Enhancing airfoil design optimization surrogate models using multi-task learning: Separating airfoil surface and fluid domain predictions

Computational fluid dynamics is essential for airfoil design optimization. Typically, it involves numerous numerical procedures such as grid generation, boundary condition setup, and simulations, leading to high computational costs and extended research periods, which pose a long-standing challenge to aerodynamic development. Recently, the data-driven deep learning method has emerged as a new approach, significantly reducing computational time. However, these models have difficulties maintaining the desired accuracy, particularly when balancing surface characteristics with internal volume features. In this study, we introduce a novel method utilizing the multi-task learning (MTL) to handle surface and volume predictions as interconnected yet distinct tasks. By employing multi-head neural network architectures and advanced MTL optimization strategies, our approach effectively resolves the inherent conflicts between airfoil surface and fluid domain predictions. Our method demonstrates significant improvement in predictive accuracy of both flow fields and the aerodynamic force coefficients. Extensive numerical experiments were conducted using an open-source dataset that includes flow field data for various airfoil shapes under different flight conditions. The results indicate that our MTL-based surrogate model outperforms existing models, providing more reliable and efficient tools for practical applications in aerodynamic engineering.