| At present, the amount of our country's private car is rapidly increased, but in relatively, the development of the automotive maintenance industry appears to lag, and now the technology used by cars is updating quickly, that makes most of the world's automobile manufactures have also increased their fault diagnosis system for automotive R&D investment, so that the establishment of an effective automotive diagnosis system can not only bring convenience for automotive maintenance, but also, on the other hand, promoting the development of the automotive industry. Papers introduce the existing technologies, and through exploring two aspects of knowledge, including neural network and expert system, combined with my understanding of the knowledge base, set up a prototype automotive fault diagnosis system to achieve the basic reasoning functions.Firstly, through look through the related books, papers made an introduction to the history about the development of the automotive fault diagnosis system, and then developed the content of my papers wanted to research and the ultimate goal of establishing an automotive fault diagnosis system's prototype. In additional, it also explored the meaning of our study. We also select the appropriate development tools and verify its feasibility on the base of existing technologies.Then, the paper through the introduction of neural networks to explore its role in the field of artificial intelligence, which found that the neural network has the ability to learn from the environment and adapt to the environment in a similar manner of biological interactions, and that is very suitable for fault diagnosis system. The feature also makes up for the lack of the reasoning ability of expert system. And there is a detailed analysis of the BP neural network; it is as a feed-forward type network has a very strong pattern recognition and classification capabilities, and is suitable for building a network-based fault diagnosis expert system. It can be a good combination of kinds of fault conditions of BP neural network training, and using this trained network to process the fault's data in order to achieve the detail information about that fault and provide suggestions to vehicle owners or maintenance personnel to complete the maintenance. At the last paragraph of chapter II, the article implemented the BP neural algorithm by MATLAB and C++ language, and simulated to achieve its role in the automotive fault diagnosis expert system.Next, it made an analysis about expert system. As an important branch of artificial intelligence, the article first has done a presentation on the overall structure of the expert system and a detailed analysis of the knowledge base and inference engine these two core components in expert system, and introduced their function and the special role in expert systems. The following is the design of our knowledge base. Combined with the car fault diagnosis expert system's some functional point, I designed the data storage format, that is, designed the relevant forms used in the expert system, and achieved them by SQL Server 2000. Then the reasoning of the inference mechanism was discussed in light of the designed knowledge base, put forward the design of expert system reasoning mechanism and the points that the inference engine and BP neural network should be linked to improve the expert system's reasoning ability.The final chapter, papers introduced the system environment about the entire system prototype's development. We choose to use C++ as the programming language, it is not only an object-oriented language, but also provides a mature library system. In the human-computer interaction interface, we use the MFC to take full advantage of the C++ class library, making our programming workload of the interface turn be much simpler, and we only need to consider the relationship between windows and the trigger event's coding, as well as the interaction with the database. In the following's system implementation, we have highlighted three functions'achievement, including the user management, knowledge base management and inference engine's implementation, and the one knowledge base management was made of rule-base management and fact-base management. Through the knowledge base management system, we can operate the knowledge stored in the knowledge base, such as made knowledge base perfect and so on. The realization of the inference engine is designed in accordance with chapter III to achieve the reasoning mechanisms, and then added a BP neural network to further enhance reasoning abilities.These are the entire paper's broad framework, and it is worth to note that, due to lack of the necessary expert knowledge, the system is only achieved with the engine as represented by the prototype, and the data is drawn from my summary, and is purely used to do experiments. That may affect the man-machine interface because it is not professional. On the other hand, because the reasoning mechanism is composed of two parts, in order to make those clear that I implement them separately, but in reality we can combine them to hide the implementation details. Therefore, this thesis is based on Visual C++ 6.0 as the main interface's development tool, SQL Server 2000 as the knowledge base's development tool, the engine failure as the main entry point, and achieve the automotive fault diagnosis system's approximate framework on a single PC, in order to become familiar with the development process of expert system and raise my own view.It is known that the research about automotive fault diagnosis system in our country is still in its infancy, and in the usual expert system development process, companies usually establish an expert system's prototype with university or research institutions together. This can get more experience in developing expert system through the development of prototypes, resulting in subsequent development process detours to develop a more competitive system. I hope that through the understanding of the expert system to get perfect the expert knowledge, and in the future I can be able to complete the entire system, to make a truly realistic automotive fault diagnosis expert system. In another aspect, I hope, in the future's study, we can come over the shortage of the related algorithms or design new neural algorithm to integrate with the expert system, in order to make diagnosis more accurate and improve system availability and accuracy in another hand.
|
PDF Full Text Request