“…Signal processing is the first procedure to confirm the existence of a fault accurately [16]. Because the initial fault is weak and masked easily with the environmental noise under bearing running, many signal processing methods for extracting initial fault signatures or denoising based on vibration signal were investigated [9], for instance, Empirical Mode Decomposition (EMD) [17], Variational Mode Decomposition (VMD) [18], Singular Value Decomposition (SVD) [19], Multi-resolution SVD (MRSVD) [20], Local Mean Decomposition (LMD) [21], Wavelet Packet Decomposition (WPD) [22], TQWT [15,23] and improved methods [9,10,22]. Although the above methods have a good effect on detecting the early fault of bearing since the solid theoretical foundation, strong robustness and the features of the time-frequency domain are considered [24], in order to overcome their disadvantage and obtain better effectiveness or computational efficiency, a lot of optimal methods are investigated, their convenience is even ignored under the optimal studies in practice.…”