Low-Reynolds number mixing ventilation flows: Impact of physical and numerical diffusion on flow and dispersion
computational fluid dynamics (CFD), numerical and physical diffusion, mixing ventilation, contaminant dispersion, artificial diffusion
Quality assurance in computational fluid dynamics (CFD) is essential for an accurate and reliable assessment of complex indoor airflow. Two important aspects are the limitation of numerical diffusion and the appropriate choice of inlet conditions to ensure the correct amount of physical diffusion. This paper presents an assessment of the impact of both numerical and physical diffusion on the predicted flow patterns and contaminant distribution in steady Reynolds-averaged Navier–Stokes (RANS) CFD simulations of mixing ventilation at a low slot Reynolds number (Re≈2,500). The simulations are performed on five different grids and with three different spatial discretization schemes; i.e. first-order upwind (FOU), second-order upwind (SOU) and QUICK. The impact of physical diffusion is assessed by varying the inlet turbulence intensity (TI) that is often less known in practice. The analysis shows that: (1) excessive numerical and physical diffusion leads to erroneous results in terms of delayed detachment of the wall jet and locally decreased velocity gradients; (2) excessive numerical diffusion by FOU schemes leads to deviations (up to 100%) in mean velocity and concentration, even on very high-resolution grids; (3) difference between SOU and FOU on the coarsest grid is larger than difference between SOU on coarsest grid and SOU on 22 times finer grid; (4) imposing TI values from 1% to 100% at the inlet results in very different flow patterns (enhanced or delayed detachment of wall jet) and different contaminant concentrations (deviations up to 40%); (5) impact of physical diffusion on contaminant transport can markedly differ from that of numerical diffusion.
Tsinghua University Press
Twan van Hooff, Bert Blocken. Low-Reynolds number mixing ventilation flows: Impact of physical and numerical diffusion on flow and dispersion. Build Simul, 2017, 10(4): 589–606.