MULTIOBJECTIVE OPTIMIZATION OF TURBULENT HEAT TRANSFER PERFORMANCE WITH RESPECT TO TEARDROP-SHAPED DIMPLE ARRANGEMENTS AND FLOW-PULSATION CONDITIONS BY USING RANS AND LES

This study performed a multiobjective optimization of turbulent heat transfer in a channel with teardrop-shaped dimples on one wall and with flow pulsation. Four design parameters consisted of inclination angle and slanted spanwise shift of teardrop-shaped dimples as surface-geometry parameters and flow pulsation frequency and root means square (rms) velocity amplitude as flow-pulsation parameters. Objective functions were the mean Nusselt number of the dimpled surface and the friction factor of the channel. By using Latin hypercube sampling, 50 combinations of the four design parameters were selected, and Reynolds-averaged Navier-Stokes (RANS) simulation using shear stress transport (SST) <i>k-ω </i>model was performed. Using response surfaces of the Kriging model and a genetic algorithm, the Pareto solutions were found. Of the Pareto solutions, four representative combinations of the design parameters were selected and verified by performing again the RANS simulation using the grid system newly meshed for the exact surface geometries. The RANS results for the exact surface geometries showed the differences from the Pareto solutions within 2.0% and 2.9% for the mean Nusselt number and the friction factor, respectively. Further verification by the large Eddy simulation showed the differences from the RANS results were within 3.3% and 5.2% for the mean Nusselt number and the friction factor, respectively. Multiple linear regression analysis for the Pareto solutions identified the inclination angle of the dimples had the strongest effects on both the mean Nusselt number and the friction factor. The surface geometry parameters had stronger effects on the results than the flow-pulsation parameters.