| Using computer technology, neural networks, expert systems, the fault of the engine to make accurate judgment timely, to improve vehicle efficiency and quality of maintenance is very important. With the widespread use of automobiles, Research of automobile engine fault diagnosis system is of great significance.The Fault Diagnostician System of Automobile Engine is a combination of Neural Networks and expert systems, which overcome the traditional shortcomings of expert system and neural network of their existence. So the expert system and neural network put together and can be better applied to fault diagnosis field. The expert system contains the five components which include a knowledge base, database, consequence engine, knowledge acquisition and interpret part. The System developed and accomplished based on Windows XP platform, used MATLAB for the data-processing software and Visual C++7.1 of object-oriented programming language for programming languages and Access for database language.Narrated the four kinds of interconnections pattern and the three kind of learning algorithm of the neural network. This paper introduces artificial neural networks, MATLAB neural network toolbox, database connectivity, the calling method of Visual C + + in MATLAB, and the calling method design of MATLAB calculation engine in Visual C + +.This paper take exhaust gas (CO, CO2, NOx, HC, and O2) for the training samples in order to make the final sample pretreatment Based EQ6102 gasoline engine in the different load condition speed without their emissions, then create the trained neural network based reusable training samples, Finally, test the state of engine cylinder using the trained neural network so as to judge whether Cylinder is in the normal condition, and get a specific fault. The design of man-machine interface systems and the users realize the numerical value of import and export based Visual Programming Language C++7.1,and simulation of the experimental results based MATLAB give full play to the interpreter.Based on BP network, network-based Elman, PNN probabilistic neural network-based and network-based RBF four kind of engine fault diagnosis system for the training; take three samples of fault diagnosis system for the certification; It also conducted a comparison of several diagnostic methods, the results shows, the correct judgment rate of the network is 100%, the network is fully able to meet the requirements of automobile engine fault diagnosis; Elman network identification error is big, but it will not affect the practical application. In addition, because of the introduction of a feedback network in Elman network, the error curves of network training is smoother than the BP network. |
