TY - JOUR
T1 - A comparison of RANS models used for CFD prediction of turbulent flow and heat transfer in rough and smooth channels
AU - Kadivar, Mohammadreza
AU - Tormey, David
AU - McGranaghan, Gerard
N1 - Publisher Copyright:
© 2023 The Authors
PY - 2023/11
Y1 - 2023/11
N2 - CFD of convective heat transfer remains a challenge, especially in channels with irregular rough walls due to the complex fluid dynamics over and between these roughness elements. This study investigates the performance of four commonly utilised RANS CFD models, namely Spalart-Allmaras (SA), Realisable k-ε, SST k-ω, and Reynolds Stress Model (RSM) in estimating the velocity and temperature profile as well as the skin friction coefficient and Nusselt number in smooth and rough channels. The CFD results were compared with DNS and experimental data in the literature. The roughness-resolving approach was used for all models to resolve the realistic geometry of irregular roughness in channel flows. The smooth-channel flow study was conducted at Reτ = 500, 1000, 2000, and 5000 and Pr = 0.71. Over this range, compared to smooth-wall DNS data from literature, the RSM showed consistent and satisfactory performance in estimating the velocity and temperature profiles within 2.54% and 4.21%, respectively; leading to the skin friction coefficient and Nusselt number in the smooth channels being predicted very well within 1.94% and 2.03%, respectively. The rough-channel flow study was performed at Reτ = 240, 360, 540, 720, 1000 and Pr = 1, with computational cells 7.8 times that of the smooth channel to resolve the irregular roughness geometry. Over this range, the Realisable k-ε showed the best performance in estimating the roughness function within 6.85%, while the results of the other models were wider at uncertainty levels of up to 36.9%. Similarly, the Realisable k-ε demonstrated the best performance in estimating the skin friction and Nusselt number in rough channels within 2.76% and 9.50%, respectively, while the performance of the other models was to a wider latitude ranging between 15.62% to 26.34%. In the present study, the Realisable k-ε with enhanced wall treatment showed the best capabilities in capturing the mean flow characteristics of both smooth and rough channels; however, its usage should be approached with caution in flows at higher Reynolds numbers. RANS models are also discussed for forced convection heat transfer at high Reynolds numbers. The results of this study should be useful in CFD studies on the effects of roughness in complex geometries without the need for LES and DNS.
AB - CFD of convective heat transfer remains a challenge, especially in channels with irregular rough walls due to the complex fluid dynamics over and between these roughness elements. This study investigates the performance of four commonly utilised RANS CFD models, namely Spalart-Allmaras (SA), Realisable k-ε, SST k-ω, and Reynolds Stress Model (RSM) in estimating the velocity and temperature profile as well as the skin friction coefficient and Nusselt number in smooth and rough channels. The CFD results were compared with DNS and experimental data in the literature. The roughness-resolving approach was used for all models to resolve the realistic geometry of irregular roughness in channel flows. The smooth-channel flow study was conducted at Reτ = 500, 1000, 2000, and 5000 and Pr = 0.71. Over this range, compared to smooth-wall DNS data from literature, the RSM showed consistent and satisfactory performance in estimating the velocity and temperature profiles within 2.54% and 4.21%, respectively; leading to the skin friction coefficient and Nusselt number in the smooth channels being predicted very well within 1.94% and 2.03%, respectively. The rough-channel flow study was performed at Reτ = 240, 360, 540, 720, 1000 and Pr = 1, with computational cells 7.8 times that of the smooth channel to resolve the irregular roughness geometry. Over this range, the Realisable k-ε showed the best performance in estimating the roughness function within 6.85%, while the results of the other models were wider at uncertainty levels of up to 36.9%. Similarly, the Realisable k-ε demonstrated the best performance in estimating the skin friction and Nusselt number in rough channels within 2.76% and 9.50%, respectively, while the performance of the other models was to a wider latitude ranging between 15.62% to 26.34%. In the present study, the Realisable k-ε with enhanced wall treatment showed the best capabilities in capturing the mean flow characteristics of both smooth and rough channels; however, its usage should be approached with caution in flows at higher Reynolds numbers. RANS models are also discussed for forced convection heat transfer at high Reynolds numbers. The results of this study should be useful in CFD studies on the effects of roughness in complex geometries without the need for LES and DNS.
KW - Computational fluid dynamics (CFD)
KW - Heat transfer
KW - RANS
KW - Roughness
KW - Turbulent flow
UR - http://www.scopus.com/inward/record.url?scp=85163172116&partnerID=8YFLogxK
U2 - 10.1016/j.ijft.2023.100399
DO - 10.1016/j.ijft.2023.100399
M3 - Article
AN - SCOPUS:85163172116
SN - 2666-2027
VL - 20
JO - International Journal of Thermofluids
JF - International Journal of Thermofluids
M1 - 100399
ER -