The Research And Implementation Of Model-based Fault Diagnosis System

The traditional diagnosis is based on experience, this kind of diagnosis mainly depende on the experience accumulation from domain expert, and has the greatly strengthened dependency on experience, lacks description knowledge of the system structure. Model-based diagnosis is a thought breakthrough in the intelligent diagnosis, it overcomes the shortcoming of traditional diagnosis method.What the model-based diagnosis is composed are: model building, conflict recognition, diagnostic generation, diagnostic test. In this paper after doing in-depth study for these processes, generate a diagnosis system which can be applied in actual operation. The processes can also be considered as that to build models, to compute minimal conflict set, and to compute minimal hitting sets.Firstly, the research of building model, here using a first-order language named AD2L to create the model of actual system. AD2L language has a very strong capability of model description, it can definitely or qualitatively descripte the function of a physical component, it can also definite a new type or recursive functions and so on. These features make the AD2L language has a wide range of modeling, and well flexibility, could be applied to many fields to improve the practical value of studying diagnosis system. Making the diagnosis system in this paper can be widely used.Secondly, for the research of conflict recognition and diagnostic generation, the work in this paper is on the basis of the previour research, doing some improvements of the already corresponding algorithm to reduce time-complexity and space-complexity of the algorithm, and improves the actual efficiency of diagnostic systems. On the research of solving conflicting set, we reference the constraint propagation method in GDE system, which introduces the concept of the environment, using the conection between components, great enhanced computational efficiency. On the research of solving hitting set, we use the improved HS-dag algorithm, which would generate less nodes, high efficiency, and will not lose the true solution. Meanwhile, the article also introduces some other algorithms about computing the conflicting set and hitting set.Finally, we use digital circuits (such as full adder circuit) as an example, giviing the detailed process of the modeling, conflict identification, diagnosis generation. Experimental result shows that the diagnosis system studied in this paper is correct, reliable, and has device independence. But now the system is not perfect and need further improvement, the model-based diagnosis of the application of the theory in practice will play a role in pushing forward.