Sparse Bayesian Learning with Generalized Double Pareto Prior for Bearing Fault Diagnosis

Vibration analysis has been widely used in bearing fault diagnosis. However, accurate fault impulse extraction from vibration signals with strong background noise is still a challenging task for bearing fault diagnosis. Although the sparse representation (SR) methods can achieve state-of-the-art performance, most of these methods rely on manually tuning several hyper-parameters, and performance may be degraded due to approximate regularization and/or heuristic sparse models. To overcome these shortcomings and extract the fault impulse more accurately, in the paper, we present a sparse Bayesian learning (SBL) framework for bearing fault diagnosis, where we exploit the group-sparsity structure and introduce additional Generalized Double Pareto (GDP) prior to enforce the sparsity of the fault impulse heavily. Through several groups of experiments, the effectiveness and superiority of the proposed method are proved.