Abstract: Combining the features of both local mean decomposition (LMD) and blind source separation, an underdetermined blind source separation method based on local mean decomposition is proposed. In this method, the observed signals are decomposed into a series of production functions (PF) by the LMD method, these PF and original observed signals then constitute new observed signals, and they undergo whitening process and joint approximate diagonalization, thus obtaining the estimate of source signals. This method can effectively overcome the deficiencies in the traditional mechanical fault source separation method, i.e. the traditional method is restricted to nongaussian, stationary and mutually independent source signals, and the number of observations is assumed to be more than the number of sources. The simulation result shows that the proposed method is effective, and obtains more satisfactory separation quality than the traditional blind source separation method based on time-frequency distribution, it can effectively process the underdetermined blind source separation of non-stationary signal mixtures. Finally the proposed method is applied to the separation of mixed faults of rolling bearing, and the result further verifies its effectivity.

Key words: Blind source separation, Fault diagnosis, Local mean decomposition (LMD), Underdetermined mixture