| With the rapid development of electronic information technology in automobile, the structure of modern autos becomes more and more complex and high-tech. Auto fault diagnosis is imperative nowadays. Since engine is the power source of an automobile, its fault diagnosis and maintenance is extremely important. Statistics show that45%of engine failure originates from ignition system, so whether the ignition system can function smoothly will directly effects the running of an engine. Therefore it's fair to conclude that the fault diagnosis of the ignition system is of enormous importance.This thesis mainly focuses on electronic fault diagnosis method of the electrical ignition system. Firstly, it discusses the common malfunctions and diagnosis methods. Currently, the primary way to diagnose fault is using digital multimeter, automobile oscilloscope and artificial intelligent method. Compared with the advantages and disadvantages, this thesis proposes an intelligent fault diagnosis method based on GA-BP model. The secondary ignition voltage waveform is the starting point of the method. It will change once the ignition system fails, and these changes in the region reflect the corresponding fault information. This thesis deeply analysis the secondary ignition voltage waveform, and extract the key parameters which reflect the waveform changes as the input of the fault diagnosis model based on GA-BP, according to correspondence, the output is the fault information.Secondly, modeling this fault diagnosis model. The model is composed of data collection, data processing, BP neural network, GA algorithm and fault conclusions. Automobile ignition oscilloscope works as collection unit for collecting data of fault waves. Data processing unit normalizes the data. BP neural network takes charge of model recognition and makes fault conclusions.For disadvantages of BP neural network structure, the output error become larger, and the diagnosis required a longer time, when the initial parameters of improper selection. In this thesis, genetic algorithm is adopted to optimize BP neural network, which reduces error and improves efficiency of the fault diagnosis model.Finally, simulation analysis of GA-BP model has been performed through MATLAB. The engine of a 'F23A3' car is the study object of this thesis and different rotational speed data have been prepared for diagnosis analysis. Result of simulation analysis shows that BP neural network model reaches convergence with239steps, with an average error of2.29%.While GA-BP model reaches convergence with68steps, with an average error of0.17%. By comparison, GA-BP model can automatically identify ignition system fault and improve efficiency and accuracy, therefore, it's a better solution. |
