Fault Feature Extraction Of Rotating Machinery Based On ICA-R And Stochastic Resonance

As the core components of mechanical equipment,rot ating machinery are widely used in the aviation,transportation,metallurgy and other fields.The requirements for safety and reliability of rotating machinery are getting higher and higher,therefore,the state monitoring and fast fault diagnosis of rotat ing machinery becomes more necessary and urgent.Rotating machine fault diagnosis uses signal processing methods to find useful information to determine the operation of mechanical equipment through analyzing the vibration signal analysis.Rotary machinery’s work environment is complex.Background noise is strong.There are multiple fault concurrent conditions.The collected observation signal is a mixture of multiple source signals.How to separate and extract the fault feature from the mixed signal with strong noise and multiple faults is the main problem to be solved.Aiming at the problem that it is difficult to extract the fault feature of the rotating machinery with the multi-fault concurrent,strong noise and low signal to noise ratio,a method of fault feature extraction based on ICA-R and stochastic resonance is proposed.The effectiveness of the method is proved by simulation and experimental analysis.With compared with the method based on FASTICA and stochastic resonance,it is proved that the m ethod based on ICA-R and stochastic resonance is superior to the hybrid fault diagnosis of rotating machinery.In this paper,the multi-fault mixing of rotating machinery is as the main research contents,the multi-fault hybrid model of rotating machinery is studied,and the fault type,multi-fault hybrid model and the fault model of gear and bearing are studied.According to the multi-fault linear instantaneous mixing model,a hybrid fault simulation signal of rotating machinery is established.The ICA-R algorithm is analyzed,and the improved ICA-R algorithm is proposed.Aiming at the shortcomings of the single fault signal processing method in dealing with the mixed fault signal,a hybrid fault vibration signal preprocessing method based on ICA-R fault signal extraction is proposed,and the effectiveness of the preprocessing method is verified by simulation.Aiming at the limitation of stochastic resonance processing of rotating machinery,the expression of pulse signal in random resonance is studied.By analyzing the time domain analysis method,the evaluation index of pulse signal is constructed.This index is used as the fitness function of the artificial bee colony algorithm.Combining with the variable scale stochastic resonance method,a variable scale adaptive stochastic resonance is proposed,and the effectiveness of the preprocessing method is verified by simulation.Based on the advantages of ICA-R and stochastic resonance,the fault signal extraction based on ICA-R is applied to the mixed signal preprocessing to realize the extraction of the expected fault signal.The fault feature extraction of the expected fault signal is carried out by using the variable scale adaptive stochastic resonance.The validity of the method is verified by simulation.We test and analyze the different fault types of mixed fault signal on the experimental platform of gearbox dynamic simulation system,using the method based on FASTICA and stochastic resonance and the method based on ICA-R and stochastic resonance.The results show that the method based on ICA-R and stochastic resonance can effectively extract the fault feature of rotating machinery.