Mingorance Mingorance, RocíoCrespo Pereira, DiegoUribetxebarria, JoneLeturiondo, Urko2026-05-072026-05-072026-06R. Mingorance Mingorance, D. Crespo Pereira, J. Uribetxebarria, U. Leturiondo, Enhancing proactive real-time decision making in manufacturing: A methodological-functional framework, Results in Engineering 30 (2026) 110033. https://doi.org/10.1016/j.rineng.2026.110033.2590-1230https://hdl.handle.net/2183/48193[Abstract]: The manufacturing industry faces major challenges in adapting to automation, driven by market volatility, energy efficiency demands, environmental concerns, and rising competitiveness. Recent advancements in digital technologies and artificial intelligence provide promising tools for manufacturers to tackle operational complexity and enhance sustainability in a demanding market. However, these challenges reveal a critical gap in existing operational frameworks, which often lack mechanisms for proactive, real-time decision-making in response to unforeseen changes in industrial processes. This paper presents an improved methodological-functional framework that combines physics-based and data-based models to strengthen the alignment between the digital twin and its physical counterpart. It leverages historical data to improve predictions and decision-making, minimizing disturbance impacts and adapting to evolving plant conditions. Unlike traditional models, which do not account for such variations, this framework uses real and synthetic data to generate projections that support proactive real-time decision-making. Experimental results from a steel industry case study demonstrate that this predictive approach yields an average 11.33% improvement in the objective function, corresponding to an average cost reduction of 133.82 $/t across the evaluated scenarios. Furthermore, the evaluation of multiple optimization techniques enables customized decision-making under varying timing and operational constraints. The real-time decision-making method attains objective function values up to 1.62% higher than those of metaheuristic approaches, corresponding to cost increases of up to 21.89 $/t. However, it reduces computation time to 0.27 minutes, compared with up to 25 minutes for metaheuristics, making it suitable for time-constrained, near-optimal applications.engAttribution-NonCommercial-NoDerivatives 4.0 Internationalhttp://creativecommons.org/licenses/by-nc-nd/4.0/Digital twinDecision-makingOptimizationDynamics chedulingMachine learningManufacturingjournal articleopen access10.1016/j.rineng.2026.110033