Bayesian inference-based wear prediction method for plain bearings under stationary mixed-friction conditions

Abstract

This study introduces a method to predict the remaining useful life (RUL) of plain bearings operating under stationary, wear-critical conditions. In this method, the transient wear data of a coupled elastohydrodynamic lubrication (mixed-EHL) and wear simulation approach is used to parametrize a statistical, linear degradation model. The method incorporates Bayesian inference to update the linear degradation model throughout the runtime and thereby consider the transient, system-dependent wear progression within the RUL prediction. A case study is used to show the suitability of the proposed method. The results show that the method can be applied to three distinct types of post-wearing-in behavior: wearing-in with subsequent hydrodynamic, stationary wear, and progressive wear operation. While hydrodynamic operation leads to an infinite lifetime, the method is successfully applied to predict RUL in cases with stationary and progressive wear.