Experiment Research On LCD In The Rotating Machinery Fault Diagnosis

Large rotating machinery is the key equipment in modern enterprise production. Inorder to ensure the safe and reliability operation of large rotating machinery, the faultdiagnosis technology of mechanical equipment has been paid more and more attention. Itis key to extract the fault feature from the vibration signals. Time-frequency analysis isthe usual method for feature extraction. The vibration signals are mostly nonlinear andnon-stationary. Local characteristic-scale decomposition (LCD) is a new adaptivetime-frequency analysis of vibration signals, so that this new method is deserved to befuture studied.Firstly, improved algorithm of LCD achieved in several aspects. Because ofgenerating pseudo components in the iterative process, an improved LCD based onenergy ratio and correlation coefficient for loop-termination criteria was put forward toeliminate undesirable component. Then, the reason of end effect was analyzed in theLCD decomposition process. An improved endpoint LCD based on similar triangle wasproposed. According to the low accuracy of components in LCD decomposition process,an improved LCD based on residue decomposition was put forward. Then, all improvedLCD methods were used to analyze the simulation signal, experimental vibration dataand actual fault signal. Through the above analysis, it is proved the improved LCDmethods are effective.Secondly, the reason of the cross-terms of the Wigner-Ville (WVD) was analyzed. Amethod, then a cross-term erasing method based on LCD was put forward in the paper.The analysis of simulation signal and actual fault signal indicates that this method caneliminate the cross-terms effectively.Lastly, combining the feature of both LCD and blind source separation, anunderdetermined blind source separation method of rotating machinery faults based onLCD was put forward. The analysis of simulation signal shows that this method caneffectively separate the non-stationary signal in which the number of observed signals isless than the number of source signals. While combining that method with the traditional blind source separation, experimental results of the rotor of unbalance-rubbing-loosenesscoupled faults show that the method is effective.