Structural health monitoring system for bridges based on skin-like sensor

Structural health monitoring activities are of primal importance for managing transport infrastructure, however most SHM methodologies are based on point-based sensors that have limitations in terms of their spatial positioning requirements, cost of development and measurement range. This paper describes the progress on the SENSKIN EC project whose objective is to develop a dielectric-elastomer and micro-electronics-based sensor, formed from a large highly extensible capacitance sensing membrane supported by advanced microelectronic circuitry, for monitoring transport infrastructure bridges. Such a sensor could provide spatial measurements of strain in excess of 10%. The actual sensor along with the data acquisition module, the communication module and power electronics are all integrated into a compact unit, the SENSKIN device, which is energy-efficient, requires simple signal processing and it is easy to install over various surface types. In terms of communication, SENSKIN devices interact with each other to form the SENSKIN system; a fully distributed and autonomous wireless sensor network that is able to self-monitor. SENSKIN system utilizes Delay-/Disruption-Tolerant Networking technologies to ensure that the strain measurements will be received by the base station even under extreme conditions where normal communications are disrupted. This paper describes the architecture of the SENSKIN system and the development and testing of the first SENSKIN prototype sensor, the data acquisition system, and the communication system.


Challenge and concept
The project concept is based on the currently limited use of sensors being applied to monitor the stability of civil structures and in particular transport infrastructures. Visual inspections tend to 2 1234567890

BESTInfra2017
IOP Publishing IOP Conf. Series: Materials Science and Engineering 236 (2017) 012100 doi: 10.1088/1757-899X/236/1/012100 overwhelm usual inspections in order to collect information to determine the condition of infrastructures. Inspections nowadays can be considered as rather slow in completion, usually requiring (partial or total) closure of the particular network node or link and/or expose inspectors to dangerous working environments and conditions. Detailed monitoring actions and assessments are (a) particularly slow and expensive, (b) usually based on simplistic and conservative models of structural behaviour, and (c) do not provide a rapid, convenient means of determining structural stability following a major incident or other serious event that could potential limit the structural integrity of the structure [1,2]. On top of this, quality and reliability of the visually extracted information and measurements really depend on the expertise of the inspector (objectiveness factor) and, to add on this, quite often the inspector is unable to gain access to all parts of a structure. State of the art structural health monitoring (SHM) technologies and integrated systems have proved a major role in the management of the transport infrastructure while currently existing and used SHM technologies and methods rely only on the use of dense networks of point sensors to monitor a structure, which is costly (and sometimes not practicable). All at once, conventional sensors also prove to fail at relatively low strains and their communication system is unreliable in extreme service conditions: thus, they do not provide a fool-proof alarm of an imminent structural collapse.

The SENSKIN Project
SENSKIN is an EC co-funded project that operates in the framework of EC-FP7-Transport (MG-8.1a-2014 -Smarter design, construction and maintenance). SENSKIN includes a consortium with all the expertise needed in the lifecycle from research to innovation, as well as real-life end-users that can provide solid feedback on the project results. The list of partners of the consortium has been presented below: Laboratoires Nationaux Europeens De Recherche Routiere (FEHRL) (including members that belong to Highway Departments) [1]. SENSKIN aims to: (a) develop a dielectric-elastomer and micro-electronics-based skin-like sensing solution for the structural monitoring of the transport infrastructure that will offer spatial sensing of reversible (repeated) strains in the range of 0.012% to more than 10%, that requires little power to operate, is easy to install on an irregular surface, is low cost compared to existing sensors, allows simple signal processing and includes the ability of self-monitoring and self-reporting. (b) use the new and emerging technology of Delay Tolerant Network to secure that strain measurements acquired through the 'sensing skin' will reach the base station even under extreme environmental conditions and natural disaster events such as, high winds or an earthquake, where some communication networks could become inoperable. (c) develop a Decision-Support-System for proactive condition-based structural intervention under operating loads and intervention after extreme events. It will be based on an accurate structural assessment based on input from the strain sensors in (a) above and will examine the life-cycle economic, social and environmental implications of the feasible rehabilitation options and the resilience of the infrastructure to future changes in traffic demand that these options offer. (d) Implement the above in the case of bridges and test, refine, evaluate and benchmark the monitoring system (integrated a and b) and package (integrated a, b and c) on actual bridges [2].
The practical application of the developed prototypes will significantly enhance structural health monitoring of bridges but could be extended to monitoring of other critical infrastructures like damns, energy related structures, tunnels, concrete pavements (paved or roadway) etc.

