Noise Diagnosis Research On Gearbox Failure Of Loader Based On ICA And SVM Algorithm

Loader is a kind of construction machinery widely used in highway,railway,port and other construction projects.Its use has greatly improved production efficiency.However,because of its complex structure and high correlation between parts,some parts often appear,especially the failure of the gearbox,which leads to the frequent failure of the whole equipment.If the failure can not be diagnosed in time,it will cause property loss and heavy casualties.The general fault diagnosis of loader gearbox uses accelerometer to measure vibration signal as information source.For different faults or different gearboxes,the sensor needs to be rearranged.The operation is time-consuming and troublesome.Therefore,it is of great practical significance to study an efficient gearbox fault diagnosis method.In this paperusing the non contact of noise diagnosis and the sound intensity sensor,the noise of the loader gearbox under different working conditions is taken as the object of study.A noise diagnosis method is proposed to identify the fault type quickly,so as to escort the normal operation of the gearbox of loaders.Because of the complexity of the noise mixing in the loader gearbox,it is difficult to extract the characteristic signal of different working conditions.By using the principle of ICA and correlation coefficient,the independent component of the gearbox noise under different working conditions is separated and the maximum value of the correlation coefficient of the independent component and the noise data of each channel is calculated.The characteristic vectors of the loader gearbox are constructed in different working conditions.At the same time,the excellent generalization ability of the small sample set is processed by using the SVM algorithm.The Gauss kernel function and the "one to one" method in the SVM processing of multi classification problems are used to train the fault diagnosis classifier to realize the efficient diagnosis of the loader gearbox failure.In MATLAB,four source signals of the loader gearbox are simulated with the method of replacing noise with vibration,and the convolution mixed model is used to simulate 120 groups of 6 channel vibration signals of the gearbox under four operating conditions.The frequency domain ICA algorithm is used to separate the independent components of the four types of working conditions and calculate the characteristic vectors of the gearbox under different working conditions.400 sets of data are used to train 6 class fault diagnosis classifier,and the remaining 80 is used to test the error rate of the classifier.The results show that in the 80 sets of data test,the number of false diagnosis is 5,the error rate is 6.25%,indicating the practical usability of the algorithm.According to the national standard,the 13 point test platform of the loader transmission noise test is set up.The noise data of the gearbox under normal condition and fault condition are collected by the sound intensity sensor,and the data are analyzed by the algorithm.The experimental results show that in 40 sets of test data,the number of errors is 4,the error rate is 10%,verify the feasibility of the algorithm.