Thermoelectric Coupling-Based Detection and Application Research on Leakage in Aluminum Electrolysis Cells

• Simulation analysis identifies distinct indicator patterns for bottom vs. side leakage in aluminum electrolytic cells • leakage fault formation mechanisms supported by case-validated AHP indicator ranking • PSO-LSTM prediction shows higher reliability when using simulation-derived key indicators • Integrated simulation and data-driven framework enables earlier and more accurate leakage-risk diagnosis Leakage faults stand out as a typical failure type of aluminum electrolytic cell characterized by high incidence rates and severe consequences. Therefore, leakage fault detection in aluminum electrolytic cells represents a critical technical challenge urgently requiring solutions in the aluminum electrolysis safety management. This study presents a new detection method of aluminum electrolytic cell leakage, we employed numerical simulation methods to identify the distinct key influencing indicators associated with bottom and side leakage troughs. Based on these findings, hypothetical mechanisms leading to leakage trough formation were proposed and analyzed. Through a case-based Analytic Hierarchy Process (AHP), an importance ranking of indicators consistent with the simulation results was obtained. Furthermore, a classification and prediction experiment using a PSO-LSTM deep learning model was conducted, in which different sets of indicators were applied to predict leakage risk. The results demonstrated that the indicators identified in the simulation analysis exhibited higher reliability in leakage risk prediction. This work contributes to the measurement domain of aluminum electrolytic cell leakage by proposing a simulation-informed, data-driven diagnostic methodology, offering theoretical foundations and technical support for early detection and effective safety prevention management of leakage faults in aluminum electrolytic cells.