Abstract: Aiming at the problem that it was necessary for traditional diagnosis models to set many super parameters for fault identification, and the parameters had a significant impact on the performance of the model, a rolling bearing fault diagnosis method based on enhanced hierarchical attention entropy (EHATE) and gray wolf algorithm optimized extreme learning machine (GWO-ELM) was proposed. In which the EHATE method was used to extract low-frequency and high-frequency feature information of rolling bearing vibration signals, while GWOELM was used to identify fault categories of rolling bearings. Firstly, based on fractal theory and enhanced hierarchical analysis, an index called enhanced hierarchical attention entropy (EHATE) was proposed, which could simultaneously measure the complexity of non-stationary time series in low and high frequency bands. Then, EHATE method was used to fully extract the fault characteristics of rolling bearing vibration signals to achieve accurate characterization of different sample fault states. Finally, the fault characteristics were input into the GWO-ELM classifier to identify the fault type and severity. Based on the EHATE+GWO-ELM model, three sets of rolling bearing fault data sets were tested and compared with other fault diagnosis methods. The research results show that the fault diagnosis model can quickly and effectively identify different faults of rolling bearings, and the fault recognition accuracy of the three sets of data sets respectively reaches 100%, 99%, and 96.92%, which is superior to the comparison method in terms of recognition accuracy and feature extraction efficiency. At the same time, the fault diagnosis model only needs to set a single parameter in the feature extraction stage, and the parameter has a very small impact on the recognition accuracy of the model. The study provides a new perspective and scheme for fault diagnosis of rolling bearings.
Key words: enhanced hierarchical attention entropy(EHATE); extreme learning machine(ELM); gray wolf algorithm optimize(GWO); fault feature extraction; fault type identification; rolling bearing vibration signals