Update Parameters Dynamic in Causality Diagram

Causality diagram theory is a new uncertainty reasoning model based on probability theory, which adopted direct cause-effect intensity and graphical knowledge representation. It has important theoretical meaning and application value for fault diagnosis. Linkage intensity is the basis of the inference which is the parameters not easy to obtain, it is often given by field experts. In this paper, the algorithm of EM(eta) is proposed to learn causality diagram parameters (linkage intensity) dynamic, which can make the parameters adapt with the change of environment, and this method´s feasibility and advantage are proved in theory. Experimental results show the validity and the superiority of the method as well. At last, we compared the differences with the learning of causality diagram parameters static.