Research On Independent Component Analysis Blind Source Separation Algorithm And Its Application In Mechanical Equipment Fault Diagnosis

Blind Source Separation(BSS) developing so far, gets in-depth study on the algorithm,and has applications in many areas. In this paper, based on linear mixed model to study theblind source separation algorithm based on independent component analysis, and applied tothe diagnosis of mechanical failure, and did the work are as follows:1．First introduced the basic theoretical knowledge of the blind source separation, giventhe basic model of blind source separation, and then combined with the basic theoreticalknowledge focused analysis commonly used optimization criterion for blind source separationand optimization algorithms, Finally, the evaluation of blind source separationperformanceindicators.2. Gearbox vibration signal model, discuss the basic principles and basic methods ofmechanical vibration signal characteristics and fault diagnosis. First gear and bearingvibration signal analysis, then introduced the gearbox vibration signal propagation andmeasurement, the last of the traditional fault diagnosis method to elaborate, detailed analysisof the deficiencies of the traditional diagnostic methods, which leads to a modern signalprocessing method in the independent component analysis (ICA) technology for machine faultdiagnosis.3. Studied varying adaptive blind separation algorithm and applied to the gearboxvibration signal separation. And other variable adaptive blind source separation (EquivariantAdaptive Separation via Independence, EASI) algorithm is a neural network-based ICAmethod. The only signal separation fellow in practical applications for this algorithm (allsuper-Gaussian or sub-Gaussian signal) shortcomings, the introduction of an improvedalgorithm, is estimated by the probability density function of non-parametric kernel functionmethod evaluation function estimates, both fellow signal can be separated so that theimproved algorithm but also separation of Hybrid signal (super-sub the Gaussian mixedsignal). Signal simulation analysis of conventional and gearbox vibration signal separationexperiments to improve the effectiveness of the algorithm was verified.4. EASI algorithm anti-jamming capability, the problem of bad separation, in-depthstudy of the EASI algorithm based on blind signal extraction fixed-point algorithm applied tothe gearbox fault diagnosis. Blind signal extraction (Blind Signal Extraction, BSE)fixed-point algorithms to extract the source signal through the multilayer neural network orderfrom the mixed-signal, and mixed-signal tightening. For the accumulated errors caused by the tightening process result in extracting a signal quality gradually decreased this problem, theintroduction of a simple and robust cascade decimation tightening method and theaccumulation of errors can be avoided in the tightening process. Computer simulation gearboxfault diagnosis experiment verified that the algorithm can improve the reliability and accuracyof fault diagnosis, highlighting the useful features of fault diagnosis signal to reduce the effectof interfering signals.