Application of wavelet packet-locally linear embedding algorithm in rolling bearing fault degree recognization

作者：Kang Shouqiang^*; Li Zhuqiang; Yang Guangxue; Wang Yujing

来源：Chinese Journal of Scientific Instrument, 2014, 35(3): 614-619.

• Feature dimensions
• Frequency domain indexes
• Local Linear Embedding
• Multiple samples
• Performance degradation
• Performance degradation assessment
• Rolling bearings
• Wavelet Packet Decomposition

Performance degradation degree recognization of rolling bearing is the premise of performance degradation assessment. In order to improve the recognization accuracy of performance degradation degree, based on wavelet packet combined with local linear embedding algorithm, a fault performance degradation degree recognization method of rolling bearing is proposed. First, the time domain and frequency domain indexes of a sample vibration signal are calculated, the wavelet packet decomposition and reconstruction are performed for the vibration signal. The time domain, frequency domain indexes and the singular values of the node signal are extracted to construct the feature vector, and the feature matrix is constructed from the feature vectors of multiple samples. Then, local linear embedding algorithm is used to reduce the feature dimension of the feature matrix. At last, different performance degradation degrees of the rolling bearing are recognized with support vector machine, and the feasibility and effectiveness of the proposed method are verified. Experimental results show that different performance degradation degrees of the rolling bearing can be recognized accurately using the proposed method.

• 出版日期2014