The field of fluid dynamics and thermodynamics is witnessing a significant shift towards the adoption of machine learning techniques to improve prediction accuracy and efficiency. Researchers are exploring the use of neural networks, transfer learning, and other machine learning approaches to enhance the performance of traditional computational fluid dynamics (CFD) simulations and thermodynamic models. These innovative methods are being applied to a wide range of applications, including indoor airflow and temperature prediction, aerodynamic analysis, and thermodynamic property prediction. Notably, the integration of physics-based knowledge into machine learning models is leading to improved accuracy and generalization capabilities.

Some noteworthy papers in this area include: The paper on component-based machine learning for indoor flow and temperature fields prediction, which proposes a novel surrogate modeling approach for fast and accurate prediction of indoor velocity and temperature fields. The paper on fusing CFD and measurement data using transfer learning, which introduces a non-linear method for combining simulation and measurement data via transfer learning, resulting in significant improvements over established methods. The paper on data-driven extended corresponding state approach for residual property prediction of hydrofluoroolefins, which proposes a neural network extended corresponding state model that shows significantly improved accuracy for density and energy properties. The paper on adjoint-based aerodynamic shape optimization with a manifold constraint learned by diffusion models, which introduces a framework that integrates a diffusion model trained on existing designs to learn a smooth manifold of aerodynamically viable shapes, achieving superior aerodynamic performance compared to conventional approaches.