A Hybrid Diagnostic System for Corrosion Mechanism Identification in Petrochemical Equipment: Integrating Bayesian Probabilities and Association Rule

This study proposes a diagnostic model for corrosion degradation mechanisms in fixed equipment within the petrochemical and chemical industries, based on association rule mining and Bayesian classification. The model addresses challenges associated with traditional corrosion diagnosis, including insufficient expertise, lack of specialists, and complex evaluation processes. By integrating international standards and employing a data-driven approach, the model provides quantitative diagnostic results for corrosion degradation mechanisms. Combining the theoretical foundation of corrosion mechanisms with real-world operational data, the model is capable of diagnosing 91 types of degradation mechanisms, encompassing 11 material types, a wide operating temperature range, 15 chemicals, and various loading conditions. To validate the model's accuracy, a simulation was conducted using the Chevron accident case. The results demonstrated consistency between the model's outputs and the findings of the accident investigation report, thereby confirming its effectiveness. The diagnostic tool developed in this study enhances the accuracy and efficiency of corrosion degradation assessments and offers the potential for dynamic improvement as more case data is accumulated. By presenting quantitative indicators of the probabilities of various degradation mechanisms, the model facilitates the early detection of equipment abnormalities and provides a solid foundation for subsequent analyses such as Risk-Based Inspection (RBI), Fitness for Service (FFS), and Integrity Operating Windows (IOW).