| As a key component of a mechanical system, Gear box influences the performance of thewhole mechanical system definitively by its operation status. Kinds of signal processingmethod and pattern recognition are adopted to Gear box’s condition monitoring whosefunction is discovering the fault, so that maintenance and repairing can be realized to ensurethe device work normally、safely and reliably. Base on the analysis of fault types and faultmechanisms, feature vector of every component are obtained by decomposing the measuredsignal coming from Gear box’s vibration profit from the local wave in this paper. Theclassification of the feature vectors above-mentioned is realized base on method of geneticalgorithm integrated neural network. The fault diagnosis of the Gear box is realized benefitingfrom the work above-mentioned. The main research contents include:(1) Analyzing and elaborating on the fundamental principle of local wave briefly,acquisition signal of Gear box’s vibration in different fault condition is accomplished byemploying the experimental device.(2) This paper employs the Local wave method to deal with the fault of a Gear box anddecomposes the signal sent in normal condition, gear fracture condition, when bearing fault ofinner race or outer ring, the IMF components with rich signal are analyzed by means of powerspectrum analysis, thus the provisional fault diagnosis of Gear box is accomplished.(3) The end effect on decomposing appearing with envelope demodulation method isanalyzed and restraining the end effect is realized by adopting end point symmetric extensionmethod.(4) There is a problem that local minimum will be produced back in the iteration ofpropagation neural network. The application of genetic algorithm optimizes the weight and threshold of propagation neural network depending on its global searching ability. Theoptimization above-mentioned eliminates the disadvantages of BP neural network bydecreasing the iteration number and reserving the accuracy. |
PDF Full Text Request