Direct numerical simulation of turbulent channel flow over random rough surfaces

Abstract

Direct numerical simulation (DNS) of flow in a turbulent channel with a random rough wall is performed at $Re_{\tau}=400$ and $600$. The roughness geometry corresponds to the experiments of Flack and Schultz (personal communication). The rough surface is used on the bottom wall of the channel. DNS of a smooth channel flow and a rod-roughened channel flow are also performed at $Re_{\tau}=400$ for validation and comparison. The skin friction coefficient of the random-rough channel shows good agreement with the experimental results of Flack and Schultz. Due to the roughness, the slip velocity on the rough wall increases while the mean velocity decreases in the log-law region. The streamwise and spanwise velocity fluctuations are enhanced near the rough wall. The pressure fluctuations show a significant increase in the roughness layer and exhibit a good collapse with the smooth wall in the outer layer. The streamwise mean momentum balance shows that pressure and viscous stress gradients are induced in the roughness layer, and the gradients are amplified at higher $Re_{\tau}$. The statistics of wall-shear stress fluctuations in the peak (above the mean height location) and valley (below the mean height location) regions are examined. The results indicate that reverse flow mainly occurs in the valley regions of the random roughness, and is enhanced at higher $Re_{\tau}$, but is not as strong as the recirculation within the cavities in the rod-roughened channel. The probability distribution function of wall-shear stress shows a better collapse after subtracting the mean and normalizing by the root-mean-squared value. The distribution tail is widened by the random roughness, implying that the probability of extreme events is increased. The probability of extreme events in the random-rough channel increases with increasing $Re_{\tau}$, in accordance with previous studies on smooth-wall flows.