Applications of artificial intelligence for boundary layer and multiphase flows

This chapter goes over the evolution of artificial intelligence (AI) and its impact on boundary-layer and multiphase flow modeling and simulation methods in the field of computational fluid dynamics. We explain how AI-based methods like machine learning, deep learning, and reinforcement learning are overturning established practices and improving traditional methods like turbulence modeling, automating mesh generation problem-solving, and providing real-time prediction and control of flows. Recently, coupled data-driven models in AI and hybrid approaches like physics-informed neural networks, Fourier neural operators (U-FNO), etc., have demonstrated substantial improvements in prediction accuracy, computing performance, and physical consistency in boundary-layer and multiphase flow problems. We have presented examples from the industrial application of AI in the fields of aerospace, oil, and gas, and chemical processing that have demonstrated improved opportunities for design, operation, and uncertainty quantification. Several previous discussions indicated there are continuing challenges with regard to data and other related needs, model interpretability, generalizability, and regulatory issues. The description of future research directions toward achieving the concept of robust, interpretable, and ethically regulated AI-CFD integration will set the stage for discussion in future papers. We are suggesting that AI reconciles the data and physical-driven paradigms and is not simply an augmentation as a practice, but a paradigm shift for fluid dynamics as a scientific and engineering practice.