Soft Capacitive Sensor
The design of the soft skin-like sensor used in the SENSKIN system is shown in Figure 1. The sensor is built around a thin silicone elastomer membrane (1) and can thus detect large reversible strains between 0.01% and 10%. The silicone membrane (1) obtained from Wacker (ELASTOSIL® Film 2030 250/100) is only 100µm thin and coated on both sides with compliant electrodes (2). The top and bottom electrodes are made by spraying a toluene-based solution of the silicone material loaded with carbon nanoparticles (Printex XE-2B) onto the membrane. After solvent evaporation and cross-linking of the silicone, the electrode can be stretched over the whole range of sensor motion with a typical resistance of 8 kΩ across its length (15 cm). Polyester frames (3) are employed to provide electrical terminals and for handling and mounting of the sensor. In a last step, the whole assembly is coated with protective silicone layers (4). As the active electroded area is 15x3 cm 2 , the capacitance without pre-stretching is around 1.5 nF. The resistance of the electrodes changes within 0.02-0.07 Ω per µstrain. The measurement principle of the skin-like capacitive sensor is based on continuously monitoring the capacitance changes caused by sensor deformation [1,9,10]. When the sensor is attached to a component of the transport infrastructure, its deformation will cause a corresponding deformation of the sensor and thus also a change in its capacitance ( Figure 2).  The data shown in Figure 2 were obtained with a linear testing machine (Zwick/Roell Z005) for the controlled deformation of the sensor and with a precision LCR meter (HP 4284A) for measuring the sensor capacitance. The capacitance-strain dependence is linear over the whole range of motion with a typical sensitivity of around 1.3 fF per µstrain. When the sensor output is measured with a SENSKIN Data Acquisition (DAQ) module (see section 1.2 below), the capacitance behaviour becomes more complicated (Figure 3). There is some drift in the measured values at higher strains as well as a hysteresis-like behaviour for the overall curve. The nonlinearity is due to the fact that the PCap system inside the DAQ module actually measures RC times during charging or discharging of the soft capacitor. Even if the capacitance changes linearly (Figure 2), the resistance of the electrodes can change nonlinearly over the range of 10% deformation.

Data Acquisition (DAQ) from the monitoring sensor
The DAQ is part of a complex sensor NODE situated between the sensing element based on stretchable elastomeric material and the microcontroller for signal pre-processing and communication control. From a measurement point of view, the sensor is a capacitance that must be monitored very precisely. There are three main challenges for the measurement electronics: a) The sensor being used has a mean capacitance of about 1.5 nF in the prestressed state.
However, the sensitivity of the sensor is as low as only 1.33 fF per µstrain. Therefore, a capacitance of approx. 1.5 nF with a noise-limited resolution of ca 40 fF must be determined for the desired resolution of ca. 30 µstrain at a sampling rate of 50 Hz. b) The compliant electrodes of the stretchable sensors have relatively high internal resistances between 3 and 10 kOhm, and the resistance is also strain-dependent. c) The strain-dependent capacitance is highly sensitive to temperature. Therefore, the data acquisition system contains a temperature-measuring chip with a resolution of 24 bit for calibration purposes. For the capacitance measurement, a Charge to Digital Converter (CDC) PicoCAP02 from ACAM electronic is used. It was developed by ACAM Elektronik and adapted by Teletronic to meet the measuring requirements of the project. The operating principle is illustrated in the equivalent circuit: An internal voltage reference charges the Sensor represented by the Capacitor C across the internal charging resistor R and the compliant-electrode line resistances R1 and R2. The capacitance of the Sensor as well as the two line Resistances depend on both strain and temperature, and the charging process is controlled by the switch S1. The detection process is completed by measuring the discharge time across the discharge resistor controlled by the switch S2. The measuring principle implemented in the application of the PicoCap02 CDC is radiometric, i.e. the capacitances are compared to a fixed reference, and the ratio of the discharge times is directly proportional to the ratio of the capacitances, as the discharge time is determined by the respective capacitance and the selected discharge resistor.
Both circuits, the CDC as well as the temperature measurement chip, communicate via the SPI bus with the microcontroller for averaging and data management. Both have internal processors for controlling the measuring process. Measurement of capacitance and temperature may be done in parallel so that an associated temperature is determined for each strain value. In order to compensate for the temperature-dependent internal resistance, as well as for the temperature coefficient of the capacitance, each sensor is calibrated individually. In practical applications, the EMI influence of any nearby 50 Hz power lines is significant because of the length of the sensor and its small capacitance. Up to now, the measuring process is synchronized with the power-line frequency for supressing the interference. In future developments, the influence may be supressed via higher sampling rates for the capacitance and by the use of a 50 Hz notch filter.

