Applied Sensor Fault Detection, Identification and Data Reconstruction

Sensor fault detection and identification (SFD/I) has attracted considerable attention in military applications, especially when safety- or mission-critical issues are of paramount importance. Here, two readily implementable approaches for SFD/I are proposed through hierarchical clustering and self-organizing map neural networks. The proposed methodologies are capable of detecting sensor faults from a large group of sensors measuring different physical quantities and achieve SFD/I in a single stage. Furthermore, it is possible to reconstruct the measurements expected from the faulted sensor and thereby facilitate improved unit availability. The efficacy of the proposed approaches is demonstrated through the use of measurements from experimental trials on a gas turbine. Ultimately, the underlying principles are readily transferable to other complex industrial and military systems.