Model-based prognostics of gear health using stochastic dynamical models

In this paper we present a statistical approach to estimating the time in which an operating gear will reach a critical stage. The approach relies on measured vibration signals. From these signals features are first extracted and then their evolution over time is predicted. This is done based on a dynamic model that relates hidden degradation phenomena to measured outputs. The Expectation–Maximization algorithm is used to estimate the parameters of the underlying state-space model on line. The time to reach the safety alarm threshold is determined by estimating the distribution of the remaining useful life using the estimated linear model. The results obtained on a pilot test bed are presented.