Abstract
Due to the powerful ability of EEMD algorithm in noising, it is usually applied to feature extraction of fault signal of rolling bearing. But the selective correctness of sensitive IMF after decomposition can directly influence the correctness of feature extraction of fault signal. In order to solve the problem, the paper firstly proposes a new method on selecting sensitive IMF based on Cloud Similarity Measurement. By comparing this method in simulation experiment with the traditional mutual information method, it is obvious that the proposed method has overcome the misjudgment in the traditional method and it has higher accuracy, by factually collecting the normal, damage, and fracture fault AE signal of the inner ring of rolling bearing as samples, which will be decomposed by EEMD algorithm in the experiments. It uses Cloud Similarity Measurement to select sensitive IMF which can reflect the fault features. Finally, it sets the Multivariate Multiscale Entropy (MME) of sensitive IMF as the eigenvalue of original signal; then it is classified by the SVM to determine the fault types exactly. The results of the experiments show that the selected sensitive IMF based on Cloud Similarity Measurement is effective; it can help to improve the accuracy of the fault diagnosis and feature extraction.
Highlights
Rolling bearing is the most common part of the rolling mechanism, and about 30% of the mechanical faults are caused by rolling bearing, so the detection and diagnosis of rolling bearing are always hot issues studied by scholars all over the world
The Cloud Similarity Measurement is used to select the sensitive Intrinsic Mode Function (IMF); this method has been proved with high accuracy and could overcome the misjudgment, after the simulation experiments
It is more accurate and it can help to overcome the misjudgment caused by mutual information method
Summary
Rolling bearing is the most common part of the rolling mechanism, and about 30% of the mechanical faults are caused by rolling bearing, so the detection and diagnosis of rolling bearing are always hot issues studied by scholars all over the world. Lilin et al have extracted eigenvalue of AE signal by using wavelet packet energy spectrum; it can help to improve the SNR of the signal [10] This method has the difficulties in selecting the wavelet base and determining the threshold. A part of the IMF after decomposing by EEMD is closely related to the original fault information It is significant for the proper selecting of the sensitive IMF which is closely related to the fault information in improving the accuracy of feature extraction of fault signal. CSM is extensively used in the e-business trade, biomedicine, and watermarking technology It has never be reported in the field of feature extraction of AE signal of rolling bearing. All the above can effectively extract the fault characteristic information of AE signal and improve the accuracy of fault diagnosis
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
