Research On Methods Of Blind Source Separation Applied In Fault Feature Extraction Of Rotating Machinery

Blind signal separation (BSS) is a new powerful teehnique in modern signal processing. Nowadays, it has been wildly applied in diverse fields such as speech signal processing, mobile communication, biomedical signal proeessing and image processing. As main technology of solving BSS, independent component analysis (ICA) is used to extract the mutual statistical independent sources. Independent component analysis is the approach to estimate original source signals using only the information of the mixed signals observed in each input channel. The thesis will apply ICA method to rotating machinery fault diagnosis. The method is used to extract characteristic frequency to analyze the machine faults, and study the theories of ICA and the application of mechanical vibration signals.In the paper, the mathematical model and principle of ICA are studied, and the different independent criteria and several main algorithms of ICA are discussed. Further, the FastICA algorithm is studied. FastICA is a fast algorithm of ICA, which is based on Fixed-point iteration theory to fix the non-Gaussian maximum. FastICA algorithm parallelly processes a large amount of sample point of received signals via Newton iterative algorithm, and recovers one independent component from the receiving signals one time. Aimed at the convergent speed of FastICA was depend on initial weights, an advanced FastICA was used with imported looseness agent,and reduced the dependence on initial weights.The thesis conducts a study of the question that the number of observed signals and source signals is different by the simulation, under the condition not satisfied with the separation requirements, a blind separation method based on the narrow band processing is proposed. The simulation results show that the improved method can recover the source signals under the condition not satisfied with the separation requirements.In the last part, two experiments are done on the eccentric rotor and the gearbox, and use the blind separation method based on the narrow band processing to analyze experimental data, the method proposed is used to analyze experiment results in order to extract the corresponding characteristic frequency and then analyze faults of the rotating machinery.