A robust health prediction using Bayesian approach guided by physical constraints

Abstract

Accurately predicting the remaining useful life (RUL) of industrial machinery is crucial for ensuring their reliability and safety. Prognostic methods that rely on Bayesian inference, such as the Bayesian method (BM), Kalman and Particle filter (KF, PF), have been extensively studied for RUL predictions. However, these algorithms can be affected by noise when training data are limited or uncertainty when empirical models are employed in place of accurate physics models. As a result, this can lead to significant prediction errors or even infeasible RUL predictions. To overcome this challenge, three different approaches are proposed to guide the Bayesian framework by incorporating low-fidelity physical information. The key idea is to impose inequality constraints to reduce sensitivity to noisy observations and achieve robust prediction. To evaluate the feasibility of the approaches, their performance is evaluated by a numerical example and real case study for drone motor degradation.