Research On Modeling Method For Production Inference Fault Diagnosis Based On Discrete Event System

Fault diagnosis is one of the hot research fields at present. With the development of artificial intelligence technology, fault diagnosis expert system based on production system has also been concerned by many researchers. But the production system’s reasoning efficiency is low and it can’t represent knowledge constitutively. Modeling and reasoning by automata is an effective method to overcome the above disadvantages. For more complex production systems, the correctness verification of the logic among invoking rules is also the guarantee of the production system’s normal operation. The correctness verification of the logic among invoking rules can be verified by production system’s automata model because of the logical rigor of automata.In this thesis, a series of methods based on automata are presented. A method based on ordinary automata is presented to model production system. The methods of modeling production system and fuzzy weighted production system based on lattice-valued automata have been put forward. They are summarized as follows.1. A general modeling method of production system based on ordinary automata has been put forward. An inferential model has been structured by combining the features of automata and production inference. The inferential model can be generalized to a general model which is used for production inference of multiple rules. The flow chart of modeling and the simplified method are also presented. Finally, an instance analysis is made. The effectiveness of the method is verified.2. A general modeling method of production system based on deterministic lattice-valued automata has been put forward. This modeling method is a theoretical extension of the method based on ordinary automata. In the succession of the ordinary automata model’s features, series methods based on lattice-valued automata are presented. The method of model simplification is presented. The correctness verification of the logic among invoking rules is presented. Also the model updating method through the rules’ increase and decrease is put forward. Finally, an instance analysis same as ordinary automata’s modeling method is made. The effectiveness of the method is verified, and the advantages over ordinary automata’s modeling method are also verified.3. A general modeling method of fuzzy weighted production system based on fuzzy lattice-valued automata has been put forward. This modeling method is a fuzzy extension of the method based on deterministic lattice-valued automata. The method can process fuzzy reasoning information, and it can describe the reasoning results’ evaluation language by triangular fuzzy numbers. Finally, the method is applied in the fault diagnosis system of power grid. Through the calculation of fuzzy reasoning, the effectiveness of the method is verified.The modeling methods presented in this thesis are all general methods, they can model various production system. The modeling methods can overcome the disadvantages of traditional production system, and also they can be used as an aided decision-making system of fault diagnosis expert system. These modeling methods have a good practical significance and value.