Applications of a higher-order bounded numerical scheme to turbulent flows

• Published in: International journal for numerical methods in fluids 2001-02, Vol.35 (4), p.371-394
• Main Authors: Song, B., Liu, G. R., Amano, R. S.
• Format: Article
• Language: English
• Subjects: bounded numerical scheme; Computational methods in fluid dynamics; Computer simulation; Exact sciences and technology; Fluid dynamics; Fundamental areas of phenomenology (including applications); Interpolation; Kinetic energy; Mathematical models; Method of moments; Physics; Reynolds number; turbulence model; Turbulence simulation and modeling; turbulent flow; Turbulent flows, convection, and heat transfer
• ISSN: 0271-2091; 1097-0363
• DOI: 10.1002/1097-0363(20010228)35:4<371::AID-FLD92>3.0.CO;2-J

This paper applies the higher‐order bounded numerical scheme Weighted Average Coefficients Ensuring Boundedness (WACEB) to simulate two‐ and three‐dimensional turbulent flows. In the scheme, a weighted average formulation is used for interpolating the variables at cell faces and the weighted average coefficients are determined from a normalized variable formulation and total variation diminishing (TVD) constraints to ensure the boundedness of the solution. The scheme is applied to two turbulent flow problems: (1) two‐dimensional turbulent flow around a blunt plate; and (2) three‐dimensional turbulent flow inside a mildly curved U‐bend. In the present study, turbulence is evaluated by using a low‐Reynolds number version of the k–ω model. For the flow simulation, the QUICK scheme is applied to the momentum equations while either the WACEB scheme (Method 1) or the UPWIND scheme (Method 2) is used for the turbulence equations. The present study shows that the WACEB scheme has at least second‐order accuracy while ensuring boundedness of the solutions. The present numerical study for a pure convection problem shows that the ‘TVD’ slope ranges from 2 to 4. For the turbulent recirculating flow, two different mixed procedures (Method 1 and Method 2) produce a substantial difference for the mean velocities as well as for the turbulence kinetic energy. Method 1 predicts better results than Method 2 does, comparing the analytical solution and the experimental data. For the turbulent flow inside the mildly curved U‐bend, although the predictions of velocity distributions with two procedures are very close, a noticeable difference of turbulence kinetic energy is exhibited. It is noticed that the discrepancy exists between numerical results and the experimental data. The reason is the limit of the two‐equation turbulence model to such complex turbulent flows with extra strain‐rates. Copyright © 2001 John Wiley & Sons, Ltd.