Automotive Electronic Control Engine Fault Diagnosis Based On Neural Network Research

The thesis introduces the process of the equipment developing and analyses the direction of the theory research on Automobile Fault Diagnosis(AFD),based on a large of the abroad and domestic information. First the purpose and significance about this subject is discussed and it is put forward that the theory research is an important base on automobile fault diagnosis system developing. Second, the relations of laws and properties between one vehicle fault and the same type vehicle fault are discussed. The diagnosis tactics and the evaluating methods are given. Third, the electrical control technique is used wildly in automobile engineering field, so that automobile performances are improved. But the difficult degree about AFD is also increased now and AFD becomes more and more complicated. There are some limitations for solving these problems by the Expert System(ES) based on the symbol study and inference. Therefore, some numerical value analysis methods are researched. For example Artificial Neural Network(ANN) and Fuzzy RecognitionTechnique(FRT). Because the pattern recognition technique used by computer is a mature method. It is an important base theory for the Automobile Electrical Control system Fault Diagnosis(AECSFD).According to the characteristic of AECSFD and the demand of the Fault Pattern Recognition(FPR),the principle of selection and extractation on automobile technology condition features and the main pattern recognition methods are discussed. The typical faults of the automobile electrical control system are taken for example, and their technology condition features are described. ANN is used for the fault pattern recognition.Because the construction of electrical control system is complex, the fault symptoms have the basic feature that one fault is resulted from many reasons. Therefore, the basic fault pattern is established by the fault symptom and Neural Network is trained by the sample or sample composed of the single fault pattern, so that it can discern the reasons and positions of the different faults. MATLAB programs are used for the neural network's training and to recognize the reasons of the unstable the unstable idle speed on the electrical control engine.