Bearing fault diagnosis on various operarion condition of rotating machinery based on LSTM-RNN algorithm

초록

Bearing part is the most important component in rotating machinery to support the mechanical rotating body and reduce the movement friction. Generally, it has a failure mode resulting from the relative motion betw een mating surfaces would cause a damage. Several studies have shown that the bearing fault is the major source in rotating machinery faults [ 1]. An effective fault diagnosis method could obtain the healthy condition of bearings and probe the fault conditio n, which are also the most important directions in research and practice. Conventional fault diagnosis methods generally extract features from the raw process data. Then certain classifiers are adopted to make the diagnosis system . However, the typical met hods require expert knowledge of feature extraction, classifier design, and adaptive processing of dynamic information in raw data , such as frequency domain analysis or statistical analysis [ 3 ,4 This paper proposes fault diagnosis and condition monitori ng method based on Recurrent Neural Network (RNN) and L ong S hort T erm M emory (LSTM) algorithm. The suggested method can directly classify the raw process data without specific feature extraction and classifier design. It is also able to adaptively learn th e dynamic information in raw data.