DNS of low Reynolds number turbulent flows in dimpled channels

Direct numerical simulation (DNS) is performed to study turbulent flows over dimpled surfaces in a channel. Results on mean field and second-order quantities are obtained. ‘Horseshoe’ vortices can be observed in the dimples of sparse arrays. As inter-dimple separation is reduced, the ‘feet’ of the horseshoe vortices are gradually lifted off the dimple surface, and the resulting flow structures in the cavities become flattened and stretched to become something akin to two-dimensional separation bubbles. At the higher dimple density, the stream traces near the surface also develop a distinct formation similar to what had been observed in earlier Reynolds-averaged Navier–Stokes (RANS) simulations (Isaev, S.A., Leont'ev, A.I. and Baranov, P.A., 2000, Technical Physics Letters, 26, 15; Lin, Y.L., Shih, T.I.-P. and Chyu, M.K., 1999, ASME paper, 99-GT-263; Lin, Y.L. Shih, T.I.-P., 2001, International Journal of Transfer Phenomena, 3, 1). Regions of high turbulence intensity are found above the downstream half of the dimples and along their side edges. These regions coincide with the locations of vortex shedding found in the experiments of Ligrani et al. (2001, Physics of Fluids, 13, 3442) and the locations of vorticity concentrations observed in Park et al. (2004, Numerical Heat Transfer, Part A (Applications), 45(1), 1) and Won and Ligrani (2004, Numerical Heat Transfer, Part A (Applications), 46(6), 549). For a fixed mean pressure gradient, it is observed that the flow rates through the channels are reduced by the presence of dimples. This indicates that the dimpled channels we have studied so far have larger drag than flat-wall channels. Computed friction coefficients for dimpled channels also confirmed the conclusion.

Keywords:

• Dimple
• DNS
• turbulent channel flow
• rough wall
• heat transfer
• mixing enhancement

Acknowledgments

We wish to thank Professor P.M. Ligrani and Professor N. Kornev for sharing their knowledge and information with us in the course of this work. The use of the Hydra cluster of Singapore-MIT Alliance (SMA) and the Neumann cluster of the Institute for High Performance Computing (IHPC) is gratefully acknowledged.