Adaptive Fault Diagnosis for Simultaneous Sensor Faults in Structural Health Monitoring Systems

Structural health monitoring (SHM) is a non-destructive testing method that supports the condition assessment and lifetime estimation of civil infrastructure. Sensor faults may result in loss valuable data erroneous structural assessments estimations, worst case with damage remaining undetected. As result, concepts fault diagnosis (FD) have been increasingly adopted by SHM community. However, most FD for consider only single-fault occurrence, which oversimplify actual occurrences real-world systems. This paper presents an adaptive approach systems addresses simultaneous occurring multiple sensors. The encompasses detection, isolation, accommodation, it builds upon analytical redundancy, uses correlated from sensors system. Specifically, are detected using predictive capabilities artificial neural network (ANN) models leverage correlations within sensor data. Upon defining time instances data, isolated analyzing moving average individual around instances. For ANN adapted removing faulty prior to occurrence produce virtual outputs substitute proposed validated via two tests recorded system installed on railway bridge. results demonstrate capable ensuring accuracy, reliability, performance systems, occur simultaneously.