Mechanical Fault Diagnosis of Rolling Bearing Based on Locality-constrained Sparse Coding

In the mechanical fault diagnosis and signal processing domain, there has been growing interest in sparse coding which is advocated as an effective mathematical description for the underlying principle of sensory systems in signal processing. In this paper, a natural extension of sparse coding, locality-constrained sparse coding, is introduced as a feature extraction technique for machinery fault diagnosis. Then, the vibration signals of rolling element bearings are taken as the target signals to verify the proposed scheme, and locality-constrained sparse coding is used for vibration analysis. With the purpose of diagnosing the different fault conditions of bearings, features are extracted according to the following scheme: basis functions are learned from each class of vibration signals by extracting the time-domain and frequency-domain features. A redundant dictionary is built by merging all the learned basis functions. Based on the redundant dictionary, the diagnostic information becomes explicit in the solved sparse representations of vibration signals. Sparse features are formulated in terms of atom activations. A support vector machine (SVM) classifier is used to test the discriminability of the extracted sparse features. Experiments show that locality-constrained sparse coding is an effective feature extraction technique for machinery fault diagnosis.