Research On Fault Diagnosis Of Gearbox Based On Independent Component Analysis

As an important part of mechanical equipment , the state of the gearbox’s detection andfault diagnosis has very strong practical significance. By doing the experiment to simulate thecommon fault of the gear box, this article uses the independent component analysis (ICA) topretreat the fault signal, using wavelet packet analysis method for feature extraction value,finally using support vector machine (SVM) method to recognize gearbox fault station, andfinally has achieved good results.In recent years, independent component analysis technology has developed rapidly,itprovides a new signal feature extraction separation method,and it is preliminary studied infault diagnosis field.This paper introduces the basic principle of independent componentanalysis and the theory,based on the study of the ICA algorithm, puting forward a newalgorithm as an improved particle swarm algorithm(WCPSO) to optimize the nonlinear ICAalgorithm , and using this algorithm in the separation of the three kinds of the simulationsignals to prove the validity of this algorithm.In practical part , apply ICA method to the laboratory extraction gearbox fault signal,using improved particle swarm algorithm(WCPSO) optimize the nonlinear ICA algorithm,separation results indicated that,decomposition failure information obviously enhanced afterICA, realize primary diagnosis of the gearbox fault signals.And then doing four decomposi-tion by wavelet package for the separated signals, obtain the sixteen frequency band energycharacteristics, then input these feature value to the support vector machine (SVM) for gear-box fault state identification, and the results show that using the proposed method ,the pre-cision of the fault diagnosis is ideal. This paper also use different feature extraction, classifi-cation and methods to get diagnosis results, analysis shows that, the proposed method in thispaper makes the precision of the fault diagnosis improves a lot.