A robust construction of normalized CNN for online intelligent condition monitoring of rolling bearings considering variable working conditions and sources

As one of the most important functional components, the running performances of rolling bearings in rotating machines directly affect the reliability and safety of equipments. However, in actual industrial scenarios, it usually exhibits different failure behaviors caused by harsh and variable working conditions. To effectively guarantee the reliability of operation, a robust construction based on convolutional neural network (CNN) is established for the condition monitoring of rolling bearings. Firstly, the proposed normalized CNN model extracts features from one full-life-cycle data, which contains different state information. Then, the trained model is directly employed for the online monitoring of other rolling bearings. The proposed method is designed to automatically apply to different scenarios and construct health indicators. Two famous datasets are adopted to illustrate the effectiveness and robustness of the proposed method, and the results show that it can achieve excellent performances in condition monitoring under variable working conditions and sources.

Automated summary

A robust construction based on convolutional neural network (CNN) is established for the condition monitoring of rolling bearings, which can effectively extract features and construct health indicators, showing excellent performances in condition monitoring under variable working conditions and sources.