Study On The Locomotive Diesel Engine Accessories Fault Diagnosis Based On The Abnormal Vibration Signal

Diesel locomotives play an important role in the development of China’s railwayindustry as the mainstream traction power in railway transportation. In order to improve thesafety and reliability of the diesel locomotive power transmission devices, to meet therequirements of modern high-level beforehand maintenance system, condition monitoring andfault diagnosis of traction generator has the profound significance.Generator is a typical reciprocating machinery, and its monitoring and diagnosis are verydifficult because of its complex structure. Its working process is mainly the high-speed rotation,which makes its obvious vibration, and the faults of most rotating machineries are reflected inthe vibration signal. This paper gives the diagnosis method based on vibration signal.This paper is roughly divided into the following three stages:First, according to the generators of three typical faults of generator system:rotorimbalance and the rotor crack and oil whip,we simulate the vibration signal of generator, anduse the technology of wavelet packet time-frequency analysis methods to the collected signalfor noise cancellation, and gains the fault characteristics from various changes in the energy ofeach frequency band.Then, the fault analysis process is determined,that is the BP neural network,which isdesigned by energy feature vector extracted with wavelet packet decomposition. Choosingtraining data and testing data mainly includes the following: first, the signals are decompositedand extracted features,in order to get the sample and the testing features vector; Second,training BP network with sample data, and the network has been tested with testing data afterthe successful training, and the testing results accord with corresponding states, which is provedit’s corresponded. The whole process is realized with MATLAB software.Finally, locomotive traction generator vibration detecting system is designed by VB,which is analyzed in the collected signal. The results proved that the system can diagnosis thefault of the traction generator simply and effectively, and also improve the efficiency andaccuracy of fault diagnosis of generator failure.This study shows that using the method of wavelet packet analysis and BP neuralnetwork analysis in the process of the fault diagnosis of diesel locomotive power transmissionequipments, can guide the on-site technicians finding out the complex mechanical failures torepair efficiently, and also can enrich the fault diagnosis methods and ways of powertransmission equipments.