Experimental and numerical studies of turbulent cross-flow in a staggered tube bundle

An experimental investigation was performed on turbulent cross-flow in a staggered tube bundle with transverse and longitudinal pitch-to-diameter ratios of 3.8 and 2.1, respectively. A particle image velocimetry technique (PIV) was employed to obtain detailed measurements in the bundle at inlet-velocity-based Reynolds numbers of 4800, 9300, and 14400. Quantities reported include mean velocities, turbulence intensities, Reynolds stresses, various terms in the Reynolds average Navier Stokes (RANS) equations, triple correlations of fluctuating velocity components, and energy budget terms. The results reveal higher shear rates and extremely high turbulence levels in the wake compare to other regions. The flow evolves fairly rapidly and becomes spatially periodic in the streamwise direction after a relatively short distance. The flow exhibits strong Reynolds number dependence in the developing region but no significant Reynolds number effects are observed in the spatially periodic region. The pressure gradient terms in the streamwise and transverse RANS equations are nearly balanced by the Reynolds stress terms in the recirculation zones, and by the convective terms outside the recirculation region.; The results also revealed extremely high levels of turbulence production by the normal stresses, as well as regions of negative turbulence production. The convective transport by mean flow and turbulent diffusion were observed to be significantly higher than in classical turbulent boundary layers. As a result, turbulence production is generally not in equilibrium with its dissipation rate.; A commercial CFD code, CFX5, is used to predict the turbulent flow in the bundle. The turbulent models used comprise the k--epsilon, k--o, a k--o-based shear stress transport model, and an epsilon-based second moment closure with a scalable wall function. The results show that none of the turbulence models was able to consistently reproduce the mean and turbulent quantities reasonably well. However, the o-based models predicted the mean velocities better in the developing region while the epsilon-based models gave better results in the region where the flow is spatially periodic. The poor performance of the two-equation turbulence models may partly be explained by the differences between the flow field in the tube bundle and the generic flows used to calibrate the turbulence models. In addition, the production of turbulent kinetic energy and its dissipation rate are not in equilibrium, as assumed in many basic eddy viscosity turbulent models. The negative production is also not accounted for in the models.