Synthesis of a CFD Benchmark for the Thermal Mixing in a Sharp Corner T-Junction With a Wall

Abstract

This article reports a CFD-benchmark with the purpose of validating different turbulence modelling approaches for the transient heat transfer due to mixing of hot and cold flow in a T-junction including the wall. This validation exercise has been carried out within the MOTHER project. In the framework of the project, new experiments were performed with a novel measurement sensor allowing the measurements of the fluctuating wall temperature inside the solid pipe wall. The tests were performed for two different Reynolds numbers (Re) 40000 and 60000 and for two different T-junction geometries; a sharp corner and a round corner. The present article reports the synthesis of the CFD validation for a sharp corner T-junction for Re = 40 000. The CFD validation study has been performed using four different CFD softwares, namely STAR-CCM+, Code_Saturne, LESOCC2 and Fluent. In addition, five different turbulence models i.e. wall-function Large Eddy Simulation (LES), Deatched Eddy Simulation (DES), Partially Resolved Numerical Simulation (PRNS), Unsteady Reynolds Averaged Navier-Stokes URANS and RANS were used to perform the CFD computations. The validation exercise has shown that LES gives the best agreement with the experimental data followed by hybrid (LES/RANS), URANS and RANS models, respectively. The velocity and the thermal fields in the fluid region are correctly predicted by the proper use of the LES modelling, whereas, the accurate prediction of the thermal field in the solid requires very long sampling time in order to achieve a statistically converged solution, which of course requires an enormous computational power. Therefore, the statistical convergence of the thermal field in the solid has been found to be a bottleneck in order to accurately predict the temperature fluctuations in the wall. However, measuring the small amplitude temperature fluctuations is also associated with an uncertainty so the disagreement between CFD and measurements (of the order of 10 %) can also be attributed, in part, to uncertainties in the measurements.