Application of an MAKD method via Dixon’s test for gearbox fault diagnosis based on built-in encoder signals

The weak fault feature caused by the planetary gearboxes under speed-varying condition (SVC) is easily submerged by the complex meshing vibration reference and noise due to the multi-degree-of-freedom motion coupling characteristic, which makes its fault diagnosis challenging. In contrast to vibration signals, built-in encoder signals intrinsically possess more comprehensive fault-related information, rendering them more appropriate for machinery fault diagnosis. Nevertheless, current research does not present an effective methodology for integrating deconvolution techniques with encoder signals to facilitate dependable fault diagnosis across variable speeds. Motivated by this, the paper proposes a maximum average kurtosis deconvolution via Dixon’s test method for the gearbox fault feature extraction of encoder signals under SVC. The method first converts the encoder signal into instantaneous angular acceleration signal through the second-order derivative for the amplification of the fault information. Then, based on the fault characteristic order of gearbox faults, the adjacent fault cycles are segmented and the initial kurtosis vector is constructed by calculating the kurtosis of each fault cycle. Therein, Dixon’s test is introduced to remove outliers from the kurtosis vector. Maximizing the average kurtosis as an objective to guide the optimization direction of the deconvolution process for fault feature extraction under SVC. Finally, simulation and experimental cases are used to verify the effectiveness of the proposed method.