Fluid flow is governed by the Navier-Stokes equations. At the heart of any fluid flow solver lies a numerical algorithm that discretizes and solves the Navier-Stokes equations. With the advent of high performing CPUs/GPUs/exascale-computing, numerical simulations of fluid flow are adding more and more value to our understanding of various challenging engineering problems. In this front, it is important to develop numerical algorithms and software programs that are accurate, robust, scalable and smart in utilizing the computational resources delivering both fidility and speed.

With the introduction of machine learning, we have made significant progress in addressing high-dimensional non-convex and/or non-linear problems that were previously daunting. In fluid flow research, modern optimization techniques and machine learning can improve flow predictions and enhance our understanding of flow dynamics. The image on the left shows an example task, to predict farfield noise based on nearfield flow input.