Sensor fault detection in a damage detection approach based on piezodiagnostics
Document typeConference report
Rights accessOpen Access
Online monitoring systems demand an adequate operation of sensor system used to acquire structural state measurements. If a damaged sensor record is incorporated in the diagnosis algorithm, it could be generate uncertainties and generate unsuitable alarms. Thus, appropriate operation of sensor system is a critical requirement in order to obtain a high reliability for structural damage diagnosis algorithms. In this work a data-driven procedure is studied in order to mitigate the faulty sensor effect in a monitoring system. The studied method takes advantage of piezo-diagnostics approach, where piezoelectric devices are attached to the surface of the monitored structure to produce guided waves. Thus, piezoelectric measurements are analyzed by applying principal component analysis and cross-correlation, in order to detect abnormal behaviors. In this sense, the squared prediction error Q and Hotelling squared statistical indices are used to observe a typical behaviour caused by sensor problems or structural damages. The methodology is validated on a lab carbon steel pipe section by using scenarios that include electric power failures, disconnecting power cords as well as mass adding. As concluding remark, in this work was possible to separate structural damage and fault sensor states at different clusters.
CitationRuiz, M., Camacho-Navarro, J., Villamizar, R., Mujica, L.E. Sensor fault detection in a damage detection approach based on piezodiagnostics. A: European Workshop on Structural Health Monitoring. "8th European Workshop on Structural Health Monitoring (EWSHM 2016): Bilbao, Spain, 5-8 July 2016". Bilbao: 2016, p. 1-10.