Is DNS Ready for Prime Time?: The State of ANSYS Fluent DNS in 2025

Abstract Can an industrial computational fluid dynamics (CFD) tool like ANSYS Fluent perform Direct Numerical Simulation (DNS) reliably? We answer by providing a combination of a feasibility report and a best-practice guide that will be useful for researchers and practicing engineers as pressure mounts to transition DNS from academia to industry. DNS has historically been the realm of efficient but somewhat restrictive spectral solvers. DNS has been validated on Finite Volume Method (FVM) solvers, but many of these works have been on structured grids with high-order stencils. This paper explores a canonical low-Re channel flow using ANSYS Fluent, a commercial FVM code, with lower order methods. Additionally, for the first time, this paper provides direct comparisons of the diffusion and dispersion of Fluent's momentum discretization schemes on six easily reproducible grids: two grid spacings each of grids composed of triangular prisms, hexahedrons, and polygonal prisms. We also demonstrate the substantial increase in solution speed offered by GPUs. While velocity profiles are predicted with moderate error, wall-normal profiles of turbulent kinetic energy budgets exhibited significant error, and large amounts of spectral pileup were visible. Therefore, it can be concluded that, for a canonical flow configuration, ANSYS Fluent is only partially “ready”. While it cannot be said that all FVM DNS is problematic for all possible flows, we prove, through a simple flow scenario, that healthy skepticism should be applied when the term “DNS” is used in conjunction with ANSYS Fluent and FVM codes which use similar stencils.