A method for characterising the sensitivity of turbulent flow fields to the structure of inlet turbulence

Abstract

The specification of inlet conditions for eddy-resolving simulations is an important research question in turbulence research. A large number of schemes have been proposed but comparisons to a benchmark simulation are usually undertaken in such a way that it is ambiguous as to whether the source of the improved preservation of the properties of the benchmark case is due to the physics captured in the inlet generation algorithm or to differences in the precise values or spectra generated at a particular position. This article uses gradual wavelet reconstruction to degrade a precursor simulation such that the time series input at each cell has the same histogram and Fourier spectrum as the original precursor case, thereby removing this source of variability. A parameter, ρ, is used to index the nature of the degraded inlets and as ρ increases from 0 to 1 the cross-correlative and nonlinear properties of the precursor are increasingly preserved too. We compare the results of four large-eddy simulations for the flow over a wall-mounted square rib and show that there are clear differences between these simulations for the flow before and over the rib, and in the far-field. However, in the wake, the intense shearing and mixing means that any differences are harder to attribute to the nature of the inlet. Hence, for flows that are mixed strongly, there is less of a need to preserve the cross-correlative structure of the inlet condition as long as the individual velocity components are modelled with an appropriate set of values and Fourier spectrum. By comparing particular inlet generation algorithms in terms of the value for ρ that they can attain, their suitability for modelling flows of particular configurations may be determined.

Keywords:

• turbulence
• boundary conditions
• large-eddy simulations
• wavelets
• gradual wavelet reconstruction

Acknowledgments

We gratefully acknowledge the Transportation Research and Analysis Computing Center (TRACC) at the Argonne National Laboratory, in particular, Dr. S. Lottes, for providing substantial computer time.