Fault Model Reduction And SDG Reasoning Method Based On Granularity Entropy

With the development of science and technology and the expanding size of industrial systems, the relationship between different variables had been increasingly complex. Fault of one component is likely to cause a series of accidents, even lead to paralysis of the entire system, resulting in economic losses and casualties. Therefore, detecting the fault of system, finding and dealing with the fault timely are particularly important.Fault diagnosis can be divided into two types:methods based on models and methods based on historical data. Both of them can be respectively divided into quantitative methods and qualitative methods. In process industry, the failure of quantitative fault diagnosis method result from that to establish precise quantitative model is difficult because the information is often incomplete, or some information of system cannot be accurately converted to quantitative information.In older to overcome these difficulties, qualitative methods are often used for processing. Fault diagnosis based on SDG model is an important branch of qualitative methods, which need less information to establish a system model that not only can comprehensively reveal the internal relations and links of system variables, but also can clearly reflects the spread path of the fault. So fault diagnosis based on SDG model is an effective method for fault diagnosis. However, as in the method'reasoning a large amount of quantitative information is ignored, the diagnosis resolution is not high. And since this model contains a large number of redundant information, the speed of its reasoning is low and its results are not real-time.Granular Computing (GrC) is an effective method to solve problems and deal with large number of complex information. It can scale the importantance of information in the systems to solve the problems beneficially.In order to solve the problems of SDG fault diagnosis and improve the speed and resolution of fault diagnosis, this article will introduce the idea of Granularity Entropy to the SDG model, proposed simple methods of knowledge based on the knowledge of Granularity Entropy and the fault reasoning method based on the Granularity Entropy.Specific researchs and work in this text are as follows:(1) For diagnosis speed is slow and real-time is bad in the SDG fault diagnosis method, so this paper proposed a knowledge reduction algorithm in incomplete information systems based on granularity entropy from the GrC perspective. This algorithm takes Granularity Entropy as heuristic information, reducing the redundant attributes and attribute-value, and gets the best diagnosis rules set and improves the speed of fault diagnosis.(2) In practical system because of some reasons, attribute-values in information system is often missing, leading to information system incomplete. As the algorithm above is only applicable to complete information system, this paper brings MCC into this algorithm. Before the reduction of attributes and attribute-value, we firstly convert incomplete information systems into complete information systems, in order to deal with incomplete information systems.(3) Because resolution in the algorithm of SDG fault diagnosis is low, this paper proposed SDG fault reasoning method based on granularity entropy. This method takes relative granularity entropy between model points as quantitative information, adds the quantitative information into the SDG model, and sorts multiple failure sources by priority to raise the resolution of fault results.(4) Take Hot Nitric Acid Cooling System for an example, this paper establishd SDG model based on the empirical knowledge, and obtained the decisive information system needed by diagnosis through the SDG model, and verified the feasibility of this method through information processing and fault reasoning for the system.