LAUSR

Abstract A dynamic Bayesian network (DBN) based fault detection, root cause diagnosis, and fault propagation pathway identification scheme is proposed. The proposed methodology generates evidence from monitored process data and uses the information to update the DBN that captures the process knowledge. A new dynamic Bayesian anomaly index (DBAI) based control chart is proposed for detection purpose. Following the detection of the fault(s), root cause(s) is diagnosed using the smoothing inference of a DBN, and fault propagation pathway is identified from the cause-effect relationships among the process variables. The proposed methodology is applied to a binary distillation column and a continuous stirred tank heater (CSTH). The result shows that it can detect the fault and diagnose the root cause of the fault precisely. The result has been compared to the performance of the Shewhart control chart, principal component analysis (PCA) and static BN. The comparative study confirms that the proposed methodology is a more efficient fault detection and diagnosis (FDD) tool.