Numerical simulation of turbulent convective heat transfer in square ribbed ducts

This article reports on an investigation on numerical prediction of thermal characteristics of a certain type of duct. The ducts consider ed have rib turbulators to enhance the heat transfer rate. The calculation method consi sts of a low Re number turbulence model and two methods for determining the turbulent Reynolds stresses, namely, a simple eddy viscosity model (EVM) [ I] and an explicit algebraic stress model (EASM) [2]. The model development is carried out to make the original EASM consistent with the low Re number k-epsilon turbulence model applied. A certain Method is developed to deal with the decoupling of the velocity and Reynolds stress fields in the collo cated grid arrangement that is chosen in this study. The SIMPLEC algorithm handles the pressure-velocity coupling. The computations are performed with the assumption of fully developed pe, iodic conditions. These models ale u sed to predict the convective turbulent forced convection in different test cases and the results are compared with experiments. A ribbed duct with tw o ribs on opposite walls is chosen and the obtained results including the m ean thermal characteristics of the considered duct are compared with an exp erimental correlation. Two further duct configurations, identical to an exp erimental setup, are then computed. These experimental cases are chosen bec ause detailed thermal-hydraulic information is available and then local com parisons between the two prediction models and experimental results ave pos sible. The calculated mean and local thermal-hydraulic values are compared with corresponding experimental data and the prediction capabilities of the two turbulence models (EVM) and EASM) are discussed The results show that the EASM has some superiority over the EVM in the prediction of the velocit y field structure, but the mean thermal predictions are not very different. There are also some important features of the flow field, which are not re vealed by the EVM calculations. However, the required CPU times are conside rably higher for the EASM case.