Decentralized structural health monitoring using sensor networks

Abstract

Structural Health Monitoring (SHM) and damage detection techniques have captured much interest and attention of researchers and structural engineers owing to their potential in providing spatial and quantitative information regarding structural damage and the performance of a structure during its life-cycle. This thesis focuses on investigating the challenges of using sensor networks for monitoring of civil infrastructure. Limited energy of sensors in the case of Wireless Sensor Networks (WSN) and the management and communication costs associated with enormous amounts of data collected by any SHM system are two such challenges that are still being encountered. Energy saving strategies, data compression techniques and several new concepts that achieve these tasks have been explored and developed by the research community to address these issues. This thesis focuses on decentralized damage detection of structures - a novel approach still being researched on, that has the potential of mitigating both energy and data communication issues. Compressive Sensing (CS) is a very recent development which introduces the means of accurately reconstructing under sampled signals with respect to the Nyquist's rate, by exploiting the sparsity of the signal. The potential of CS demonstrated in its recent applications motivated its application for data reduction in SHM systems which may also lead to energy saving in WSN based systems. As the first step, existing decentralized damage detection algorithms and their limitations are explored. Two of the existing algorithms - the Auto Correlation Function-Cross Correlation Function (ACF-CCF) method and the Auto Regression - Auto Regression with eXogenous input method (AR-ARX) method - are comparatively evaluated using experimental data. Based on the insight gained from this analysis, a novel algorithm for decentralized damage detection based on time-series modelling is presented. Successful damage identification is achieved in the verification of this method using experimental data. Motivated by the limitations inherent in the underlying bases of existing damage detection methods and also by the variety of these methods, a novel approach for structural damage detection using the Wiener filter is proposed. Verification of this method using experimental data resulted in successful damage identification. Next, the effectiveness of CS for compressed data acquisition in SHM systems is explored. This investigation provided accurate reconstruction of the SHM signals with good Compression Ratio (CR). Further analysis on the reconstructed signals provided successful damage identification of the structure showing that the application of CS has not compromised the important information carried by the structural responses. A laboratory experiment was carried out to further validate the developments of this thesis using wired and wireless sensors. An initial modal analysis provided evidence that wireless sensors can be as accurate and as reliable as wired sensors for effective SHM. Validation of damage detection techniques and CS confirmed the above observations while presenting successful damage identification results and CS reconstruction results.

Date of Award	2014
Original language	English