Research On Uncertainty Reasoning Algorithms

With the development of modern production and improvement of science and technique, the function of modern devices has been more and more complicated. If failures happened, it would lead to serious damage and even economic loss without finding and disposing failures rapidly and exactly. To reduce the negative effect of such failures, it not only requires people with high technique, but also requires more exact and intelligent modern diagnosis devices. So the researches of reasoning machine based on artificial intelligence become more and more important. Recent years, the research of uncertainty reasoning as the core of artificial intelligence reasoning machine has been continuous developed and innovated.This thesis expounds the basic concepts of uncertainty reasoning and chiefly illustrates the main point—the research of uncertainty reasoning algorithms, including probability reasoning, evidence reasoning and fuzzy reasoning. It analyzes these uncertainty reasoning methods and simulate their characteristics on MATLAB. These classical uncertainty reasoning algorithms has been validated and simulated, and the simulation results show the restrictions and proper conditions when using these algorithms to design uncertainty reasoning machines, which provides basic theories and experimental evidences for uncertainty reasoning. According to the literature and experiments of uncertainty reasoning methods, the thesis takes use of Visual C++ to program the algorithms with DLL and extends the access in reasoning system and discusses the extended application.Then, a reasoning machine of missile fault diagnosis system has been designed and programmed in the DreamWeaver MX 2004 development with ASP and JSP programming language, based on the fuzzy reasoning and fuzzy matching principle. The system can identify the related knowledge in the database with fuzzy theory and reasoning the correct results in LAN.