Asia-pacific Journal of Modeling and Simulation for Mechanical System Design and Analysis
| Volume 2, No. 1, 2017, pp 39-46 | ||
| Abstract |
A New Feature Evaluation and Weighting Method Using Three measurement Hybrid Model and Its Application to Fault Diagnosis of Rolling Element Bearing
|
|
Abstract
Aiming at the issue that the prevailing single measurement based feature evaluation methods merely reflect some aspect of information which would be harmful to the enhancement of fault diagnostic accuracy for rolling element bearing, this paper develops a new a hybrid feature evaluation and weighting scheme synchronously considering the correlation, distance and information measurements. Firstly, 10 time-domain statistic feature parameters and 5 frequency-domain ones are extracted from the collected vibration signals to construct a multidomain and multi-class original fault feature set. Secondly, the proposed feature evaluation and weighting scheme is utilized to endow the weighted feature set with more rich and helpful information. Finally, taking the weighted feature set as the input, Least Squares Support Vector Machines (LS-SVM) is employed to implement fault diagnosis of rolling element bearing. Experimental results demonstrate the effectiveness of the proposed method in the improvement of diagnostic accuracy.