The Research Of Fault Diagnosis Based On Knowledge Reduction Of Granular Computing And SDG

When a priori knowledge is not complete, uncertainty, can not establish a quantitative model, or qualitative information of system can not be accurately converted to quantitative values, quantitative-based fault diagnosis can not resolve the situation. The incomplete prior knowledge can use by the qualitative method to describe the system structure and function, and then establish the qualitative model, inference fault reason. Graph theory-based Signed Directed Graph (SDG) is a branch of fault diagnosis, which can reflect the process variables and the relationship between variables, express complex systems of causality, and revealing the fault propagation path. However, there are still many problems of the SDG-based fault diagnosis. Such as the ambiguity in diagnostic reasoning led to the low-resolution; slow diagnosis of large systems, did not use quantitative knowledge, and so on.Granular computing resarch different levels of grain, primarily to handle uncertain, ambiguous, incomplete, and massive information. It is a effective tool of complex problem solving, massive data mining, fuzzy information processing. Knowledge reduction of Granular computing remove redundant attributes without changing the classification ability, redundant nodes of complex systems to improve speed and real-time of fault diagnosis.This article will introduce knowledge reduction of granular computing and SDG-based fault diagnosis, and four problems are studies as below.(1) Knowledge reduction of granular computing and SDG-based fault diagnosis is studied. In complex system, there is lots of nodes in SDG-based fault diagnosis, which produce the combination explosion and diagnosis slow. After reduction of Granual Computing, it can be fast, accurate of fault diagnosis to improve the efficiency.(2) Based on the knowledge reduction of Matrix algorithm, the node importance-based knowledge reducitong algorithm is introduced, whih choose attribute importance as referece. So that it can better identify the minimal relative reduction of attributes.(3) The falut is classified according to the main fault. Judge the main falt first, then judge the other from SDG model.(4) Power Plant Deaerator system for example, introduced the SDG modeling, fault diagnosis decision-making. Then the node importance-based knowledge reducitong algorithm is test in the simulation platform, reduce the main falult diagnosis decision table, simplify the redundant nodes, improve the efficiency of fault diagnosis. Finally the diagnosis rules after the reduction is verified accuracy and validity.