Flexible Cooling Strategy for Hot-Rolled Steel Based on Physical Theories Coupled with Machine Learning

Based on the complex characteristics of nonlinearity, strong coupling, and multi-disturbance in hot-rolling steel production, developing a multi-objective optimization system for dynamic control of multi-scale process parameters is challenging. In this paper, novel solutions coupling physical mechanisms into the machine learning modeling are proposed, and their feasibility has been fully demonstrated by analyzing experimental verification and industrial trial production. By comparison with experimental data, the results show that the proposed model accurately describes the microstructure evolution during hot rolling and cooling under complex processing conditions. On this basis, the flexible cooling system is developed by multi-objective particle swarm optimization. Compared with the traditional purely data-driven optimization method, results show that the proposed model ensures that the optimization results meet the requirements of multiple property indicator collaborative optimization but also obtain the optimal comprehensive property and quality stability by controlling the cooling path. Finally, the established system applied to guide the industrial production of steel and the potential for matching optimum cooling parameters according to the fluctuation in steel composition and rolling parameters to achieve a compensatory effect is proved through metallographic observations of final microstructures under different cooling paths.