Towards a Digital Twin for Production Planning: Combining Discrete Event Simulation and ML for Flexible Job Shops

In the era of Industry 4.0, Digital Twins are transforming industrial processes through real-time monitoring and predictive analytics. This study addresses the challenge of embedding full-scale Discrete Event Simulation models within simulation-based Digital Twins due to their high computational demands. To overcome this, we propose using Discrete Event Simulations to generate synthetic data for training Machine Learning models, which are then integrated into the Digital Twin framework for efficient production planning. A Discrete Event Simulation model was developed to represent a flexible job shop production system. The data generated from this model was used to train various Machine Learning models, with XGBoost achieving the best performance, yielding a Mean Absolute Percentage Error (MAPE) of 15.47%. This case study demonstrates the practical applicability and benefits of the proposed approach.