Data Modelling Challenges in Material Flow Optimization: A Machine Learning Perspective

Material flow optimization poses a unique challenge due to the heterogeneous and multidimensional nature of the data involved in representing factory layouts and processes. Consequently, the encoding and representation of this factory data significantly influences the choice and efficiency of machine learning (ML) solutions. In this research, we explore three levels of model complexity to represent production systems with the goal of predicting machine utilization of each functional unit. Through a comprehensive evaluation, we assess how the complexity of the modeling framework affects the predictive accuracy of ML models. Our results indicate that incorporating information about the interconnections between functional units yields the most favorable complexity/accuracy ratio. This study highlights the potential of ML in material flow and factory layout optimization, emphasizing the critical role of data representation and the importance of considering interdependencies within production systems. The insights gained can guide practitioners and researchers in effectively applying ML to address the challenges of material flow optimization in dynamic manufacturing environments.