| The predictive capability of implementations of the Spalart-Allmaras, the kappa - epsilon, and the kappa - o turbulence models was assessed for subsonic compressible internal flows using implied friction coefficient to determine simulation accuracy. The sensitivity of the solution to the near-wall node's location, measured in the computational domain by the normalized wall distance, y+, is also considered. CFD simulations were compared to the experimental pipe flow data of Keenan and Neumann. The results show that when the viscous sublayer is resolved, the kappa-epsilon model with Enhanced Wall Treatment produces the best results, generally within 3% of the data. When law-of-the wall implementations are used, the Spalart-Allmaras and kappa - epsilon models show similar predictive performance, within 4% of the data. However, the one-equation Spalart-Allmaras model produces results in roughly half the computational time and is less sensitive to variation with first node wall distance. The standard kappa - o models failed to converge probably because of the high free-stream cell Reynolds numbers, while the kappa - o Shear Stress Transport model showed poorer results with average errors of about 11%. |
