RUB » Bibliotheksportal

Nina Krautgartner, Yara Salloum

• Agent-based simulation (ABS) and machine learning (ML) can be combined to model and analyze complex systems. ABS involves creating agents that simulate decision-making processes in real-world systems, while ML uses data-based algorithms to handle big data and gain insights into these systems. This paper proposes a method to exchange data between an ABS model of a mechanized tunneling project and an ML prediction model to improve the simulation’s outcomes. In mechanized tunneling projects, the initial estimation of project duration often deviates significantly from the actual duration due to unplanned disturbances, leading to increased costs. To address this issue, simulation models developed during the planning phase can be updated with real-time data to make them useful in real-time scenarios. Here, ML models are employed to train sensor data from the tunnel boring machine (TBM) to predict the ring-building duration parameter of the ABS accurately for the next steps. This enhances the estimation of the total tunnel construction duration. This approach is adaptable to various applications depending on data availability. The paper concludes by suggesting further research directions for leveraging data exchange to enhance modeling capabilities in complex systems.