Experimental and numerical investigation of developing turbulent flow over a wavy wall
Martins Segunda, Vinicius
Turbulent flow over a wavy wall in a horizontal channel is investigated by experimental and numerical methods. The thorough problem understanding can advance turbulent flow physics knowledge for separating and reattaching flows. Another important consideration is the performance evaluation of mathematical models used in computational fluid dynamics (CFD) codes to predict the flow characteristics. This study explores numerical models because they are critically important to the design and performance evaluation of engineering systems. The experimental data are obtained to provide repository data and more insights into the flow physics considering both the flow development and fully periodic regions. A channel with a wavy bottom wall is considered for this study, and its main characteristic is a value of 10 for the ratio between the wave length and wave amplitude. A high-resolution particle image velocimetry (PIV) system is used to obtain detailed measurements of velocity at Reynolds number of 5040, 8400, 10700 and 13040 in both the developing and fully periodic regions. The numerical simulations are performed with a commercial CFD code using four eddy viscosity turbulence models and three Second-Moment Closure (SMC) turbulence models. This work assessed the predictive accuracy of a total of seven turbulence models. The experimental study covered a lack of data for the flow development within the waves, prior the periodic condition region, and it supported the turbulence models evaluation. The experiments provided features of the flow such as the recirculation regions, Reynolds stresses, and turbulent kinetic energy production at different channel locations. A comprehensive comparison between models and experimental data revealed a significant dependency on the turbulence model formulation and on the wall treatment selection for the flow development and fully periodic regions predictions.
Wavy wall; Turbulent flow; RANS model; PIV data