SDG-based Fault Diagnosis And Application Based On Granular Reasoning Method

Fault diagnosis based on the signed directed graph (SDG) of graph theory is an important research branch in the field of fault diagnosis. There is a good correlation between knowledge representation of SDG and fault characteristics in complex systems.And it can effectively express fault knowledge in complex systems, but also it contains a flexible reasoning way and effective reasoning algorithm. So,the approach of SDG for fault diagnosis in complex systems, is of great theoretical significance and application value for shortening the time to determine the fault, reducing the loss improving the accuracy of maintenance and saving maintenance costs, etc. But there is a lot of problems to be solved: (1)the multivocal reasoning leading that resolution is not high; (2) bad real time , slow diagnosis rapidity when calculating in large systems; (3) not improving its qualitative reasoning and decision-making by making use of quantitative knowledge .Granular Computing (GrC) is a new way to simulate human thinking and to solve complex problems in the current intelligent computing field,it has become an effective tool for solving complex problems, mining massive data and processing fuzzy information. And granular computing can solve the problem that is knowledge acquisition, expression, and reasoning in fault diagnosis. Professor Liu Qing introduced information granule into rough logic, presented granular language, logic and reasoning, that is a research direction of granular computing.To solve the problem of complex deduction in SDG model and low resolution, the reasoning method based on granular computing and SDG-based fault diagnosis are unifyed and a theory on SDG-based fault diagnosis based on reasoning method of granular computing is proposed, which provides a new approach for reasoning of SDG. The detailed method is that granule is used to formally express the elements of SDG model, the sample sets of the fault are obtained according to the experiment and analysis, after that the sample sets of the fault decision is transformed into decision table which reflects the relation of faults and symptoms, then the decision table is transformed into the corresponding granular base, consequently fault source that most likely occure to fault is found by searching granular base and computing the most similarity.And the following work is done:(1) The definition of SDG model based on granule is proposed, granule is used to formally express the elements of SDG model, granule is integrated with the SDG, which is a new description method of SDG model.(2) study exploringly threshold p and validate that its value is appropriate. The method is used in steam boiler system in this paper, and the feasibility is validated on simulating platform"Matlab". The resolution could be improved by the method.