Adaptive stochastic resonance method for impact signal detection based on sliding window

Aiming at solving the existing sharp problems in impact signal detection by using stochastic resonance (SR) in the fault diagnosis of rotating machinery, such as the measurement index selection of SR and the detection of impact signal with different impact amplitudes, the present study proposes an adaptive SR method for impact signal detection based on sliding window by analyzing the SR characteristics of impact signal. This method can not only achieve the optimal selection of system parameters by means of weighted kurtosis index constructed through using kurtosis index and correlation coefficient, but also achieve the detection of weak impact signal through the algorithm of data segmentation based on sliding window, even though the differences between different impact amplitudes are great. The algorithm flow of adaptive SR method is given and effectiveness of the method has been verified by the contrastive results between the proposed method and the traditional SR method of simulation experiments. Finally, the proposed method has been applied to a gearbox fault diagnosis in a hot strip finishing mill in which two local faults located on the pinion are obtained successfully. Therefore, it can be concluded that the proposed method is of great practical value in engineering.

► An adaptive SR method for weak impact signal detection is proposed.
► Weighted kurtosis index is proposed based on SR characteristics of impact signal.
► A new scheme of sliding window-based data segmentation is developed to improve SR.
► This method can effectively detect one or more impact faults in rotating machine.
► The superiority of the proposed method is verified by simulation and gearbox case.