Research And Application Of The Confidence Rule Library Expert System Modeling Method

For comprehensive utilization of various incomplete or uncertain subjective information,a belief rule base(BRB)expert system is proposed to build the modeling of the complex decision-making problems.Compared with the IF-THEN rules,BRB is on the basis of the belief framework,which is used to capture nonlinear causal relationships as well as vagueness and incompleteness.On account of advantages of the knowledge representation and the intelligence,BRB has attracted the attention of many scholars.BRB allows experts to get involved directly.Nevertheless,experts knowledge has the nature of subjectivity.And,most practical engineering problems are more complex.All these issues have brought challenges to set parameters of the belief framework quickly and accurately.In order to solve the above problem,this work gives emphasis on research on modeling methods of belief rule base expert system.The main work of the paper is summarized as follows:(1)Particle swarm intelligent optimization algorithm(PSO)is used to improve the disadvantage that the solution efficiency of BRB optimization model is low.The proposed method is verified by adulteration detection of edible vegetable oil.Compared with the traditional optimization strategy,PSO improves the efficiency of BRB model fourfold.It shows that PSO is an effective method to to increase the efficiency of BRB model.(2)In order to overcome the limitations of expert knowledge and enhance the performance of BRB system,this thesis takes both the referential values of the antecedent attributes and the utilities of the consequent as parameters in numerical BRB model,along with rule parameters.As such,a model,optimizing the structure and parameters of BRB(OSP-BRB),is proposed.At last,an instance of the tipping paper porosity measuring in tobacco factory is introduced to prove the functionality of OSP-BRB.Compared with the method of identifying BRB structure,OSP-BRB reflects the situation of the porosity more actually.It shows that OSP-BRB is a reasonable approach to get the the structure of BRB.(3)Aiming at the problem of too many belief rules in a BRB system,caused by multiple decision attributes with uncertainty,the antecedent attribute weights optimization is presented based on the concept of the importance of attribute.Finally,the paper provides a practical case on fault diagnosis for oil-immersed power transformer to illustrate the validity.The inputs of the model is to be cut by 40 percent and the correct recognition of fault diagnosis is increased by 3 percent.It shows that BRB-R is an effective method of attribute reduction.To deal with the disadvantages of BRB in modeling,the study has done from three aspects,which are include the structure and parameters identification and the model reduction.The application effect in three areas shows that the improved methods can effectively overcome the limitations of expert knowledge and accurately set the confidence framework.It has important practical engineering value.