Fractional Fourier Domain Feature Fusion Combining Multichannel Targeting Extreme Learning Machine for Bearing Fault Diagnosis
Zong Meng, Ruxue Bai, Quansheng Xu, Jimeng Li, Fengjie Fan
Abstract
The dependence on big data and lengthy training time discounts the advantages of deep learning method in fault diagnosis to a certain extent. In this article, we proposed an easy-to-use and cost effective method for bearing fault diagnosis that incorporates maximum kurtosis based fractional Fourier transform (MK-FRFT) as a feature extractor and multi-channel targeting extreme learning machine (MCT-ELM) as feature fusion and classifier. Experimental studies demonstrate that MCT-ELM paired with MK-FRFT is fast to achieve an accurate fault diagnosis with small sample length and limited data volume. The superiority is also validated by comparison with some advanced deep learning and machine learning methods in terms of diagnostic accuracy, time cost, robustness to noise and load variation as well as dependence to data volume. The proposed approach is promising to be applied in some practical scenarios with difficulties obtaining large data volume and/or limited hardware configuration.