Diagnostics and prognostics using switching Kalman filters

The use of condition monitoring data for diagnostic and prognostic of vehicle health has been growing with increasing use of health and usage monitoring systems. In this article, an approach using the switching Kalman filter framework is explored for both diagnostic and prognostic using condition monitoring data under a single framework. The switching Kalman filter uses multiple dynamical models each describing a different degradation process. The most probable underlying degradation process is then inferred from the observed condition monitoring data using Bayesian estimation. By using the dynamical behavior of the degradation process, pre-established fault detection threshold is no longer required. This approach also provides maintainers with more information for decision-making as a probabilistic measure of the degradation processes is available. This helps maintainers to predict remaining useful life more accurately by distinguishing between the degradation states and performing prediction only when unstable degradation is detected. The proposed switching Kalman filter approach is applied onto sets of condition monitoring data from gearbox bearings that were found defective from the Republic of Singapore Air Force AH64D helicopter. The use of in-service data in a practical scenario shows that the switching Kalman filter approach is a promising tool for maintenance decision-making.