Accuracy comparison of a standard CFD code for the thermal analysis of non-simple geometries with baseline experiments

The aim of the research is to investigate the temperature prediction accuracy of Computational Fluid Dynamics (CFD) analyses for relatively simple model geometries (consisting of more than one component) in a natural and a mixed convection environment. A two-step approach has been taken. First, a baseline set-up was built consisting of a heated cubic component inside a (Perspex¿) PMMA enclosure for which all material properties and boundary conditions are known with sufficient accuracy. This baseline experiment is conceived to ensure that the CFD code predicts the temperature and velocity fields with an accuracy better than say 3%. The second step was to add complexities one by one. Temperature predictions for mixed convection of two cubic components in a low-velocity channel have been investigated, as well as the influence of cables, grilles etc. In particular the accuracy of the heat transfer from a heated upstream to an unpowered downstream component has been attended. Smoke visualization and 3D PIV measurements have been used to qualitatively and quantitatively check the numerically obtained velocity fields.