Fault diagnosis of machines based on D–S evidence theory. Part 2: Application of the improved D–S evidence theory in gearbox fault diagnosis

Fault diagnosis requires reasoning and decision-making based on diagnostic knowledge and features extracted from raw data. In practice, fault features may be uncertain and imprecise due to sensor errors, fluctuating working conditions, and limitations of feature extraction methods. Features may not be apparent when a fault is in the early stages of development. In addition, diagnostic knowledge is not always accurate because most of it is obtained from experts’ experience. In Part 1 of this study, a new decision method is proposed that can deal with these issues, combine multi-evidence information from different methods, and provide more accurate diagnostic results. It is an improvement on conventional D–S evidence theory. Part 2 of this study reports an application of the improved D–S evidence theory in gearbox fault diagnosis. Compared with conventional diagnostic methods, the proposed method can enhance diagnostic accuracy and autonomy.