Comparison of Finite-Volume and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>Spectral</mi><mo>/</mo><mi>h</mi><mi>p</mi></mrow></math></inline-formula> Methods for Large-Eddy Simulation of Combustor Port Flow

You are here: Homepage > Search

Comparison of Finite-Volume and $Spectral / h p$ Methods for Large-Eddy Simulation of Combustor Port Flow

Saini, Vishal / Xia, Hao / Page, Gary J.

High-order accurate methods provide significant accuracy/cost benefits for well-resolved and geometrically simple scale-resolving turbulent flow simulations. However, the benefit on under-resolved unstructured grids for complex industrial geometries is unclear. The purpose of this work is to contribute to understanding of the benefit of high-order schemes for large-eddy simulations (LES) in practical applications. A crucial requirement is a high-order solver that supports hybrid unstructured grids. In this context, accuracy and cost of a high-order spectral/ $h p$ Nektar++ solver and a standard second-order finite-volume OpenFoam solver are systematically compared for performing LES (with subgrid scale treatments) on two configurations. The first configuration is Taylor–Green vortex, which shows that the accuracy benefit of fifth-order LES depends upon the mesh type and the level of under-resolution. The second configuration is geometrically more complex and represents the dilution port flows of gas turbine combustors. Specifically, the flow consists of crossflow radial jets impinging onto each other, providing a rich variety of features and thus making the case challenging. It is found that, for a given cost, fifth-order LES reproduces the unsteady flow features significantly better and offers moderate improvements in the mean quantities. Moreover, results from a finer $5 \times$ more costly OpenFoam simulation suggest that, for a similar accuracy, fifth-order LES could be $3 - 8 \times$ cheaper.