Anomaly Diagnosis of Catalytic Cracking Process Based on Autoencoder and Causal Model

The regeneration process of catalytic cracking reactions is the core process of refining and chemical enterprises, which localized anomalies can propagate in cascade, leading to severe consequences. Accurate anomaly diagnosis is crucial for identifying root causes and establishing a basis for implementing refined protective measures to prevent the further propagation of anomalies. Based on this, the paper proposes a method for abnormal detection of root causes in the regeneration process of catalytic cracking reactions. In this method, the Diagonal State Space Autoencoder (DS2AE) is utilized to detect abnormal process variables. In addition, the Maximum Information Coefficient and Time-Delayed Transfer Entropy-based causal analysis method (MIC-TDTE) is employed to construct causal models among variables. Finally, causal graphs are traversed to trace the propagation paths of anomalies, thereby detecting the root causes of anomalies. The proposed method is applied to a dataset of abnormal working conditions constructed from a catalytic cracking process simulation system, demonstrating the effectiveness of the approach.