Digital Twin-Based Hybrid Simulation–Prediction Framework for KPI Optimization in Sustainable Digital Printing

The increasing emphasis on sustainability in digital printing requires quantitative methods for optimizing key performance indicators (KPIs) under technical and operational constraints. The term digital twin is used here in a methodological and analytical sense, as a simulation framework for analyzing interdependence, prediction, and multi-criteria optimization of KPIs, rather than as a direct virtual replica of a specific physical production system. This paper proposes a hybrid simulation–prediction model based on a digital twin framework for optimization of KPIs in sustainable digital printing, with particular emphasis on overall equipment effectiveness (OEE). Due to the limited availability of structured industrial data, the model is developed using a synthetically generated dataset constructed in accordance with industry-reported operating ranges and technically realistic digital printing process variables. Random Forest and XGBoost algorithms are applied to model nonlinear relationships between process parameters and KPIs, including material waste, energy consumption, machine downtime, and OEE. Based on these predictive models, a constrained multi-objective optimization procedure is performed to identify Pareto-efficient configurations that reduce material waste and energy consumption while maintaining acceptable downtime and OEE levels. The results characterize structural trade-offs among environmental and operational KPIs within a formally defined decision space.