SECOND ORDER BLIND IDENTIFICATION OF SINGLE-CHANNEL SIGNAL BASED ON KERNELS

Weight-adjusted variant second order blind identification（ WASOBI） algorithm had been used in fault diagnosis field,but it couldn’t separate source signal when the observation signal’s dimension was insufficient. This paper combined kernels with this algorithm to achieve underdetermined blind source separation,and applied it to multi-fault diagnosis. First,the collected single-channel signal was decomposed into multi-dimensional signal by kernels. Then the K-SVD source number estimation algorithm was used to estimate the number of sources. According to the estimated result,a positive definite matrix was reconstructed. Finally,weight-adjusted variant second order blind identification algorithm was applied to separate the source signals. The experiment results indicate that it can well diagnose the multi-fault on rolling bearing with single-channel mechanical signal,feasibility of this method is also validated