Turbulent Fluid Flow over Aerodynamically Rough Surfaces Using Direct Numerical Simulations

Abstract

Incompressible turbulent fluid flow in aerodynamically rough channels is investigated using direct numerical simulations. A comprehensive database of simulation data for rough surfaces with different topographical properties has been developed for 17 industrially relevant rough surface samples. It includes numerous commonlyseen industrial rough surfaces such as concrete, graphite, carbon-carbon composite and ground, shotblasted and spark-eroded steel. Other surfaces such as cast, filed and gritblasted steel are also studied, along with replicas of ship propeller surfaces eroded by periods of service. The Reynolds number considered is Re τ = 180, for which the flow is in the transitionally rough regime. A study with variable δ /S q ratio while keeping S + q constant, where S q is the root mean squared roughness height, is conducted for one of the samples with the mean profiles showing convergence for δ /S q >≈ 25. A Reynolds number dependence study is conducted for two of the samples with Re τ up to 720 showing a more complete range up to the fully rough flow regime, allowing the equivalent sandgrain roughness height, k s to be computed. A correlation based on the frontal and wetted roughness area is found to be superior to the surface skewness in predicting ∆U + based on the topographic surface parameters.

An assessment of the ship drag penalty arising from light calcareous tubeworm fouling

“…In recent years, there are increasing examples in the literature where DNS has been used to investigate realistic scanned rough surfaces Busse et al [2013Busse et al [ , 2015. There are also examples where DNS or Large Eddy Simulation (LES) have been employed to compute the equivalent sandgrain roughness of particular surface topologies Yuan & Piomelli [2014], Chan et al [2015], Thakkar et al [2015]. Though these studies are promising, the surfaces tested to date are all relatively homogeneous, with compact wall-parallel scales, and hence lend themselves well to efficient computations in limited numerical domains.…”

Numerical study of Fourier-filtered rough surfaces

“…As seen in table I, the blockages based on S z,max or S z,5×5 associated with the different surfaces are considerably larger than the value of 1/40δ recommended by Jiménez [29]. A blockage study carried out by Thakkar [24] on a sub-section of the same surface scan used here revealed little sensitivity of the flow to blockages based on S z,5×5 up to ∼ 0.17δ. In their case, the dimensions of the sub-sections used to compute S z,5×5 were smaller than here, since the domain size (5δ by 2.5δ) was significantly smaller.…”

“…In their case, the dimensions of the sub-sections used to compute S z,5×5 were smaller than here, since the domain size (5δ by 2.5δ) was significantly smaller. Accounting for that difference by computing S z,10×10 (defined as S z,5×5 by splitting the surface into 10 by 10 tiles) yields, as seen in table I, mean peak-to-valley heights comparable to the S z,5×5 reported by Thakkar [24].…”

“…[20,25]). A sub-section of the scan was selected following the procedure outlined in Thakkar [24], which was then treated as periodic along both directions. To ensure the resulting surface had broad spectral content, the stream-( x ) and span-wise ( y ) dimensions of the sub-section were chosen to be respectively 10δ by 5δ, with δ the half-channel height, roughly twice the size used by Busse et al [20], Thakkar et al [25].…”

“…Flow visualisation and analysis of two-point correlations reveals how different scales contribute to different flow features present in the original surface. The baseline surface in question has been extensively studied and its roughness function ∆U + was found to closely follow that of Nikuradse-type roughness [24]. Specific questions we aim to address are: whether existing correlations can predict the roughness function for these surfaces; to which extent is the turbulence influenced by the surface topography and whether any of the surfaces gives rise to specific structures in the mean flow and the turbulence.…”