Fault Diagnosis Method of Rolling Bearings Based on Supercomplete Dictionary Learning

Dou An, Chunlin Hu

Abstract: As one of the most important technologies to guarantee the safe operation of industrial equipment, fault diagnosis technology has gained wide attention. Rolling bearings are indispensable and wearing parts for large machinery equipment. It’s greatly significant to find out the fault type, fault severity, and fault location in time for maintaining the normal operation of a mechanical system. Based on the advantages of the supercomplete dictionary learning model, this paper proposes a new feature extraction method, which is combined with the Softmax classifier for rolling bearing fault diagnosis. We use a measured rolling bearing data set to prove the effect of our method. Then we design contrast experiments to compare our method with traditional methods. The experiment results show that our method can accurately diagnose multifarious bearing faults, and the supercomplete dictionary model can extract the characteristics of vibration signals well, which is superior to traditional research efforts.