Why Delay-/Disruption-Tolerant Networking technology is necessary?
WSNs (Wireless Sensor Networks) deployed in open environments are prone to intermittent connectivity due to power scheduling, channel fading, node failure, and severe packet losses from unpredictable external factors, such as interference. On top of that, the connection of the Gateway to the Internet might also present its own challenges, as it is not uncommon for bridges to be located at places where communication infrastructure to the Internet is only available intermittently (e.g. remote places where only a satellite link is available) or it is not sufficient to cover monitoring needs due to the employment of low-capacity links (e.g. GPRS links).
In this context, SENSKIN communication system utilizes DTN (Delay-/Disruption-Tolerant Networking) protocols [3,4,5,6,7], which synergistically with mesh networking, address the majority of the aforementioned problems. In particular, DTN communication protocols provide reliable data communication across failure-prone networks by employing various purpose-specific network mechanisms, including store-and-forward data forwarding, multi-path routing and hop-byhop reliability.  Figure 5, constitutes a fully distributed and autonomous WSN that is able to self-monitor and report itself.   At link layer, communication modules adopt a dual interface approach, by employing in particular the IEEE 802.11 and 802.15.4 standards on an as-needed basis. This approach helps to increase communication robustness and data delivery reliability, as well as, to balance out the trade-off between performance and energy consumption. Bundle Protocol (BP) runs on top of them [3]. BP constitutes the basic protocol through which a DTN architecture can be established among the nodes of a network. Through appropriate convergence-layer protocols, BP runs directly over IEEE 802.15.4. The same does not stand true for IEEE 802.11, where a UDP/IP stack is employed to allow BP pass its data over IEEE 802.11. Custom-tailored networking solutions for supporting SENSKIN system operation logic, such as the SENSKIN Operation Logic Protocol (SOLP), run at the Bundle layer. Finally, all SENSKIN communication applications used for sending and receiving data (i.e. solpd, solpSend), along with a panic communication mechanism that is implemented as an extension of the node's management application (i.e. mngd), run at the application layer. Panic communication mechanism is triggered under certain network circumstances (e.g. low signal quality, loss of neighbour nodes, extreme weather conditions). In particular, it allows network self-recovery and in extreme cases, (e.g. major catastrophic event like an earthquake) it enables data from potentially isolated

System Experimental Evaluation
The findings of a review showed that it was necessary to (a) design and fabricate purpose-built test rigs for undertaking calibration tests, and (b) that an off-specimen camera-based system was the only practicable means of calibrating the sensor. The calibration tests are undertaken using uniaxial test rigs (in the main) and a flexural rig (see Figure 9). Both types of rig can accommodate steel and concrete test coupons. The strains applied to the test coupons, and sensors, are measured using an off-specimen camerabased system obtained from iMETRUM. The variation in the output of two sensors with ambient temperature is investigated using the environmental cabinet at TRL. Prior to undertaking the calibration tests it was necessary to undertake proving tests of the equipment and calculation routines to establish a reliable standard reference procedure.
Tests on the prototype sensor showed that the application of a uniaxial load generated differential movements within the silicone encapsulation. And, as shown in Figure 10, a highly non-uniform strain field was developed around the end-blocks of the sensor. The results of preliminary tests on the prototype sensor showed that the output of a sensor varied non-linearly with the ambient temperature; this was expected. However, more importantly, the temperature sensitivity of the output of the prototype was higher than would be accepted for most structural health monitoring applications.
The findings of the tests show that the prototype needs to be refined to meet a few key target requirements for bridge monitoring: in particular to (a) improve the reliability of the relationship between the applied strain and the output of the sensor, and (b) reduce the temperature sensitivity of that output. The findings also provided suggestions for effecting such refinements.

Conclusions
This publication has presented the challenges and system concept of operation as well as the system modules of the SENSKIN monitoring solution that is being developed in the framework of the SENSKIN EC co-funded research project. The project is currently at the first integrated version where the second version of the sensor has been developed followed by the development of the first version of the data acquisition system (DAQ). The first integrated version of the system has been also implemented and tested including the design of the system electronics, interfaces and data considerations. The next steps of the project will be a set of simulated environment testing of the prototype device that will be executed by the partner TRL (TRL Limited) that will provide direct feedback to the technical activities towards the final implementation of the SENSKIN system. In turn the final system prototype will be validated in the actual bridges of Egnatia Motorway (Greece) and the Bosporus Bridge (KGM). SENSKIN is expected to provide high impact to monitoring and management systems increasing infrastructure capacity and optimizing maintenance costs for all transport modes, new construction and maintenance techniques that enhance the performance and reliability of infrastructure, innovative and cost effective approaches to use Green Infrastructure for transport, extension of the life span of ageing transport infrastructure, development and application of effective and efficient materials, technologies and tools to meet cost-effectiveness and sustainability goals, reduction of multi-modal infrastructure construction and maintenance energy intensity and subsequent CO2 pollutants and noise emissions, support the transition towards zero traffic disruption from inspection, construction and maintenance by 2030 and boost the overall performance of the European transport infrastructure and reduce nuisances generated by transport. The first steps towards the system commercialization have already started though a draft exploitation strategy identifying the SENSKIN market opportunity, initial partners' IPRs as well as a concise strategy for the SENSKIN commercialization.