Enhancing Automotive Paint Curing Process Efficiency: Integration of Computational Fluid Dynamics and Variational Auto-Encoder Techniques

The impetus of the present work is to propose a comprehensive methodology for the numerical evaluation of drying/curing, as one of the most complex and energy-consuming stages in the paint shop plant, to guarantee a decrease in energy costs without sacrificing the final paint film quality and manufacturability. Addressing the complexities of vehicle assembly, such as intricate geometry and multi-zoned ovens, our approach employs a sophisticated conjugate heat transfer (CHT) algorithm, developed under the OpenFOAM framework, providing efficient heat transfer with the accompaniment of the Large Eddy Simulation (LES) turbulence model, thereby delivering high-fidelity data. This algorithm accurately simulates turbulence and stress in the oven, validated through heat sink cases and closely aligning with experimental data. Applying modifications for the intake supply heated airflow rate and direction leads to optimal recirculation growth in the measured mean temperature within with the curing oven and along the car body surface, saving a significant amount of energy. Key adjustments in airflow direction improved temperature regulation and energy efficiency while enhancing fluid dynamics, such as velocity and temperature distribution. Furthermore, the study integrates machine learning to refine the oven’s heat-up region, which is crucial for preventing paint burnout. A data-based model using a variational auto-encoder (VAE) and an artificial neural network (ANN) effectively encodes temperature and velocity fields. This model achieves an impressive 98% accuracy within a 90% confidence interval, providing a reliable tool for predicting various operational conditions and ensuring optimal oven performance.