Researching On The Improved Method Of Causality Diagram Inference

The dynamic causality diagram developed on the basis of belief network is a knowledge representation model based on probabilistic theory.There are many similarities with the belief network,but also making up the deficiency of the belief network.Causality diagrams are characterized by the ability to express uncertain knowledge and to be flexible in reasoning.Especially in the fault diagnosis can be very good application,in the actual project has the important application value,therefore there is the important academic significance to causality diagram theory research.This paper mainly studies the inference method of causality diagram,which includes:Firstly,the paper introduces the knowledge expression model of causality diagram,by summarizing the traditional inference methods,it is found that the traditional inference method of causality diagram needs a lot of logical operation to solve the final cut set and the disjoint cut set,especially in the complex causality diagram model,the computational volume will be very large,and the reasoning process is very complex.To solve this problem,a new reasoning method is proposed.The method of transforming causality diagram into binary decision diagram is to show the whole inference process through the graph structure of binary decision diagram,and the direct search path gets the disjoint cut set,reducing the reasoning complexity and improving the inference speed.Considering the minimum cut set of causality diagram for fault diagnosis,in order to find out the fault source as soon as possible,the step function is introduced to analyze causality diagram,and the causality diagram is transformed into function form,which is convenient for programming and can improve the calculation speed.Secondly,in practical engineering applications,it is very difficult to get the exact probability value of the event,and a method based on Dempster-Shafer evidence theory is proposed to replace the interval number with the probability value,and the relative superiority degree is used to compare the interval number to determine the fault source and apply it to the fault diagnosis in the practical example,avoiding the problem that the exact value is difficult to obtain.