Computation fluid dynamics on drag reduction II

Abstract

Concentrated research activities are done on drag reduction and one of the main subjects being investigated frequently is the presence of dimple on rigid surfaces. Both the nature of flow and the advantage of drag reduction embedded within have been studied by Isaev et al. (2000) & A.M Cary (1979) [7][11] have been studied with experiments and computational fluid dynamics. In this project, a number of Reynolds-Average Navier-Stokes (RANS) models are verified with benchmark values of Direct Numerical Solutions (DNS) provided by the past research to identify appropriate RANS models which allow the simulation of dimpled flow with specific Reynolds number and initial conditions in a 3D channel. Five RANS turbulence models: Spalart-Allmaras (Vorticity-based production); Spalart-Allmaras (Strain / Vorticity-based production); k-ε Standard (Enhanced Wall Treatment); k-ε Realizable (Enhanced Wall Treatment) and k-ω Standard (Shear Flow Correction), are chosen after comparing with DNS data published by Moser et al (1999). [20] Subsequently, the single and multiple dimpled channel flows are modeled and simulated by the five RANS models chosen in the verification stage with periodic boundary condition, a Reynolds number of 20000, and a three-dimensional channel of L = 1.5 m, B = 0.454 m and H = 1 m. Both the flow pattern in the form of velocity contours and the quantitative drag coefficient are extracted and compared in several ways: single dimple along the wall with depth (h/D = 0.21); multiple (three) dimples along the wall with different separation distances between adjacent dimples; single dimple but with several other h/D ratios (i.e. 0.25, 0.14, 0.10, 0.08). The post-processed results reveal the wide existence of reversal flow as an eddy current that contributes to the reduction of viscous drag on the surface. However, the induced pressure drag exerts great adverse effects on drag reduction capability and efficacy of dimpled surface. Several useful observations and inferences are made through the investigation process, such as smaller separation distance between dimples facilitate the drag reduction mechanisms. Lastly, the grid size and RANS models are subjected to further verification to ensure the observations and calculations are indeed trustworthy and relatively accurate by studying skin friction coefficient and comparing to empirical formula; also by simulating more complex dimpled flow with different Reynolds number regimes. The result establishes agreements in friction coefficient while the simulation of complex flow doe s not. Further work can be done to optimize dimple separation distance or h/D ratio under certain Reynolds number by RANS.