Hot rolling in the age of artificial intelligence: towards enhanced efficiency, quality and sustainability in steel industry

Sustainable manufacturing has emerged as a critical paradigm for the future of industry, integrating environmental stewardship with operational excellence. Against this backdrop, the rapid evolution of machine learning and deep learning is catalyzing the intelligent transformation of hot rolling production—a sector with substantial energy and material demands. This paper focuses on the core processes of hot rolling and provides a systematic review of machine learning applications for predicting and controlling key quality indicators, including crown, thickness, and width. By fusing multi-source process data with advanced intelligent algorithms, we construct a predictive quality framework tailored to the hot rolling environment, enabling precise monitoring and real-time adjustment of crown deviations, thickness variability, and width fluctuations. Empirical results demonstrate that machine learning approaches effectively capture complex, nonlinear interdependencies between process parameters and product quality, significantly enhancing both production stability and consistency. In addition, we propose a multimodal perception-based framework for early warning and dynamic optimization, facilitating coordinated, real-time parameter adjustments. This integrated strategy not only offers a viable technical pathway for smart manufacturing but also contributes practical insights toward more sustainable and resource-efficient hot rolling operations.