question about turbulence model selection and sensitivity

karananand · February 25, 2010, 20:32

Hi,
This might be a simple question, but I have less experience with turbulence models. To select the best turbulence model, do I have to check for mesh independence for each turbulence model? In simple words, say if I have an independent grid when I first tested my model with K-e RNG model, do I have to test for mesh sensitivity again for a different turbulence model (say k-w or k-e SST)?

Also I have another question. When we talk about low RE turbulence models, how low (approx magnitude) should the Re be, say for impingement flow via a circular nozzle (cylindrical pipe in 3D) Re_transition = 2000?
And, since it is turbulence model, can i safely assume low Re turbulence model for Re <= 8000?

RossFS · February 26, 2010, 05:41

If the Reynolds number of the flow is not very high, you could get away with using some of the best turbulence models your CFD software has as the computational requirements should not be that much.

Re =<8000 is near the bottom end of Re's from what I gather.

Note that the above two comments are based on only some basic research into DNS, k-epsilon, k-omega and SST-SAS models and a little in ANSYS CFX. (4th year engineering thesis)

February 25, 2010, 20:32	question about turbulence model selection and sensitivity	#1
karananand New Member Karan Anand Join Date: Feb 2010 Posts: 23 Rep Power: 16	Hi, This might be a simple question, but I have less experience with turbulence models. To select the best turbulence model, do I have to check for mesh independence for each turbulence model? In simple words, say if I have an independent grid when I first tested my model with K-e RNG model, do I have to test for mesh sensitivity again for a different turbulence model (say k-w or k-e SST)? Also I have another question. When we talk about low RE turbulence models, how low (approx magnitude) should the Re be, say for impingement flow via a circular nozzle (cylindrical pipe in 3D) Re_transition = 2000? And, since it is turbulence model, can i safely assume low Re turbulence model for Re <= 8000?

February 26, 2010, 05:41		#2
RossFS Member Join Date: Feb 2010 Location: Australia Posts: 65 Rep Power: 16	If the Reynolds number of the flow is not very high, you could get away with using some of the best turbulence models your CFD software has as the computational requirements should not be that much. Re =<8000 is near the bottom end of Re's from what I gather. Note that the above two comments are based on only some basic research into DNS, k-epsilon, k-omega and SST-SAS models and a little in ANSYS CFX. (4th year engineering thesis)