Traction Motor Bearings For SS_7E Electric Locomotive Research On Fault Diagnosis Method

With the continuous development of high-speed,intelligent and informational trains in China,the safety of train operation and passengers’ lives is particularly important,and also represents the economic prosperity of a country to a certain extent.As an important equipment on AC-driven electric locomotive,the operation status of traction motor plays a vital role in the safety of the whole train operation,but the traction motor will gradually appear various failures in the process of long time operation,which directly threatens the safety of train operation.After researching and investigating,it was learned that the overhaul workshop of a railroad bureau locomotive section diagnosed the motor bearing of SS7 E electric locomotive as follows: firstly,the JK00430 walking section detection device on the locomotive issued an alarm,and the TCMS display indicated the traction motor failure.Secondly,the overhaul workshop disassembles all the components of the traction motor and diagnoses all the bearings on the motor,and finds that the faulty bearings are directly replaced with new ones.Such an overhaul process consumes huge manpower and time,and causes waste of resources when replacing the repairable bearings,resulting in unnecessary labor and financial losses.Therefore,it is considered important to design an accurate and intelligent fault bearing diagnosis system.In this paper,the traction motor simulated fault test bench data from the maintenance workshop of a railroad bureau’s machine section is used as the data set for this test,and all the data are extracted centrally,and the extracted vibration signal data are decomposed by CEEMDAN,and the PSO optimized SVM algorithm is used to classify the collected signals after the decomposition is completed.The human-computer interaction interface design is completed by the GUI design of MATLAB software.The system test has a high recognition rate for fault diagnosis and has a large practical application value,and the main research contents are as follows:(1)The collected signal is decomposed by CEEMDAN,and the decomposition results of EMD and CEEMDAN are given respectively.The above two decomposition results are compared to verify that the CEEMDAN decomposition results are more accurate than the EMD decomposition method,and eliminate the mode aliasing phenomenon caused by EMD decomposition;(2)The time domain and frequency domain analysis of the four bearing signals of the bearing without fault,the outer ring fault,the inner ring fault and both inner and outer ring faults are carried out to make sufficient preparation for the next step of extracting the signal features.Considering that the fault in the frequency domain also contains certain oscillation information in its features,a fast Fourier transform is performed to plot the time-frequency domain curve;(3)Extracting the time-frequency domain statistical features of the original and decomposed signals,including the mean,standard deviation,mean squared deviation,maximum and minimum values,skewness,kurtosis,median,information entropy,frequency domain mean,center of gravity frequency,root mean square of frequency,and standard deviation of frequency of the signals,and the set of the above statistics is used as the feature set for signal extraction;(4)The feature set is input to the support vector machine network used to train the network and PSO optimized for SVM fault identification classification test.In addition,DNN,DBN,SVM and ELM algorithms are used to diagnose the same fault data,and the results show that the SVM algorithm based on PSO optimization selected in this paper has better diagnostic effect.(5)GUI of MATLAB is used for the interface design.Controls are added to complete the interface design,and callback functions are written to facilitate the completion of each control function.The main interface mainly includes four modules,signal acquisition and reading module,signal feature extraction module,fault pattern recognition module and diagnosis result analysis module.The four successfully written modules are debugged to confirm the normal operation status of the functions.The designed system is used to diagnose the actual measured data in the field,and the diagnosed results are compared with the simulated data diagnosis to verify the higher diagnostic accuracy and reliability of the designed system.