Data mining methodology employing artificial intelligence and a probabilistic approach for energy-efficient structural health monitoring with noisy and delayed signals

Numerous methods have been developed in the context of expert and intelligent systems for structural health monitoring (SHM) with wireless sensor networks (WSNs). However, these techniques have been proven to be efficient when dealing with continuous signals, and the applicability of such expert systems with discrete noisy signals has not yet been explored. This study presents an intelligent data mining methodology as part of an expert system developed for SHM with noisy and delayed signals, which are generated by a through-substrate self-powered sensor network. The noted sensor network has been demonstrated as an effective means for minimizing energy consumption in WSNs for SHM. Experimental vibration tests were conducted on a cantilever plate to evaluate the developed expert system for SHM. The proposed data mining method is based on the integration of pattern recognition, an innovative probabilistic approach, and machine learning. The novelty of the proposed system for SHM with data interpretation methodology lies in the integration of the noted intelligent techniques on discrete, binary, noisy, and delayed patterns of signals collected from self-powered sensing technology in the application to a practical engineering problem, i.e., data-driven energy-efficient SHM. Results confirm that the proposed data mining method employing a probabilistic approach can be effectively used to reconstruct delayed and missing signals, thereby addressing the important issue of energy availability for intelligent SHM systems being used for damage identification in civil and aerospace structures. The applicability and effectiveness of the expert system with the data mining approach in detecting damage with noisy signals was demonstrated for plate-like structures with an accuracy of 97%. The present study successfully contributes to advance data mining and signal processing techniques in the SHM domain, indicating a practical application of expert and intelligent systems applied to damage detection in SHM platforms. Findings from this research pave a way for development of the data analysis techniques that can be employed for interpreting noisy and incomplete signals collected from various expert systems such as those being used in intelligent infrastructure monitoring systems and smart cities. (C) 2019 Elsevier Ltd. All rights reserved.