A new interval native Bayes uncertain fault diagnosis method based on the firefly algorithm

Traditional fault diagnosis methods assume that the rolling bearing fault samples are precise data. However, this assumption may be wrong when there are problems of measurement uncertainty etc. Due to this, this paper proposes an interval native Bayes uncertain fault diagnosis method based on the firefly algorithm. First, the interval fault vibration signals are decomposed by intrinsic time scale decomposition, and several proper rotation components (PRC) are obtained. Features of PRC, such as interval kurtosis etc., are extracted as fault samples. Then, an interval native Bayes uncertain fault diagnosis method is designed for these uncertain rolling bearing interval features. Conventionally, the fault diagnosis method utilizes the same interval features to distinguish different fault types. However, each type of fault has its own distinctive classification accuracy for different features. Thus, this paper uses the firefly algorithm to extract different optimal interval fault features for different fault types. Experimental results show: (i) the proposed interval native Bayes method can be effectively applied in the interval fault diagnosis of rolling bearing under measurement uncertainty conditions. (ii) Compared with two traditional methods which extract the same fault features for all fault types, the new method can obtain higher classification accuracy.