Development of a GCN-LSTM Based Prediction Model for Sticking Breakout in Continuous Casting

Continuous casting is a critical stage in steelmaking, yet it is prone to breakout accidents with complex influencing factors and mechanisms. Among these, sticking breakout is the most frequent type, making the development of an effective prediction model essential for operational safety. This paper conducts an in-depth analysis of the sticking breakout mechanism, tracing the complete process from sticker initiation to crack propagation, and clarifies the spatial characteristics of thermocouple temperature variations during such events. A novel deep learning framework is proposed, utilizing the first-order derivative of thermocouple temperature as model input. This framework integrates Long Short-Term Memory (LSTM) networks and Graph Convolutional Networks (GCN) to learn and extract spatiotemporal features from the temperature data. Through designed ablation experiments and extensive analysis of field data to investigate the impact of model parameters, optimal parameters were identified, validating the effectiveness and superiority of the GCN-LSTM model for sticking breakout prediction. The finalized model, with optimal iterations, learning rate, timestep, and hidden layer dimension set to 176, 0.01, 10, and 5 respectively, achieved a prediction rate, correct alarm rate, and F1-score of 99.58%, 100%, and 99.79%.