Coupling of an analytical rolling model and reinforcement learning to design pass schedules: towards properties controlled hot rolling

Abstract Rolling is a well-established forming process employed in many industrial sectors. Although highly optimized, disruptions can still lead to undesired final mechanical properties. This paper demonstrates advances pass schedule design based on reinforcement learning and analytical rolling models guarantee sound product quality. Integrating an established physical strengthening model into allows tracking the microstructure evolution throughout process, furthermore prediction of yield strength ultimate tensile rolled sheet. The trained algorithm Deep Deterministic Policy Gradient (DDPG) automatically proposes schedules by drawing upon scheduling rules combined with novel rule sets maximize designed trialed using laboratory mill while predicted properties are confirmed micrographs materials testing. Due its fast calculation time, prospectively this technique be extended also account for significant such as longer inter-pass times adapting online reach desired avoid scrapping material.