No abstract

No abstract

The rolling bearing fault detection technology solution can be used to maintain the machinery in advance and avoid downtime and safety accidents. However, the current studies primarily focus on utilizing complex algorithms to improve the accuracy of fault diagnosis, while neglecting the influence of extracted features on diagnosis results, resulting in redundancy and conflict among features. To address such an issue, a feature selection method based on compensation distance evaluation technique (CDET) is proposed to optimize the feature set extracted from acoustic emission (AE) signals. Firstly, multiple failure data of the bearing are modeled by the fault simulation experiment platform. Then, the proposed feature selection method is used to evaluate the sensitive features. Finally, hidden markov modelsupport vector machine (HMM-SVM) series model is used to predict works. The experimental results show that the proposed method selects 7 sensitive features from the feature set composed of 13 features and achieve a high accuracy (99.3%).