shield tunnel to extreme surcharge: a case study." Struct. Infrastruct. Eng., 1-16.

1 Huang, H. W., and Zhang, D. M. "Quantitative Geotechnical Risk Management for Tunneling

2 Projects in China." Proc., 5th Int. Symp. Geotech. Safety & Risk (ISGSR) 2015, IOS Press, 613 75.

4 ILO (2003). "Annual report for record of deadly accident for occupational safety in the world."

5 International Labor Organization, Geneva.

6 ITA (2002). "Guidelines for tunneling risk management." International Tunnel Association

7 Working Group No. 2 217-237.

8 ITA (2014). "Guidelines on Monit oring Frequencies in Urban Tunnelling." International

9 Tunnelling and Underground Space Association, Longrine, France.

10 Lacasse, S. (2015). "Hazard, Risk and Reliability in Geotechnical Practice: 55th Rankine

11 Lecture." British Geotechnical Association, London.

12 Langford, J. C., Vlachopoulos, N., and Diederichs, M. S. (2016). "Revisiting support

13 optimization at the Driskos tunnel using a quantitative risk approach." Journal of Rock

14 Mechanics and Geotechnical Engineering, 8(2), 147-163.

15 Latham, M. (1994). "Constructing the Team: Final Report of the Government/ Industry Review

16 of Procurement & Constractual Arrangements in the UK Construction Industry." HSMO,

17 London.

18 Lee, K. M., Hou, X. Y., Ge, X. W., and Tang, Y. (2001). "An analytical solution for a jointed

19 shield-driven tunnel lining." Int. J. Numer. Anal. Methods Geomech., 25(4), 365-390.

20 Li, Z., Huang, H., and Xue, Y. (2014). "Cut-slope versus shallow tunnel: Risk-based decision

21 making framework for alternative selection." Engineering Geology, 176, 11-23.

22 Li, Z., Nadim, F., Huang, H., Uzielli, M., and Lacasse, S. (2010). "Quantitative vulnerability

23 estimation for scenario-based landslide hazards." Landslides, 7(2), 125-134.

24 Liu, Q., Zhang, D. M., and Huang, H. W. "Tunnel squat due to surcharge in soft soil and

25 visualization monitoring conception with LEDS." Proc., International Symposium on

26 Advances in Ground Technology and Geo-Information, 185-195.

27 Ni, J. C., and Cheng, W. C. (2012). "Characterising the failure pattern of a station box of Taipei

28 Rapid Transit System (TRTS) and its rehabilitation." Tunnel. Under. Space. Tech., 32, 26029 272.

30 O'Rourke, T. D. (2010). "Geohazards and large, geographically distributed systems."

31 Geotechnique, 60(7), 505-543.

32 Peck, R. B. (1969). "Advantages and Limitations of the Observational Method in Applied Soil

33 Mechanics." Géotechnique, 19(2), 171-187.

34 Titi, A., and Biondini, F. "Resilience of concrete frame structures under corrosion." Proc.,

35 Safety, Reliability, Risk and Life-Cycle Performance of Structures & Infrastructures, Taylor &

36 Francis Group, London.

37 Uzielli, M., Nadim, F., Lacasse, S., and Kaynia, A. M. (2008). "A conceptual framework for

38 quantitative estimation of physical vulnerability to landslides." Engineering Geology, 102(3),

39 251-256.

1 Vardanega, P. J., and Bolton, M. (2016). "The Design of Geo-Structural Systems." ASCE-ASME

2 Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering, 2(1),

3 0401507.

4 Wang, K., and Huang, H. W. "Smart crack sensor and its application in subway shield tunnel."

5 Proc., Proceedings of the 6th International Conference on Structural Health Monitoring of

6 Intelligent Infrastructure, PS 01-08.

7 Webb, G. T., Vardanega, P. J., and Middleton, C. R. (2015). "Categories of SHM Deployments:

8 Technologies and Capabilities." J. Bridge Eng., 20(11), 04014118.

9 Whitman, R. V. (1984). "Evaluating calculated risk in geotechnical engineering." Journal of

10 Geotechnical Engineering, 110(2), 143-188.

11 Whitman, R. V. (2000). "Organizing and evaluating uncertainty in geotechnical engineering." J.

12 Geotech. Geoenviron. Eng., 126(7), 583-593.

13 WISEN (2016). "Introduction of WSN system for all MEMS based sensors."

14 <http: //www sp?id=119>.

15 Zhang, D. M., Huang, H. W., Hu, Q. F., and Zhang, Y. J. "How smart sensoring improve tunnel

16 resilience: from theoretical model to future application." Proc., Proceedings of the

17 International Conference on Smart Infrastructure and Construction, ICE Publishing, 721-726.