Adaptive detrended fluctuation analysis as a feature extraction method for gear's vibration signal

作者：Wu Yong Heng; Xiao Han^*

来源：Journal of Vibration and Shock, 2015, 34(9): 154-157.

DOI：10.13465/j.cnki.jvs.2015.09.028

• Clustering
• Detrended fluctuation analysis
• DFA
• Extraction efficiencies
• Feature extraction methods
• Long range correlations
• Neural network algorithm
• Scaling exponent

Scaling exponent is often used as a feature measure for gear fault identification in detrended fluctuation analysis (DFA), but it leads easily to the problem of aliasing among gear failure modes. According to a logarithmic scale wave function diagram, the scaling exponents and the intercept of characterization of signal intensity were combined to make up a feature vector of gear vibration signal. According to the dual scaling character of gear vibration signal, a sliding windowing algorithm was projected to extract adaptively the scaling exponent, and used for fault classification combining with the neural network algorithm. The results show that the method can improve the extraction accuracy and extraction efficiency of the scaling exponent.

• 出版日期2015-5-15
• 单位武汉科技大学