6. Summary of results

Since the reference solution is not analytical, but numerical (obtained by 3D modeling), the differences observed (from \(\text{1\%}\) to \(\text{3\%}\)) can be considered reasonable. To obtain a better correspondence between the 3D and TUYAU solutions, it would be appropriate to model the parts straight over a longer length, and to adopt a finer mesh for each of the models. This was not done as part of this test, to keep execution times reasonable.