Study On Repeated Blind Separation Of Independent Component Analysis And Its Application In Mechanical Fault Diagnosis

Independent component analysis (ICA) is a new signal processing method which was developed at the end of last century. It is a particular method of Blind Source Separation (BSS), and was widely used in mechanical vibration, signal processing, biomedical signal processing, communication, array signal processing, sound signal processing and other fields, where it gains good effect.The function of ICA is to decompose a linear mixed signal from many sources by the method of optimization algorithm. Based on BBS, the ICA assumes that the source signal is independent from each other, and extract the source signals one by one with specific optimization. ICA evolved by the blind source separation is also known as blind separation. ICA has a good effect in feature extraction, as well as weak signal detection. In this research, the knowledge, principle, application of ICA are illustrated in details to prove the possibility of ICA's function, and its feasibility is verified though idealized simulation.Every method has its advantages and disadvantages, ICA is no exception. The advantages shall be inherited but the disadvantage should be innovative. In accordance with the problem of the ICA, that is, the number of the test channel can not be less than that of the signal source. This research has put forward a Re-ICA method, which was proved to be a successful breakthrough. In this paper, frequency domain analysis is integrated into the ICA which is the most classic method in signal processing. Create a virtual channel according to the frequency domain so that the number of channels will increase. After this step, the independent component analysis will be processed. The nature of this new method is the same as the ICA's, but to some extent, this method overcomes the shortcomings of the ICA. This paper proves the availability of this new method through the simulation and practical application. Of course, this new method has its deficiencies, but these deficiencies show its potential.