Hybrid mechanistic-data modelling for dynamic water distribution in slab caster secondary cooling zone

The water flow rate distribution in the secondary cooling zone is a critical control parameter determining the slab internal quality. Aiming at the limitations of low precision and delayed response of conventional water distribution systems, the study proposes a hybrid mechanistic-data modelling approach. First, the prediction model of slab surface temperature based on the solidification heat transfer principle was established, and a genetic algorithm was used to optimise the model parameters to improve the prediction accuracy. Secondly, a water flow rate optimisation model based on the particle swarm algorithm was established with metallurgical mechanism as the constraint. Finally, a dynamic water flow rate distribution model based on the XGBoost algorithm was established, and the high-quality samples generated by the temperature prediction model and the water flow rate optimisation model were used for model training, so as to achieve the deep integration of metallurgical mechanisms and industrial big data, and to regulate and control the dynamic water distribution in real time. On-site application results show that the dynamic water distribution systems can effectively reduce the response time and improve the control accuracy. The research provides a new idea for the dynamic water distribution and efficient control of the secondary cooling zone.