Deep Clustering Bearing Fault Diagnosis Method Based on Local Manifold Learning of an Autoencoded Embedding

Abstract

In many practical fault diagnosis applications, the acquisition of fault data labels requires substantial manpower and resources, which are sometimes impossible to achieve. To address this, an unsupervised bearing fault diagnosis method based on deep clustering is proposed. In this method, an autoencoder is initially applied to the signal spectrum to learn the initial representation. Then, its potential manifold is further searched, and a Gaussian mixture model is finally used for clustering. Experiments conducted on the Case Western Reserve University bearing datasets show that the proposed method can find the optimal clusterable manifold. Moreover, its clustering performance is better than those of the current advanced baseline methods, and it is only slightly complex. Thus, the effectiveness of the proposed method is verified.

Keywords

• Bearing fault diagnosis
• autoencoded embedding representation
• local manifold learning
• manifold re-embedding
• deep clustering
• clusterable manifold