Study On The Reasoning And Application Of Dynamic Causality Diagram

Research of Artificial Intelligence(AI) has come to practical application stage, and these applications have almost covered all the domains. These applications have raised amount of complexity which have not been fairly solved by the early AI theory. One of them is the uncertainties in the knowledge and available information on which the reasoning is based. Uncertainty is the essence character of the intelligent problems, and whether the human intelligence or the artificial intelligence can not avoid coping with the uncertainty. So the ability of resolving the uncertain problems represents the intelligence of system, and the reasoning model based on uncertainty has become a key research project in AI and Expert System(ES).Uncertainty knowledge representation can be classified into two categories: probabilistic and non-probabilistic. The probabilistic approaches include the Belief Network, the dynamic causality diagram, the Markov network, the approach used in PROSPECTOR, etc. The non-probabilistic approaches include the certainty factor theory in MYCIN, Fuzzy Set Logic, Dempster-Shafer theory, etc. The non-probabilistic approaches have reached some achievement in their respective application domain, and shown their shortage while applying. Among the probabilistic approaches, Pearl's belief network is the most representative, due to its rigorousness and consistence in theory, the efficient local computation mechanism and intuitive graphical expression of knowledge. The dynamic causality diagram developed from the Belief Network adopts its graphical expression of knowledge and then innovates and extends this expression approach, abolishes the limit to logical structure of system, and introduces Boolean calculation. Thus, the dynamic causality diagram overcomes the shortage of Belief Network, and possesses more greater advatage.This paper describes the details about knowledge representation and reasoning based on uncertainty in AI, and the outline of Belief Network. After introducing the causality diagram model and summarizing conventional reasoning algorithm, a new reasoning approach of causality diagram has been presented, which is aimed at the defects in conventional reasoning algorithm, which are the large amount of Boolean computation and its complexity. The new approach is on the basis of the former disjoint. A reasoning algorithm based on the causality influence possibility distribution ispresented while the causality diagram is discussed in multivalued case.At present, research of Artificial Intelligence is almost combined with the practical application domain, and there no exception with the dynamic causality diagram. So the problem of applying the theory to practical engineering, such as the fault diagnosis of industrial system, is our target all the time. In the last of the paper, this important problem is discussed with the instance of the control system of the helm and oar of ship. The causality diagram for fault diagnosis of its electric-hydraulic loops is constructed, and applied to diagnose fault offline. This approach is more efficient comparing with the fault tree, expert system and neural network approaches in the original system, and the diagnosis proceeding is fast and effective, while the result is consistent with the fact.