Health condition monitoring and early fault diagnosis of bearings using SDF and intrinsic characteristic-scale decomposition

Early fault diagnosis is crucial to reduce the machine downtime. This paper presents a novel method based on symbolic dynamic filtering (SDF) for early fault detection and intrinsic characteristic-scale decomposition (ICD) for fault type recognition. SDF is first applied to extract the fault feature for depicting bearing performance degradation. Then, a fault alarm is triggered using cumulative sum. Finally, the extracted abnormal signal is decomposed by the ICD method, and the kurtosis method is used to select a principal product component that contains most fault information for fault detection. The real life experimental results validate the effectiveness of the proposed method in early detection of bearing fault and fault diagnosis in comparison with Fourier transform, Hilbert envelope spectrum, original local mean decomposition and spectral kurtosis.