Advancing Airfoil Design: A Physics-Inspired Neural Network Model

Abstract Turbomachines are an integral part of the energy and industrial landscapes, and improvements to their efficiency benefit the environment, profitability of operation, and in turn, society at large. Therefore, the application of advanced methods for rapid design and development of high-performance turbomachinery components is of significant interest. In the past decade, the use of optimization methods has made inroads in improving turbomachinery aerodynamics. Recent advances in machine learning (ML) methods have the potential to augment design systems by providing the ability to explore larger design spaces and generate high-quality initial designs. Physics Informed Neural Networks (PINNs), based on the Navier-Stokes equations, are used to incorporate physical laws into the design process. This approach leverages the power of deep learning while ensuring that the designs conform to fundamental principles of fluid dynamics. The use of Physics Informed Neural Networks (PINNs) not only accelerates the design process by reducing the need for extensive simulations but also improves the accuracy of the designs by ensuring physical consistency as opposed to designs made using Generative Artificial Intelligence (AI) models. However, combining PINNs with Generative AI for airfoil optimization could provide a fruitful avenue in improving compressor blade designs.