An experimental and numerical study of channel flow with rough walls

Abstract

A fully turbulent channel flow with smooth and rod-roughened walls has been investigated using hot-wire anemometry and direct numerical simulations (DNS). The mean flow follows the law of the wall for both surfaces and the velocity defect suggests that the outer layer is very little affected by the roughness. The Reynolds stresses appear to be very similar for the two surface geometries outside $y\,{\approx}\,5k$, where $k$ is the roughness height. A quadrant analysis shows that the structural differences close to the wall extend somewhat further out. The turbulence structure is further investigated using stress ratios and the anisotropy tensor, which corroborate the findings from the Reynolds stresses. Many of the recent investigations on boundary layers seem to find large differences between smooth and rough wall data in the outer layer also. A tentative explanation for the apparent dependence on flow type of the surface roughness effects is given.

Progress in Urban Meteorology :A Review

“…Turbulent flows within and above regular obstacle arrays have been investigated by direct numerical simulations (Miyake et al 2001;Nagano et al 2004;Krogstad et al 2005;Coceal et al 2006), LES (Hanna et al 2002;Cui et al 2003;Kanda et al 2004b;Kanda 2006a), and RANS (Lien and Yee 2004;Hamlyn and Britter 2005). Results from LES applications (Kanda et al 2004b;Kanda 2006a) suggest some important physical aspects that are currently not emphasized in simpler CFD simulations: i.e., large dispersive momentum flux within the urban canopy layer due to a mean stream such as recirculation, intermittent urban canyon flow, non-persistent stream patterns, and longitudinally elongated streaks of low speed over building arrays with a scale an order of magnitude larger than individual buildings (Fig.…”

Similarity of the streamwise velocity component in very-rough-wall channel flow

“…Antonia & Krogstad (2001) conducted measurements in the boundary layers developed over two walls with different roughnesses but with matching U + , and found significant differences in the observed turbulence characteristics in the outer layer, though the surfaces had similar values of k/δ ≈ 1.8 and 2.1 %. However, the combined experimental and numerical study of Krogstad et al (2005) found that, for a rough-wall channel with k/δ = 3.4 % and a surface topology similar to the transverse grooves of Bhaganagar et al (2004), the influence of the roughness did not extend beyond approximately five roughness heights from the surface, and the outer layer was unaffected. The authors attributed the apparent discrepancy to fundamental differences between channel and boundary layer flows.…”

“…By expressing the roughness size in terms of the equivalent sandgrain roughness, k s (which is associated with the total losses in fully developed turbulent pipe flow rather than any geometric parameter, Nikuradse 1933), Wu & Christensen showed that the relative roughnesses, k s /δ, of Krogstad et al (1992), Keirsbulck et al (2002) and Bhaganagar et al (2004) were 6.7, 13 and 13 %, respectively, and so that the application of the Townsend (1976) scaling assumptions may have been inappropriate in these cases. For the turbulent channel case of Krogstad et al (2005), Wu & Christensen noted that k s /δ was as high as 25 %, rendering the lack of observed effects of the roughness upon the turbulent flow structures even more surprising, regardless of boundary conditions. Clearly, the effect of large roughness upon turbulence structure is still not well understood, especially in the case of internal flows.…”

Free surface flow over square bars at intermediate relative submergence

“…For a flow of given bulk Reynolds number, R, over two-dimensional roughness elements, the transition between d-and k-type roughness depends solely on the streamwise spacing of the elements. Many researchers have chosen to focus on square bar roughness to investigate the effect of spacing on turbulence structure and mean flow characteristics, both experimentally (Coleman, Nikora, McLean, & Schlicke, 2007;Djenidi, Antonia, Amielh, & Anselmet, 2008;Djenidi, Elavarasan, & Antonia, 1999;Krogstad, Andersson, Bakken, & Ashrafian, 2005;Okamoto, Seo, Nakaso, & Kawai, 1993;Roussinova & Balachandar, 2011) and numerically (Cui, Patel, & Lin, 2003;Ikeda & Durbin, 2007;Stoesser & Nikora, 2008;Stoesser & Rodi, 2004). Simpson (1973), Tani (1987), Jiminez (2004) and Coleman et al (2007) all proposed that the transition from d-to k-type roughness occurs at around l/k = 5, where l is the crest-to-crest bar spacing and k is the roughness height.…”