An Offline Digital Twin Case Study for Data-Constrained Energy-Intensive Foundry Production

Energy-intensive foundries require methods to explore trade-offs among delivery performance, production horizon, electricity use, and cost when industrial data are incomplete. This paper presents an offline, process-level digital twin case study for the melting and casting area of a Danish cast-iron foundry. A multi-agent simulation represents production orders, enterprise resource planning and manufacturing execution system functions, induction furnaces, holding furnaces, crane-based transfer of molten metal, vertical moulding lines, the operating calendar, and electricity cost accounting for the induction furnaces. The model is assessed through boundary definition, assumption registration, implementation checks, material flow plausibility, a diagnostic comparison of furnace temperature, controlled scenario experiments, and local sensitivity analysis. These activities support internal consistency and bounded interpretation but do not constitute independent operational validation of the full production system. In the simulated 200-order monthly case, First-Come-First-Served and Earliest Deadline First complete the same 288,620 pieces and 5482.00 t. Earliest Deadline First increases the simulated on-time completion rate from 87.5% to 100%, while makespan, model-estimated electricity use by induction furnaces, and model-estimated electricity cost increase by 7.52%, 0.58%, and 3.42%, respectively. The case indicates that deadline-oriented sequencing may improve delivery performance but lead to a longer production horizon and higher energy use and cost within the defined model boundary. The contribution is an auditable foundry-specific modelling workflow that links heterogeneous data conditions to modelling choices, supporting evidence, and interpretation limits. The model is therefore intended for preliminary offline scenario exploration rather than validated operational decision support.