Blind Separation Method of Complex Mechanical Vibration Signals Based on Sparse Coding
There exist many excitation sources in complex mechanical vibration signals,so the source signals are mutually correlated sources,and they are difficult to meet the statistical independent characteristics,which leads to the poor separation effect of the traditional blind source separation method.Therefore,a blind separation method of mechanical vibration signals based on sparse signal coding is proposed.The key to the blind source separation lies in the precise estimation of the mixing matrix.However the presence of relevant components in the sources of the machine severely affects the estimation of the mixing matrix.In the proposed method,the short-time Fourier transform is carried out on the observed signal to increase the signal sparsity.Then,the sparse coding is used to screen the time-frequency observation points which have the strait line clustering feature,and the K-means clustering method is used to find the clustering center.Finally,the proposed screening method is used to find the estimated mixing matrix and reconstruct the source signal.The effectiveness of the proposed method is verified by analyzing the fault data of a reciprocating compressor.
vibration and waveblind source separationrelated sourcesparse codingstraight line clusteringcompressor fault signal
万方数据
维普
