(187j) Root Cause Diagnosis of Process Fault Using Modified Convergent Cross Mapping and Bayesian Network

Modern plants’ measurement devices and control systems are cross-linked with each other. A disturbance in major equipment may lead to plant-wide disturbances, which makes it difficult to find the root cause. This paper develops a method to combine data-driven methods and process knowledge to diagnose the root cause of an abnormal status in industrial process. Convergent cross mapping(CCM) is an algorithm to detect causality of two time series variables. CCM is based on the theory of Takens’ embedding theorem. Compared to other causal algorithms, CCM can be applied to systems where causal variables have synergistic effects. The causality detected by the CCM between process variables and process knowledge are combined to construct the causality network. The process data is used to estimate the conditional probabilities of the Bayesian network. After an alarm is detected, the Bayesian network is updated by changing the status of the node. The proposed method enable the operator to diagnose the fault. To test the effectiveness of the proposed method, the method is applied to Tennessee Eastman Chemical Process.