Wavelet And Wavelet Neural Network-based Locomotive Traction Motor Gear Fault Diagnosis System

Traction motor is a nuclear part of electric locomotive, whose normal work directlyinfluences the railway traffic safety. Due to poor working conditions, frequent load changes,space limitation and the dynamic action and so on, the gear of traction motor often havesfailures, such as pitting, wear, broken teeth, shedding, etc. Hence, it is necessary to make aanalysis of condition monitoring and fault diagnosis for traction motor.The conventional thinking of gear detection and fault diagnosis is to monitor and judgethe gear fault type according to the research of the vibration or noise spectrum. Based on thespecial position of the locomotive traction motor gear box and the characteristics of enclosurespace, the diagnosis research of traction motor gear fault is carried out according to theanalysis of the stator current signal.Stator current contains a wealth of information of the motor running, and the breakdownsuch as motor gear pitting, wear, broken teeth, fallen off and so on, will cause a newfrequency component in the stator current. The effective information can be obtained bywavelet analysis. The main advantage of the stator current detection method is that it is moreeasily to extract the fault signal compared to the vibration methods and the vibration signalsoften contain a variety of interference noise, while the stator current is independent to theenvironmental impact of motor work which is very important for the extraction of usefulsignals.This paper takes our country freight mainstream traction type and harmonious numberHXD1B locomotive asynchronous traction motor YQ1633as the research object. Selectingthe methods of wavelet transform and wavelet neural network as the main analysis tools,motor current signals were collected by the Hall elements and made this initial noisereduction. In addition, the signals were decomposed using wavelet processing in order toextract the useful time-domain signal. Experiments were carried out to analyze the inputcharacteristic parameters of neural network repeatedly and selecting kurtosis and signal poweras input characteristics quantity of neural network, the types of motor gear fault werediagnosed by the wavelet neural network. The paper built system fault diagnosis platform anddesigned GUI application interface with MatlabAfter the debugging of QPZZ-II gear fault simulation experiment platform for the systemand the online actual operation test of locomotive, it was effective to detect traction motorgear fault with locomotive traction motor gear fault diagnosis system based on the analysis ofstator main current, which provides guarantee for the safe operation of locomotive.