An Investigation of Weighted Neural Networks for Rolling Bearing Fault Classification Under Uncertain Speed Condition

Abstract

The fault recognition of bearing under uncertain speed condition is an important task for rotating machinery health monitoring. Since the speed shows a serious influence on the intrinsic characteristics of the acquired vibration signal, there is a significant difference for the characteristic distribution of the vibration signal under different speed conditions, and thus will lead a high misjudgment rate for fault identification of rolling bearings. In this manner, this paper proposes a method of bearing fault recognition under uncertain speed condition based on weighted neural network. Compared with the traditional network, the proposed architecture builds a speed-insensitive fault identification network by embedding a new unit with the consideration of different speed nodes, through the weight network and hence greatly improves the fault recognition accuracy of bearing faults under uncertain speed condition. Compared with the conventional methods, the identification result shows the feasibility and effectiveness of the proposed method.