Multiwavelet denoising with adaptive block thresholding and its application in gearbox diagnosis of rolling mills

In order to efficiently extract weak fault features of key equipments immersed in strong background noise, a multiwavelet denoising method with adaptive block thresholding is proposed and it is applied to gearbox fault diagnosis of the rolling mills. The effect of wavelet denoising mainly depends on the optimal selection of wavelet functions and threshold. Multiwavelets have more than two multiscaling functions and multiwavelet functions. They possess such properties as orthogonality, symmetry, compact support and high vanishing moments simultaneously. Therefore, multiwavelets are extensively used for fault diagnosis of incipient faults and weak faults. Based on the correlation of multiwavelet coefficients, this paper uses the minimum principle of Stein's unbiased risk estimate to estimate the true fault features. The optimal block length and threshold are selected for effective feature extraction and noise elimination at each decomposition level. The simulation signal validates the effectiveness of the proposed method, the gearbox fault diagnosis of the rolling mills indicates that the proposed method can successfully detect two local scuffing fault features of the pinion simultaneously.