Sparse dictionary analysis via structure frequency response spectrum model for weak bearing fault diagnosis

Rolling element bearings are the critical parts of every rotating machinery and their failure is one of the main reasons for the machine downtime and even breakdown. It is a big challenge to extract the weak transient impulse related to fault features of rolling bearing under strong background noise. Dictionary Learning (DL) based on Sparse Representation Theory (SRT), an effective means in handling this problem, has received tremendous attention both in academia and industry, However, the methods based on DL inevitably has its own drawbacks in either experimental or engineering implementation in terms of anti-interference, the great majority of methods fail to achieve a good balance between the robustness of anti-noise and the generalization. Therefore, a novel sparse structure frequency analysis framework based on DL is proposed to address this problem in this paper, A K-SVD(generalizing the K-means clustering process) based DL algorithm is firstly introduced to capture the inner structure information of the row fault signal, then a novel Sparse Frequency Response Spectrum Model (SFRSM) is proposed to expose the fault information with sparse frequency structure graph (SFSG) in a straightforward and detailed manner, and a new global filtering feature extraction algorithm called frequency response function editing filtering (AFEF) is also proposed to obtain relevant fault features masked by noise. Simulation analysis validates the effectiveness of the method. The experimental result based on the bearing fault test bed demonstrate the superiority of the proposed method in terms of anti-noise and adaptability compared with other state of art algorithms.