부산대학교 도서관

Measurements with both two-dimensional (2D) two-component and three-component stereo particle image velocimetry (PIV) and computation in 2D and three-dimensional (3D) using Reynolds stress turbulence model with commercial code are carried out in a square duct backward-facing step (BFS) in a turbulent water flow at three Reynolds numbers of about 12,000, 21,000, and 55,000 based on the step height h and the inlet streamwise maximum mean velocity [U.sub.0]. The reattachment locations measured at a distance of [DELTA]y=0.0322h from the wall are 5.3h, 5.6h, and 5.7h, respectively. The inlet flow condition is fully developed duct flow before the step change with the expansion ratio of 1.2. PIV results show that the mean velocity, root mean square (rms) velocity profiles, and Reynolds shear stress profiles in all the experimental flow cases are almost identical in the separated shear-layer region when they are nondimensionalized by [U.sub.0]. The sidewall effect of the square BFS flow is analyzed by comparing the experimental statistics with direct numerical simulation (DNS) and Reynolds stress model (RSM) data. For this purpose, the simulation is carried out for both 2D BFS and for square BFS having the same geometry in the 3D case as the experimental case at the lowest Reynolds number. A clear difference is observed in rms and Reynolds shear stress profiles between square BFS experimental results and DNS results in 2D channel in the spanwise direction. The spanwise rms velocity difference is about 30%, with experimental tests showing higher values than DNS, while in contrast, turbulence intensities in streamwise and vertical directions show slightly lower values than DNS. However, with the modeling, the turbulence statistical differences between 2D and 3D RSM cases are very modest. The square BFS indicates 0.5h- 1.5h smaller reattachment distances than the reattachment lengths of 2D flow cases. [DOI: 10.1115/1.2746896]