Data-Driven and Model-Driven Approaches in Predictive Modelling for Operational Efficiency: Mining Industry Use Case

In this study, we explore the effectiveness of a hybrid modelling approach that seamlessly integrates data-driven techniques, specifically Machine Learning (ML), with physics-based equations in Simulation. In cases where real-world data for industrial processes is insufficient, a simulation tool is employed to generate an extensive dataset of process variables under varying operating conditions. Subsequently, this dataset is utilized for training the Machine Learning model. The paper showcases a practical use case of this hybrid modelling approach, revealing a model that consistently demonstrates strong predictive accuracy and reliability within the specific industrial context we investigate. By merging the insights derived from physics-based understanding with the adaptability of data-driven Machine Learning, the hybrid model offers a comprehensive solution for precise and accurate predictions.