Study Of Signed Directed Graph Fault Diagnosis Method Based On Neighborhood Rough Set

Signed Directed Graph (SDG) is a kind of qualitative fault diagnosis method, which can explain the fault propagation paths and has a good completeness of diagnosis result. But SDG ignores a lot of quantitative information, leading to the resolution of the diagnosis is not high, and the information redundancy increases the amount of calculation; in addition, estimates of the unobservable nodes based on fuzzy probability SDG are not accurate enough. For these reasons, this paper puts forward the SDG fault diagnosis method based on neighborhood rough set and the improved fuzzy probability fault diagnosis based on SDG method.Granular Computing (GrC) reduction algorithm can obtain the attribute reduction effectively according to the importance of the attributes and eliminate redundant information. So the reasoning way based on granular computing combined with SDG fault diagnosis has reduced the amount of calculation. But most of the granular computing reduction method can only be used for processing the discrete data, and for the continuous data, it is processed with the discrete way, as a result, some quantitative information is lost. It is not conducive to the improvement of resolution of the diagnosis result and the rapidity of diagnosis process. Neighborhood Rough Set (NRS) is able to deal with numeric attributes directly, without discretization processing on it. This paper proposes a SDG fault diagnosis method based on the NRS. Firstly, this method uses SDG to establish each fault propagation path. Secondly, it blurs the nodes on the paths which can not only indicate the degree of node deviation, but also make the data have the characteristics of continuity. And then use the NRS to do the attribute reduction for the decision table of fault diagnosis. Lastly, obtain the decision table after the attributes reduction. When a fault occurs, use the real-time collected sample data to compare similarity with the data in the decision table which has done attribute reduction, judge whether the case fault occurs.When the fault rules library of the SDG fault diagnosis method based on neighborhood rough set is not complete, the improved fuzzy probability fault diagnosis based on SDG method is applied to do the fault diagnosis. This paper introduces a constraint of fault contribution weights to express the different magnitude of the child node’s contribution to the parent node.The method plays an important role in estimating the unobservable nodes.Finally, Tennessee Eastman (TE) process is used as an example for simulation analysis, to validate the feasibility and the accuracy of the diagnosis models for fault diagnosis presented in this paper.