Diagnosis of an incipient defect in a worm gearbox using minimum entropy deconvolution and local cepstrum

Abstract

Abstract A worm gearbox is a slow speed gear arrangement which can greatly reduce velocity (in some cases 100:1 in a single stage). The early detection of a gear fault in this high reduction gearbox is a challenging task as the fundamental frequencies in the spectrum are close together. The sliding mechanism of its operation and the presence of loud noise further complicate the task. A signal processing technique has been proposed to extract weak fault features in the worm and wheel gearbox signal. In the preprocessing stage, the signal is decimated, and an autoregressive minimum entropy deconvolution (AR-MED) is applied. The decimation process enhances the data handling capability. The AR-MED filter denoises the signal. The signal obtained is further processed for the local cepstrum (LC) to identify the quefrencies. The quefrencies provide information regarding the period of repetition of impulses corresponding to the defect. The results of quefrencies validate the values of fault frequency for the faulty gear. A comparison with the autocorrelation LC and squared envelope spectrum kurtosis is presented to establish the effectiveness of the proposed scheme. The accuracy of the results of the proposed method is 99.27%.