A New Method of Multi-feature Fusion Rolling Bearing Fault Diagnosis Based on GA-DHMM

In the process of rolling bearing fault diagnosis, single feature has the limited ability to characterize a bearing vibration signal, and the number of training samples is also limited. To solve these problems, this paper proposes a new method of multi-feature fusion rolling bearing fault diagnosis based on Genetic Algorithm-Discrete Hidden Markov Model (GA-DHMM). Firstly, the features including time-domain, frequency-domain, wavelet packet energy and multi-scale sample entropy are extracted from the preprocessed vibration signals. Then, the features are evaluated and selected by the distance discriminant factor, the selected features form the joint feature space after normalization. Finally, GA-DHMM is used to realize rolling bearing fault diagnosis with limited training samples. The bearing dataset from Case Western Reserve University is used to verify the effectiveness of the proposed method. Compared with the method with single feature and BP neural network with multi-feature, the effectiveness of the proposed method is proved.