Investigation on Mold Heat Transfer Control by Integrating Measurement, Coupled Heat Transfer and Flow Simulation, and Machine Learning
摘要
The heat transfer in copper mold plays a crucial role in the initial solidification behavior of molten steel, making its control essential for maintaining the steel quality. In this study, a coupled model of water flow and copper mold heat transfer is developed based on thermocouple measurements and inverse heat flux calculations. The influence of copper mold thickness and primary cooling processes on heat transfer is investigated. The results show that for every 5 mm reduction in mold thickness, the maximum temperature on the hot face decreased by an average of 21.3 °C. Additionally, for every 5 °C decrease in the inlet water temperature, the maximum temperature on the hot face decreased by about 5 °C, while a reduction in cooling water velocity by 2 m s−1 resulted in an average increase of about 6.8 °C in the maximum temperature on the hot face. Based on the numerical heat transfer model, five machine learning regression models are developed to predict the hot face temperature of the copper mold, with the support vector regression model demonstrating superior prediction performance. Furthermore, an integrated method combining machine learning and numerical modeling is proposed to control the heat transfer of the copper mold. This method enables real-time adjustment of the cooling water velocity to maintain the hot face temperature within a specified range under different copper mold thicknesses and inlet water temperatures, which provides a novel strategy for controlling heat transfer in the copper mold.