Sparse representation based fault diagnosis of bearings

摘要

This paper propose a novel fault diagnosis of bearings approach based on sparse representation. Three steps are conducted to classify the fault types. In the dictionary learning step, dictionary is learned using training set with known fault types; in the sparse coding step, testing samples with unknown fault types are represented through spares representation model with sub-dictionaries extracted from the dictionary learned in the first step; in the last step, fault types are evaluated through a Hough voting process. Different from most conventional fault diagnosis methods, rejection mechanism can be easily introduced to guarantee the diagnosis accuracy, which is suitable in areas where high diagnosis accuracy is required such as nuclear and aerospace industries. Finally, an experiment is undertaken to verify the effectiveness of our model.