DECENTRALIZED AUTONOMOUS FAULT DETECTION IN WIRELESS STRUCTURAL HEALTH MONITORING SYSTEMS USING STRUCTURAL RESPONSE DATA

Sensor faults can affect the dependability and the accuracy of structural health monitoring (SHM) systems. Recent studies demonstrate that artificial neural networks can be used to detect sensor faults. In this paper, decentralized artificial neural networks (ANNs) are applied for autonomous sensor fault detection. On each sensor node of a wireless SHM system, an ANN is implemented to measure and to process structural response data. Structural response data is predicted by each sensor node based on correlations between adjacent sensor nodes and on redundancies inherent in the SHM system. Evaluating the deviations (or residuals) between measured and predicted data, sensor faults are autonomously detected by the wireless sensor nodes in a fully decentralized manner. A prototype SHM system implemented in this study, which is capable of decentralized autonomous sensor fault detection, is validated in laboratory experiments through simulated sensor faults. Several topologies and modes of operation of the embedded ANNs are investigated with respect to the dependability and the accuracy of the fault detection approach. In conclusion, the prototype SHM system is able to accurately detect sensor faults, demonstrating that neural networks, processing decentralized structural response data, facilitate autonomous fault detection, thus increasing the dependability and the accuracy of structural health monitoring systems.

[1]  Oliver Obst.  Poster abstract: Distributed fault detection using a recurrent neural network , 2009, 2009 International Conference on Information Processing in Sensor Networks.

[2]  D. E. Rumelhart,et al.  Learning internal representations by back-propagating errors , 1986 .

[3]  Keith Worden,et al.  An introduction to structural health monitoring , 2007, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences.

[4]  Martin A. Riedmiller,et al.  A direct adaptive method for faster backpropagation learning: the RPROP algorithm , 1993, IEEE International Conference on Neural Networks.

[5]  Ramaswamy Vaidyanathan,et al.  Process fault detection and diagnosis using neural networks , 1990 .

[6]  Ka-Veng Yuen,et al.  On the complexity of artificial neural networks for smart structures monitoring , 2006 .

[7]  Kay Smarsly,et al.  Decentralized fault detection and isolation in wireless structural health monitoring systems using analytical redundancy , 2014, Adv. Eng. Softw..

[8]  Geoffrey E. Hinton,et al.  Learning representations by back-propagating errors , 1986, Nature.

[9]  Kay Smarsly,et al.  A Decentralized Approach towards Autonomous Fault Detection in Wireless Structural Health Monitoring Systems , 2014 .

[10]  Russell Beale,et al.  Handbook of Neural Computation , 1996 .

[11]  Kishan G. Mehrotra,et al.  Elements of artificial neural networks , 1996 .

[12]  Asad I. Khan,et al.  Graph neuron and hierarchical graph neuron, novel approaches toward real time pattern recognition in wireless sensor networks , 2009, IWCMC.