Fault diagnosis in gearbox using adaptive wavelet filtering and shock response spectrum features extraction

A wavelet adaptive filtering technique is presented for enhanced fault identification in gearboxes. Based on Morlet wavelet analysis and conventional optimization methods, an adaptive filtering is performed for the background noise removal of vibration signals emanating from gearboxes. A fourth-order statistical moment, kurtosis, is used as an objective function to optimize. A filtered signal is obtained by choosing the suitable Morlet wavelet that maximizes the kurtosis. The optimization framework uses one-dimensional and multidimensional accelerated search techniques to speed up the convergence in solution search space. A novel, transient-based features extraction method based on the shock response spectrum is used to extract characteristic features representing the health state of the gearbox. The effectiveness and feasibility of the proposed method have been demonstrated on experimental gearbox data. The proposed technique enables a high signal-to-noise ratio for gearbox fault detection.