Weighted Kurtogram Method With Application To Fault Diagnosis

Modern machineries are becoming more and more large-scale,automated and intelligent,which will inevitably lead to more complexity of the operating conditions for the machinery and equipment.Therefore,the test signal for large rotating machinery is also subject to more interference,which will inevitably lead to the reduction of signal-to-noise ratio(SNR)and the difficulties of the machine fault defect.This paper proposes the L-kurtosis of envelope spectrum.By combining with kurtosis,the weighted kurtogram method is proposed to determine the optimal resonance frequency band.It overcomes the sensitivity problem that kurtosis based methods have with interference impulses.The results show the proposed methods work well in bearing diagnosis.The research contents are listed as follows:Firstly,this dissertation proposes the weighted kurtogram method to detect the optimal resonant frequency band.It overcomes the sensitivity problem that kurtosis based methods have with interference impulses using the kurtosis weighted by L-kurtosis of envelope spectrum.It improves the accuracy of determining the optimal resonance frequency band.Secondly,the weighted kurtogram method is robust to random impulses and effective for the periodic impact fault signal analysis.The performances have been proved by the simulations in different operation conditions.By adjusting the weights,the weighted kurtogram method can also be used to detect impact faults at variable speed.Thirdly,the weighted kurtogram method is applied to bearing fault diagnosis and bearing life prediction well,which proves the proposed method is effective for practical engineering signals.By using the signal demodulated with the proposed method,the impact fault diagnosis can be realized accurately.Finally,this dissertation investigates the online monitoring system of rotating machinery.Two sets of on-line monitoring systems are designed by analyzing the fault characteristics of condenser and steering gear.