Turbomachinary and Boundary Layers

EDE16 · August 5, 2016, 20:33

Hi all,

Im modelling a radial compressor and applying a sliding mesh in STAR. MY issue is my blade is very close to the inner housing walls. So I dont have alot of room for my rotating region which leads to the issue of having room for prism layer, I can really only get one layer and even then I get alot of skewed cells. HOw bad would it be not to have any prism layers on the inner housing walls in order to have better cell quality in that area??

Any suggestions appreciated,

Ed

DarylMusashi · August 6, 2016, 06:48

"MY issue is my blade is very close to the inner housing walls." When I understand you correctly you are talking about the tip gap of the blade, which is very small in your case?

One prism layer is definitely not enough to resolve the boundary layer of the flow field sufficiently when using a Low-Reynolds approach without wall functions (like Spalart-Allmaras, k-omega or SST). For a good boundary layer resolution you need about 5 - 10 layers by an expansion ratio of rougly 1.2. A good standard value for the height of the first cell in turbomachinery applications is 1E-05 m.
Which turbulence model are you going to use? If it uses wall functions (e.g. k-epsilon) one layer might be enough, but please don't expect accurate results.

Are you forced to use an unstructured mesh for this? For meshing of turbomachinery applications of all kind NUMECA's AutoGrid5 is the superior meshing tool in my opinion. In general you get a high quality structured mesh with very good boundary layer resolution within minutes.

EDE16 · August 6, 2016, 06:57

Yes the tip gap area. I am using kw SST, using STAR CCM unstructured with polyhedral cells. majority of the rest of the simulation model has y+ of 0-1, with very few cells (less than 100 in a 5x10^6 mesh) then above 5-10. I had thought maybe that the compressor spinning at over 150000rpm would mean no flow in that area would be in the streamwise direction through the compressor, all spinning around with the compressor wheel due to centrifugal forces

August 5, 2016, 20:33	Turbomachinary and Boundary Layers	#1
EDE16 Member Join Date: Jun 2016 Posts: 52 Rep Power: 9	Hi all, Im modelling a radial compressor and applying a sliding mesh in STAR. MY issue is my blade is very close to the inner housing walls. So I dont have alot of room for my rotating region which leads to the issue of having room for prism layer, I can really only get one layer and even then I get alot of skewed cells. HOw bad would it be not to have any prism layers on the inner housing walls in order to have better cell quality in that area?? Any suggestions appreciated, Ed

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Boundary layer in Fluent---urgent	haamed	FLUENT	0	October 25, 2009 05:51
Boundary layer in Fluent	haamed	Main CFD Forum	0	October 25, 2009 05:43
Unsteady Boundary Layer Flow	Wen Long	Main CFD Forum	0	July 29, 2002 23:08

August 6, 2016, 06:48		#2
DarylMusashi Senior Member Holger Dietrich Join Date: Apr 2011 Location: Germany Posts: 174 Rep Power: 15	"MY issue is my blade is very close to the inner housing walls." When I understand you correctly you are talking about the tip gap of the blade, which is very small in your case? One prism layer is definitely not enough to resolve the boundary layer of the flow field sufficiently when using a Low-Reynolds approach without wall functions (like Spalart-Allmaras, k-omega or SST). For a good boundary layer resolution you need about 5 - 10 layers by an expansion ratio of rougly 1.2. A good standard value for the height of the first cell in turbomachinery applications is 1E-05 m. Which turbulence model are you going to use? If it uses wall functions (e.g. k-epsilon) one layer might be enough, but please don't expect accurate results. Are you forced to use an unstructured mesh for this? For meshing of turbomachinery applications of all kind NUMECA's AutoGrid5 is the superior meshing tool in my opinion. In general you get a high quality structured mesh with very good boundary layer resolution within minutes.

August 6, 2016, 06:57		#3
EDE16 Member Join Date: Jun 2016 Posts: 52 Rep Power: 9	Yes the tip gap area. I am using kw SST, using STAR CCM unstructured with polyhedral cells. majority of the rest of the simulation model has y+ of 0-1, with very few cells (less than 100 in a 5x10^6 mesh) then above 5-10. I had thought maybe that the compressor spinning at over 150000rpm would mean no flow in that area would be in the streamwise direction through the compressor, all spinning around with the compressor wheel due to centrifugal forces