The Effects of Anisotropic Surface Roughness on Turbulent Boundary-Layer Flow

Measurements of a turbulent boundary-layer developing over systematically generated roughness are acquired for friction Reynolds numbers ranging between 3000 < Ret < 6000. A set of near-Gaussian surfaces with matched amplitude parameters and specified effective slopes in streamwise and spanwise directions are synthesised using a roughness generation algorithm. Three cases are considered: (i) an isotropic surface with equal streamwise (ESx =0:34) and spanwise effective slope (ESy = 0:34); (ii) an anisotropic spanwise elongated surface with ESx = 0:34 and ESy = 0:17, and (iii) an anisotropic streamwise elongated surface with ESx = 0:17 and ESy = 0:34. The surfaces are manufactured from square sheets of acetal copolymer using an in-house CNC router. Note that surface (iii) is obtained by simply rotating surface (ii) by 90 degrees. The principal interest here is to quantify the sensitivity of the Hama roughness function to systematic changes in surface anisotropy. To this end, hot-wire anemometry measurements are acquired at three different freestream velocities under zero-pressure gradient conditions for each surface. Relative to the isotropic case, an increase in the turbulence intensity is seen in the near-wall region for the anisotropic cases. As expected, decreasing ESx leads to a lower mean momentum deficit which confirms the findings of many previous experimental and numerical studies. However, results also suggest that ESy plays an important role. Even for the mildly anisotropic case considered here, the roughness function is seen to vary by up to 15% as ESy is reduced while ESx is held constant. In addition, regions of high streamwise dispersive velocity are seen to extend further into the flow field as ESy reduces. These observations suggest that existing models for drag prediction need to be modified to account for surface anisotropy.