Research On Open Fault Of Drive System In Permanent Magnet Motor For Electric Vehicles

Since1990s, with the improvement of people’s consumption level, the increasingmaturity of automotive technology and the reducing of traditional internal combustion enginevehicles’ costs, car is becoming the most important and popular means of humans’ transport,its holdings is rising sharply worldwide. At the same time, the rapid development of theautomobile industry has accelerated the consumption of non-renewable resources andenvironmental pollution. Therefore, countries are putting focus on the study of electricvehicles. As such the fault diagnosis of drive systems for electric vehicle are becoming moreand more meaningful.This paper mainly research on the open fault of the motor controller which is a corecomponent of the drive system. Propose a method for detecting the open faults of MOSFET inthe motor controller and verify the feasibility of this method by experiments.This method is based on detecting the DC side current of the motor controller, alsoreferred to as the bus current. Firstly, build the equivalent circuit model of permanent magnetbrushless DC motor by analyzing and summarizing the electric vehicle drive system.Operating state of the drive system is divided into two according to the modulation scheme ofinverter circuit and motor theory. Establish bus current contact with the inside of the motor byparsing each of these two states. Secondly, collect data when drive system work properly andthe three MOSFETs of inverter’s up bridge in the open state respectively. Thirdly, dataprocessing and fault diagnosis. Decompose the bus current in three layers by using waveletpacket decomposition technique. The bus current signal is decomposed into eight sub-bands.The bandwidth of the eight sub-bands are equal. Then reconstruct to the original signal.Calculate the energy value for each sub-band, the eight energy value form a8-dimensionalfeature vector. Finally, establish the BP neural network, and Combine the feature vectors andneural network Loosely. Feature vector as the input of BP neural network. To identify faultsby the adaptive, non-linear characteristics of the neural network. In actual processing, first totrain the neural network with a certain amount of samples. When the training is completed,and then test the neural network with the test samples. The tests proved that this method canbe used to detect open circuit faults of the MOSFET in the motor controllers.