Analysing the role of physics-informed neural networks in modelling industrial systems through case studies in automotive manufacturing

When modelling industrial systems, two major challenges need to be addressed: complexity and uncertainty. Traditional modelling approaches can be categorized into data-driven and physics-based methods. Data-driven models excel at identifying patterns but struggle with interpreting the complexities of physical systems and often require large amounts of data. In contrast, physics-based models rely on detailed information that is often incomplete or uncertain. To overcome these limitations, Physics-Informed Neural Networks (PINNs) have emerged as a hybrid approach, combining the flexibility of neural networks with the rigour of physical constraints. This approach leverages both data and physics to provide more accurate predictions. The paper reviews the applications of PINNs in industry, demonstrating their advantages over traditional methods, particularly in terms of efficiency and versatility. To validate the model, two case studies from the automotive industry are presented. The first case study involves predicting the thermal power of an industrial air handling unit (AHU) used in a topcoat painting process, while the second focuses on forecasting temperature profiles in a degreasing tank. The results from both case studies show that PINNs outperform traditional data-driven techniques, offering significant benefits such as enhanced efficiency even with simplified physical models and reduced requirements for large training datasets. Future research directions include exploring different neural network models or alternative ways to integrate physical knowledge, aiming to further improve the performance and applicability of PINNs in industrial settings.