A dynamic method for prescribing realistic inflow boundary conditions is presented for simulations of spatially developing turbulent boundary layers. The approach is based on the rescaling–recycling method proposed by Lund, Wu & Squires (J. Comput. Phys, vol. 140, 1998, pp. 233–258) and the multi-scale method developed by Araya, Jansen & Castillo (J. Turbul., vol. 10, no. 36, 2009, pp. 1–33). The rescaling process requires prior knowledge about how the velocity and length scales are related between the inlet and recycle stations. Here a dynamic approach is proposed in which such information is deduced dynamically by involving an additional plane, the so-called test plane located between the inlet and recycle stations. The approach distinguishes between the inner and outer regions of the boundary layer and enables the use of multiple velocity scales. This flexibility allows applications to boundary layer flows with pressure gradients and avoids the need to prescribe empirically the friction velocity and other flow parameters at the inlet of the domain. The dynamic method is tested in direct numerical simulations of zero, favourable and adverse pressure gradient flows. The dynamically obtained scaling exponents for the downstream evolution of boundary layer parameters are found to fluctuate in time, but on average they agree with the expected values for zero, favourable and adverse pressure gradient flows. Comparisons of the results with data from experiments, and from other direct numerical simulations that use much longer computational domains to capture laminar-to-turbulence transition, demonstrate the suitability of the proposed dynamic method.
A pioneer direct numerical simulation (DNS) of a turbulent boundary layer at $R{e}_{\theta } = 2077{{\unicode{x2013}}}2439$ , was performed, on a rough surface and with a zero pressure gradient (ZPG). The boundary layer was subjected to transitional, 24-grit sandpaper surface roughness, with a roughness parameter of ${k}^{+ } \simeq 11$ . The computational method involves a synergy of the dynamic multi-scale approach devised by Araya et al. (2011) for prescribing inlet turbulent boundary conditions and a new methodology for mapping high-resolution topographical surface data into a computational fluid dynamics (CFD) environment. It is shown here that the dynamic multi-scale approach can be successfully extended to simulations which incorporate surface roughness. The DNS results demonstrate good agreement with the laser Doppler anemometry (LDA) measurements performed by Brzek et al. (2008) and Schultz & Flack (2003) under similar conditions in terms of mean velocity profiles, Reynolds stresses and flow parameters, such as the skin friction coefficient, boundary and momentum thicknesses. Further, it is demonstrated that the effects of the surface roughness on the Reynolds stresses, at the values of $R{e}_{\theta } = 2077{{\unicode{x2013}}}2439$ , are scale-dependent. Roughness effects were mainly manifested up to $y/ \delta \approx 0. 1$ . Generally speaking, it was observed that inner peak values of Reynolds stresses increased when considering outer units. However, decreases were seen in inner units. In the outer region, the most significant differences between the present DNS smooth and rough cases were computed in the wall-normal component $\langle {v}^{\prime 2} \rangle $ of the Reynolds stresses and in the Reynolds shear stresses $\langle {u}^{\prime } {v}^{\prime } \rangle $ in outer units. From the resulting flow fields a proper orthogonal decomposition (POD) analysis is performed and the effects of the surface roughness are distinctly observed in the most energetic POD modes. The POD analysis shows that the surface roughness causes a redistribution of the kinetic energy amongst the POD modes with energy being shifted from low-order to high-order modes in the rough case versus the smooth case. Also, the roughness causes a marked decrease in the characteristic wavelengths observed in the POD modes, particularly in the streamwise component of the velocity field. Low-order modes of the streamwise component demonstrated characteristic wavelengths of the order of $3\delta $ in the smooth case, whereas the same modes for the rough case demonstrated characteristic wavelengths of only $\delta $ .
Direct numerical simulation of highly accelerated turbulent boundary layers (TBLs) reveals that the Reynolds shear stress, $\overline{u^{\prime }v^{\prime }}^{+}$ , monotonically decreases downstream and exhibits a logarithmic behaviour (e.g. $-\overline{u^{\prime }v^{\prime }}^{+}=-(1/A_{uv})\ln y^{+}+B_{uv}$ ) in the mesolayer region (e.g. $50\leqslant y^{+}\leqslant 170$ ). The thickness of the log layer of $\overline{u^{\prime }v^{\prime }}^{+}$ increases with the streamwise distance and with the pressure gradient strength, extending over a large portion of the TBL thickness (up to 55 %). Simulations reveal that $V^{+}\,\partial U^{+}/\partial y^{+}\sim 1/y^{+}\sim \partial \overline{u^{\prime }v^{\prime }}^{+}/\partial y^{+}$ , resulting in a logarithmic $\overline{u^{\prime }v^{\prime }}^{+}$ profile. Also, $V^{+}\sim -y^{+}$ is no longer negligible as in zero-pressure-gradient (ZPG) flows. Other experimental/numerical data at similar favourable-pressure-gradient (FPG) strengths also show the presence of a log region in $\overline{u^{\prime }v^{\prime }}^{+}$ . This log region in $\overline{u^{\prime }v^{\prime }}^{+}$ is larger in sink flows than in other spatially developing FPG flows. The latter flows exhibit the presence of a small power-law region in $\overline{u^{\prime }v^{\prime }}^{+}$ , which is non-existent in sink flows.
A zero-pressure-gradient turbulent boundary layer flowing over a transitionally rough surface (24-grit sandpaper) with $k^{+}\approx 11$ and a momentum-thickness Reynolds number of approximately 2400 is studied using direct numerical simulation (DNS). Heat transfer between the isothermal rough surface and the turbulent flow with molecular Prandtl number $Pr=0.71$ is simulated. The dynamic multiscale approach developed by Arayaet al.(J. Fluid Mech., vol. 670, 2011, pp. 581–605) is employed to prescribe realistic time-dependent thermal inflow boundary conditions. In general, the rough surface reduces mean and fluctuating temperature profiles with respect to the smooth surface flow when normalized by Wang & Castillo (J. Turbul., vol. 4, 2003, 006) inner/outer scaling. It is shown that the Reynolds analogy does not hold for $y^{+}<9$ . In this region the value of the turbulent Prandtl number departs substantially from unity. Above this region the Reynolds analogy is only approximately valid, with the turbulent Prandtl number decreasing from 1 to 0.7 across the boundary layer for rough and smooth walls. In comparison with the smooth-wall case, the turbulent transport of heat per unit mass, $\overline{v^{\prime }v^{\prime }{\it\theta}^{\prime }}$ , towards the wall is enhanced in the buffer layer, but the transport of $\overline{v^{\prime }v^{\prime }{\it\theta}^{\prime }}$ away from the wall is reduced in the outer layer for the rough case; similar behaviour is found for the vertical transport of turbulent momentum per unit mass, $\overline{v^{\prime }u^{\prime }v^{\prime }}$ . Above the roughness sublayer (3 $k$ –5 $k$ ) it is found that most of the temperature field statistics, including higher-order moments and conditional averages, are highly similar for the smooth and rough surface flow, showing that the Townsend’s Reynolds number similarity hypothesis applies for the thermal field as well as the velocity field for the Reynolds number and $k^{+}$ considered in this study.
1 January 2016 in European Journal of Mechanics - B/Fluids