Multi-signal fusion diagnosis of gearbox based on minimum Bayesian risk reclassification and adaptive weighting

Abstract

Gearbox mainly including gears and bearings plays a significant role in adjusting the speed and torque in the rotating machinery, whose failure will seriously affect the production safety and industrial process. In this paper, a multi-signal fusion fault diagnosis method based on a reclassification model and an adaptive weighting mechanism is proposed. Different from the previous decision-making process of PNN, our reclassification model can make full use of the diagnosis information and prior knowledge, and a classification risk coefficient is introduced to derive a lower risk and more accurate results. The adaptive weighting mechanism designed for decision information fusion could assign more reasonable and flexible weights without depending on the experts. Finally, the proposed method is verified by a power transmission experimental platform. The results show that the proposed multi-signal fusion fault diagnosis method can effectively fuse multiple signals to achieve robust gearbox fault diagnosis and yields the promised results.