Research On Reasoning Algorithm Based On Dynamic Uncertain Causality Graph

In the field of modern artificial intelligence technology,knowledge representation and reasoning algorithm based on it are very important in the process of constructing intelligent system.The knowledge information that modern intelligent system needs to process is mostly uncertain causal information,so it's necessary to study the intelligent system that can process uncertain causal information easily.Dynamic uncertain causality graph is a method to deal with uncertain causal information in practical problems on the premise of conforming to the basic laws of probability theory.The concise knowledge representation and reasonable and effective reasoning methods make the wide application of dynamic uncertain causality graph in fault diagnosis,risk assessment and prediction,which is more and more accepted by the world.This paper focuses on the knowledge representation and reasoning algorithm of dynamic uncertain causality graph(DUCG).The main contents list as follows:(1)Aiming at the construction of dynamic uncertain causality graph model,in this paper,the traditional construction method is given at first,secondly,for the existing typical fault tree,three steps of DUCG model transition are summarized by using the idea of taking part as a whole,which are event transition,numerical transition and logical transition,and then the transition algorithm is proposed to realize the transition to a more advantageous model.(2)In view of the complexity of conditional events in DUCG model,conditional concatenation events in single-valued DUCG and conditional events in multi-valued DUCG are discussed in detail at first,then two practical theorems are proposed according to the type of conditional event,the results show that the given theorems are more suitable for the expansion of variable logic expression under complex conditions than traditional reasoning algorithm.(3)For the directed cyclic graphs in DUCG model,two classes of cyclic graphs are defined,and the general forms are induced according to the traditional reasoning process of single-valued and multi-valued DUCG,respectively.A new reasoning algorithm based on causal intensity matrix is proposed to solve the DUCG model involving complex directed cycles.The algorithm searches and breaks the loops according to the prescribed path rules.The results of the examples further prove the rationality and reliability of the algorithm.