Feature Extraction Method Of Equipment Dynamic Signal Based On Blind Searation Theory

The mechanical equipment is usually in a bad environment,the signal collected from the sensor is a mixed signal containing other noises,and the effective fault features are often submerged,making the fault difficult to diagnose and identify.In this paper,based on the theory of blind source separation,the feature extraction method of vibration signal of mechanical equipment is studied.First introduced the theory of blind source separation(BSS),it is completely unknown in the source case,through independent component analysis(ICA)algorithm the source signals from the mixed signals.Secondly,the ICA and the constrained independent component analysis(c ICA)algorithmare introduced.The characteristics of ICA and c ICA are studied by numerical simulation.The experiments show that c ICAalgorithm combines the ICA algorithm with the natural characteristics of the research object,which is better than ICA algorithm.In addition,the construction method of the reference signal for the c ICA algorithm is discussed,and the influence of the parameters of the reference signal on the separation of the observed signal is analyzed.In order to solve the single channel signal using BSS algorithm of underdetermined problem of signal decomposition method of empirical mode decomposition and empirical mode decomposition,the overall frequency slice wavelet transform and variational modal decomposition,the results show that the VMD method can be adaptive with single component decomposition for multiple fixed mode component,high accuracy and less decomposition decomposition,which can effectively solve the modal aliasing problem,enhance the ability of feature recognition.In order to solve the problem of single channel blind source separation,a method of source number estimation based on singular value decomposition is proposed.Then,the optimal component is selected by the Mahalanobis distance between the original signal and the decomposed component.Experimental results show that this method can effectively eliminate the interference between signals and signals.A single channel blind source separation method based on c ICA is studied in this paper.The algorithm is used to decompose the single channel observation signal by VMD.Study on the separation method of envelope c ICA based on single channel blind source,after the VMD decomposition of the signal,the signal vector observation to extract the envelope of each component as the input of c ICA,so as to isolate the fault characteristics of the ideal equipment information.The two methods are applied to the fault diagnosis of rolling bearing,which can effectively extract the fault characteristics of rolling bearing,but the c ICA single channel blind source separation method based on component envelope is more effective.Single channel blind source separation method using only a single sensor signal acquisition,ignoring the differences between the sensors,signals from multiple sensors can provide abundant information,is conducive to the separation of more effective fault feature.A blind source separation method of c ICA based on dual channel signal,after VMD decomposition in each channel signal,signal vector observation to extract the envelope of each component as the input of c ICA,thus separating equipment fault information.It is applied to the fault diagnosis of generator bearing looseness and locomotive bearing fault diagnosis.