Research On Diagnosis Method Of Remanufacturing Automatic Transmission

With the rapid development of the automobile industry and automatic transmissionloading rates, the retirement capacity of automatic transmission is growing rapidly. Theremanufacturing is a preferred choice for all kinds of methods due to its advantage of energyand material conservation.Since the raw material of remanufacturing process is the scrap gearbox, the qualityfluctuations is unavoidable. The fault diagnosis is designed in this paper, data is gatheredfrom factory, experiment table in order to study on parameter target, Fourier Transform,Wavelet Transform and neural network. Based on the analysis of remanufacturing studystatus and fault diagnosis method at home and abroad, it includes four parts as follows:1、By the relevant investigation on automatic transmission remanufacturing process,there goes a detailed analysis of the transmission working principle and the cause of damage.Design experiment plan, measure the vibration signal of remanufacturing automatictransmission in nomal work conditions, shifts gears impact, slip, lock moment insufficient ofhydraulic torque converter. The corresponding time domain and frequency domain featuresare extracted by methods.2、The feature extraction of remanufacturing gearbox is by wavelet analysis. The basictheory of wavelet analysis and wavelet packets analysis are introduced, a feature extractionmethod is based on wavelet packet energy in this paper. The failure frequency components areanalysed by the wavelet theory which it can capture the short-term impact and transientcharacteristics with the help of time-frequency feature.3、Based on the analysis of basic fault diagnosis process and method combine with thecharacter of neural network, it is used for fault diagnosis of remanufacturing automatictransmission in this paper. By taking five time domain characteristic parameter (margin index,skewness, the kurtosis, the peak value and the pulse target)as the fault vector, it isestablished the neural network of fault diagnosis based on the vibration signal by the use ofMATLAB neural network toolbox.4、Aimed at the problems of BP neural network easily to fall into the minimum point,slow convergence rate,this paper use the improvement algorithm based on value optimize andused two time multinomial approximation determination concealed level nodal point numberin the network architecture design. Neural network which establishes based on the above improvement and the optimized algorithm can make quick and accurate judgment for theautomatic gear box’s working condition.