Construction and selection of lifting-based multiwavelets for mechanical fault detection

• New multiwavelets with diverse vanishing moments are constructed by lifting schemes.
• Multiwavelets selection rules of local spectral entropy minimization are proposed.
• The selection rules are classified for shaft, gear and bearing faults.
• The method is applied to incipient fault diagnosis of rolling bearing.
• Also, it is applied to gearbox fault diagnosis of rolling mill.

The essence of wavelet transforms is a similar measurement between the signal and the wavelet basis functions. Thus, the construction and selection of the proper wavelet basis functions similar to the fault feature and possessing good properties such as vanishing moments have vital importance to the effective fault diagnosis. In this paper, the construction of lifting-based adaptive multiwavelets with various vanishing moments and the selection rules for different mechanical fault detection are proposed. On the basis of the fixed cubic Hermite multiwavelets, lifting schemes are adopted to construct new changeable multiwavelets with diverse vanishing moments. Then, the defined local spectral entropy minimization rules are proposed to determine the optimum multiwavelets providing the proper vanishing moments, classified into the typical shaft faults, gear faults and rolling bearing faults. The proposed method is applied to incipient fault diagnosis of rolling bearing and gearbox fault diagnosis of rolling mill to verify its effectiveness and feasibility in comparison with different wavelet transforms and spectral kurtosis. The results show that the proposed method can act as a promising tool for mechanical fault detection.