Application of HMM-BK Reasoning Method in Multiple Sectioned Bayesian Networks

摘要

HMM-BK method under MSDBNs architecture is proposed in order to conduct state detection and realtime fault diagnosis for complex uncertain systems. This method can update node parameters of the entire model network timely according to the partial data which are provided by the sensors and then being inputted into the network as new proofs. It can meet the deficiency of reasoning ability caused by deletion of sensor data during actual system operation. Deaerator damage model is established and the method is applied for fault diagnosis. Experimental analysis shows that the method is correct and effective.