Abstract: Dynamic blind source separation is a focus in the blind source separation of multi-fault. Traditional blind source separation (BSS) is restricted to the stable statistical characteristics and static mixture system, and ignores the sequential information. Based on this deficiency, combining to canonical variate analysis (CVA) and independent component analysis (ICA), a dynamic blind source separation method based on CVA-ICA is proposed. In the proposed method, the source signal is regarded as state variable in the state space, observation signal as output variable, thus the dynamics ICA is transform into the state space ICA. The proposed method employs CVA as a reduction tool to construct a state space, then the statistically independent sources are separated by the conventional ICA algorithm. The simulation results show that the CVA-ICA method is superior to traditional blind source separation in the dynamic blind source separation, and has satisfactory separation performance. The proposed method is applied in blind separation of bearing inner and ball fault, the experiment results further validate the effectiveness of the proposed method.

Key words: canonical variate analysis, dynamic blind source separation, fault diagnosis, independent component analysis