The Design Of GrC&SDG Based Real-Time Fault Predictive Diagnosis Expert System

Fault Diagnosis Method Based on Signed Directed Graph(SDG) is an important branch in qualitative fault diagnosis field.Researches indicate that the method based on SDG has following advantages. First, SDG model excels in completeness. SDG model contains the whole information in a system, could enumerate all possible faults situation. Second, SDG model can be used for safety assessment and operating guidance, can give fault propagation and provide explanation of faults evolution.Thus, fault diagnosis method based on SDG has a good research prospect. But it also has some drawbacks. First, the process of building SDG Model in Large-scale complex systems is cumbersome. Second, the model build is complex. Third, existing SDG reasoning used graphic reasoning algorithm.It may obtain multiple faults solution, and have low reasoning rate and low fault resolution.Granular Computing(GrC) has a great research value in intelligent information processing field. It simulates the idea of granulation and hierarchical in human thinking and problem solving, to solve complex problems in reality. It has become a effective tool for massive data mining and intelligent information processing. GrC can solved knowledge acquisition, expression, reasoning and other issues in a fault diagnosis expert system.This thesis combines the advantages of SDG and GrC,and proposes a real-time fault predictive diagnosis method. This method use granular knowledge reduction algorithm to reduct the full-nodes decision table, get the core nodes decision table. And use granular similarity calculation, compared the granular samples collected in real time with the granular in the granular library built by decision table to judge the possibility of fault case may occure.Contrast with the single SDG diagnosis method, the method increased diagnosis resolution and reduced complexity of inference algorithm,to some extent.On the basis of the method, proposed the actual and simulation framework for GrC&SDG based real-time fault predictive diagnosis system. And developed a real-time fault predictive diagnosis simulation test system based on configuration software. Finally, the practicality and effectiveness of the method proposed and the system built is proved by instances in pump liquid level simulation system.This paper is divided into three major parts. Specific details are as follows1. On the basis of in-depth study the theory of SDG and GrC, combine their advantage, proposed a real-time fault predictive diagnosis method based on GrC&SDG.The method is applied to SDG model safety evaluation system for the rules of fault cases to ensure completeness of the rules obtained, the formation of the full-nodes decision table. Then use granular knowledge reduction algorithm of GrC to reduct the full-nodes decision table, get the core nodes decision table of system. Create the granular library of two decision tables, then do granular similarity calculation between granular of samples collected from the scene in real time with granular of granular library, and come to the possible degree of every fault case, then sort the granular by their possible degree,and then come to the fault case most likely occurred.2. On the basis of proposed the method above, combined expert system and related technologies as configuration, simulation in industrial control domain. Proposed the system framework of GrC & SDG based real time fault predictive diagnosis, and build its simulation system in software KingView and KingAct.3.Do fault simulation test in pump liquid level system built in simulation system, the results show the practicality and effectiveness of the method proposed and the system built.