Fault Feature Extraction of Rolling Bearing Based on LFK

Based on multiple embedding theory, traditional multi-scale entropy is optimized by local characteristic-scale decomposition (LCD) and fast kurtogram (FK) which can be called LFK for short. In the improved method, the vibration signal of rolling bearing is decomposed by LCD and the two component signals whose cross correlation coefficient with the original signal is bigger than others are selected. FK is applied for filtering the reserved component signal and highlighting the fault feature. Multivariate multi-scale entropy is extracted from the processed signal to characterize the degradation state of rolling bearings. Compared with multi-scale entropy of the original signal, multivariate multi-scale entropy has a better performance.