Bearing Single And Compound Fault Diagnosis Based On Singular Spectrum And Deep Belief Network

Under the complicated working environment,the weak fault feature of the bearing is difficult to extract under the interference of noise,which brings difficulties to the diagnosis.In addition,when the compound faults are generated,the faults of different positions are coupled with each other,which is a challenge to the fault diagnosis.According to the characteristics of bearing single fault and compound fault,the vibration signal of rolling bearing is taken as the research object,and the method based on singular spectrum and deep belief network is studies.Firstly,a new metric-chaotic singular spectrum is studies,and the feature space and noise platform of singular spectrum are explained from the perspective of mathematics Then,from the perspective of geometric space,it is proved that the singular spectrum is a description of the spatial geometry based on variance maximization,which shows that the chaotic singular spectrum can extract the characteristic signal with low signal-to-noise ratio and has strong anti-noise ability.The stability and anti-noise ability of chaotic singular spectrum are analyzed by the typical chaotic system.The vibration signal was analyzed by the chaotic singular spectrum to further verify the validity and practicability of the method.Secondly,the deep belief network(DBN)is studied.Combined DBN with variational mode decomposition sample entropy,bearing fault diagnosis experiments are carried out which verifies it has advantages when there are few samples or few label samples.Thirdly,a compound fault diagnosis method based on singular spectrum decomposition(SSD)and independent component analysis(ICA)is proposed.The SSD constructs a trajectory matrix and adaptively selects the length of the embedding dimension,and is used to decompose the non-stationary fault vibration signal into several singular spectrum components with different frequency bands.Then,the valuable singular spectral components and original observation signal form new observation signals.Furthermore,ICA is introduced to perform blind source separation on the new observation signals.Finally,simulation examples verify the effectiveness of the method.At last,for a single fault,diagnostic method based on chaos singular spectrum and deep belief network is used to diagnose and classify 10 different states of vibration signals,and the result verifies the feasibility of the method.For the compound fault,SSD and ICA are used to deal with the compound faults.Take the compound fault of the inner outer rings as an example,the method can realize the extraction of different fault feature information.The experimental results show that this method can diagnose bearing compound fault.