Special Seminar: Spring 2013
Speaker: Dr. Stefano Leonardi (University of Puerto Rico at Mayagüez)
Title: "Direct Numerical Simulations of Flows over Rough Walls"
Date: Tuesday, February 26, 2013
Time: 12:00 p.m. (Noon)
Location: Gilman Hall 132
Turbulent flows over rough surfaces are often encountered in practice; in the atmosphere, the underlying surface is usually rough, while, in an engineering context, pipes and ducts cannot be regarded as hydraulically smooth, especially at high Reynolds numbers. Rough surfaces may be used to enhance heat transfer, albeit at the expense of increasing the drag; alternately, the roughness geometry may be selected so as to decrease the drag, e.g. by using riblets, or delay transition. Roughness can seriously degrade the performance of airfoils, wings and turbomachinery blades. In the last two decades several numerical and experimental studies have tried to explain the physics of the flow over rough surfaces. Most of the studies have dealt with idealized roughness: square bars, meshes, rods, cubes, chopped transverse bars. However, an understanding of idealized roughness may not properly extrapolate to more practical cases of highly irregular surface roughness. Computations of a turbulent channel flow with a rough wall made of 3D wedges of random height will be discussed and compared to our previous DNS studies of the flow over square bars and staggered arrays of cubes. Results have been obtained using a finite difference code combined with the immersed boundary method. The code is parallel and typical grids have 50-100 million cells.
Figure 1 Sketch of the Random Rough Wall
The surface drag is the sum of the frictional and pressure drag. It is shown that the surface drag is predominantly form drag, which is greatest at an area coverage around 15% for both staggered cubes and square bars. Frictional and form drag present a similar dependence with the roughness density.
Two-point velocity correlations, with the fixed point at several locations within one roughness wavelength show a decreased coherence in the streamwise direction with respect to a smooth wall. On the other hand, the coherence in the wall normal direction has increased. This is consistent with an increased communication between the flow in the roughness layer and the outer layer and with an increase in heat transfer or passive scalar mixing. The increase mixing in wall normal direction makes the flow more isotropic. This is examined in the context of its anisotropy invariants. The Reynolds stress budgets showed a redistribution of energy between the streamwise velocity to the normal wall and spanwise velocity components. The main sink in the budget of <uu> is not the dissipation but rather the pressure-velocity correlation and the turbulent diffusion. Roughness seems to affect the overlying flow through the pressure and wall normal velocity. In fact, a satisfactory collapse of the data is achieved when the roughness function is plotted against the wall–normal velocity rms at the roughness crests. This can be a new way to model roughness in regional scale simulations.