Bearing Fault Diagnosis using Deep Belief Networks

摘要

This paper presents an experimental study on bearing fault diagnosis using a Deep Belief Network and the Genetic Algorithm for parameter optimization. Arnbearing test-rig is proposedly built to simulate various bearing operation conditions in the study, namely, Healthy, Inner Race fault, Ball fault and Outer Racernfault. The diagnosis technique is then employed to analyse the experimental data acquired from the bearing test-rig and to recognize the bearing operationrnconditions based on the fault patterns detected by the algorithm. It is shown that the diagnosis technique proposed in this study can successfully discriminaternthe four bearing fault conditions with rather high accuracy and a good computational efficiency.