Digital Twin-Enabled Thermal Energy Management System for Sustainable Manufacturing Process Optimization
Abstract
Manufacturing processes consume substantial thermal energy, yet siloed management approaches cannot exploit facility-wide synergies. This study develops and validates an integrated Digital Twin (DT) that fuses physics-based thermal models with machine-learning forecasts and multi-objective optimization to coordinate process heat, waste-heat recovery, thermal storage, and on-site renewables in real-time. Deployed across four heterogeneous manufacturing facilities, the DT generated operator-ready knee-point recommendations that balanced energy use, operating cost, and emissions under changing production and weather conditions. Across sites, deployment produced substantial, sustained gains in thermal-energy efficiency and marked reductions in carbon intensity (approximately 27% higher efficiency and about one-third lower emissions in aggregate), demonstrating that system-level orchestration outperforms isolated component upgrades. Novelty lies in plant-scale, real-time co-optimization of process heat, waste-heat recovery, thermal storage, and on-site renewables using a hybrid physics–ML digital twin with uncertainty-aware multi-objective control, field-validated across four heterogeneous manufacturing sites.