Weighted KPCA Based on Fault Feature Selection and Feature Information Fusion
Zhang Heng;Zhao Rong-zhen;
Journal of Vibration and Shock ›› 2014, Vol. 33 ›› Issue (9) : 89-93.
Weighted KPCA Based on Fault Feature Selection and Feature Information Fusion
Aimed at the complexity for the corresponding relation between fault characteristics and fault categories of rotating machinery, a twelve channel fault information set for a double span rotor system is constructed, a new method about the feature extraction based on weighted KPCA is proposed. At first, the time domain, frequency domain, time and frequency domain for a single channel vibration signal is extracted, and the original fault feature set is obtained from the twelve channel of the monitoring system. Secondly, a sensitive feature subset of fault is screened from the original feature set by using multiple criterion of feature selection method. And then, a fusion feature vector is obtained by fusing the sensitive feature subset. Finally, the main components of fusion feature are extracted by using the weighted kernel principal component analysis. Experiment result shows that this method can search a sensitive feature subset, the kernel main components can show the gap between the different fault categories effectively.
feature selection / information fusion / weighted kernel principal component analysis / fault diagnosis. {{custom_keyword}} /
/
〈 | 〉 |