Rolling bearing fault diagnosis based on wavelet energy spectrum, PCA and PNN

In order to solve the problem that the excessive dimensions of feature vector will lead to probabilistic neural network (PNN) ´s structure becoming complicated and recognition rate slowing down when we take the wavelet energy spectrum of the rolling bearing vibration signal as the feature vector, a novel approach based on wavelet energy spectrum, principal component analysis (PCA) and probabilistic neural network (PNN) is proposed. The method firstly decomposes the vibration signal by wavelet transform algorithm, separately reconstructs the wavelet coefficients of each level, and calculates each frequency band´s signal energy in the time domain as the feature vector. Then, we use the principal component analysis (PCA) technology to process wavelet energy spectrum so as to reduce its dimensions. Lastly, we feed the principal components into the PNN for recognition. The experimental results show that the proposed method not only can accurately recognize the test set, but also can reduce the dimensions of input feature vector in order to simplify network model, reduce the time required for recognition, and improve the recognition efficiency.