Is comparison with experimental data a reasonable method of validating computational models

In the case of laminar flow, once all property variations are included, properly modelled and correctly solved a good agreement with experimental data means that numerical model has been validated. However, for turbulent flows the difference between experimental and numerical results could be the effect of an inadequate model or the use of inappropriate boundary conditions. A better agreement with the experiment, might be obtained by "improving" the model, or by obtaining "better" boundary conditions or a combination of the two. This issue is explored in the paper using the example of LES of buoyancy-driven flow in a tall rectangular cavity. Based on comparison with experimental data, it is shown that introducing heat losses through the top wall in simulations brings the numerical results closer to the experimental data. However, an improved subgrid model with all non-isothermal walls being adiabatic also brings numerical results and experimental data closer, at least near the hot wall. Therefore, experimental results cannot be used to validate numerical model of turbulent flows with heat transfer unless thermal boundary conditions can be established with reasonable accuracy.