what is the proper intlet nuTilda for Spallart-Allmaras IDDES?

KTG · April 5, 2021, 18:09

Hello Foamies,

for just SA, I have read that it should be zero, but if that causes solver issues to use 3nu. I have also seen a few papers that use zero for SA IDDES, but don't really justify why. When I use nuTilda=0 at the inlet for my simulations, I get results that look good, but nuTilda = nut = 0 for the whole domain, all the time (shouldn't turbulent flow at the solid boundaries produce nuTilda and nut values that are not 0?). Also the DESModelRegions function object is constant through time. If I use nuTilda > 0 at the inlet, the DES regions change with time, but the flow looks less turbulent as nuTilda increases. If I choose values that look reasonable from boundaryFoam without using an artificial turbulence inlet condition, things look downright viscous.

Can anyone shine any light on any of these issues with this turbulence model? Thanks!

chegdan · April 6, 2021, 15:30

Dear KTG,

Just a few thoughts...

Are you assuming that your turbulence is well-developed when it enters your domain? If so, how? For things like a backward facing step, you perform channel flow simulations and extract a channel exit profile as input into your inlet profile for you backward facing step simulation. For atmospheric boundary layer problems, there are ABL profiles that you can fit to industry accepted values. You can also do a sensitivity study on nutlida to see if it actually dissipates before it reaches an area of interest, if so then zero might be your best inlet condition. Finally, you may have laminar flow entering your domain and in that case, nutilda is zero. The answer is that there is no fixed answer, but there are heuristic approaches that depend on the problem and what you are seeking to answer.

KTG · April 6, 2021, 16:29

Well, this thread is going to look like a personal blog until I have some kind of breakthrough. Here is a plot of the forces on a cube. y2 is with nuTilda = 0, and y1 is nuTilda = 3nu. You can see that the shedding frequency is pretty different.

KTG · April 6, 2021, 16:33

Sorry, looks like you posted that while I was writing that last entry! In these simulations I was just applying a laminar log layer inlet. What is strange to me though is that the DES regions are static throughout the simulation when nuTilda=0 at the inlet - even though turbulence is produced around the solid wall boundaries. Also, nut and nuTilda are zero throughout. How can the SA part of the model be functioning properly like this?

Thanks for the response-

April 5, 2021, 18:09	what is the proper intlet nuTilda for Spallart-Allmaras IDDES?	#1
KTG Senior Member Abe Join Date: May 2016 Posts: 119 Rep Power: 9	Hello Foamies, for just SA, I have read that it should be zero, but if that causes solver issues to use 3nu. I have also seen a few papers that use zero for SA IDDES, but don't really justify why. When I use nuTilda=0 at the inlet for my simulations, I get results that look good, but nuTilda = nut = 0 for the whole domain, all the time (shouldn't turbulent flow at the solid boundaries produce nuTilda and nut values that are not 0?). Also the DESModelRegions function object is constant through time. If I use nuTilda > 0 at the inlet, the DES regions change with time, but the flow looks less turbulent as nuTilda increases. If I choose values that look reasonable from boundaryFoam without using an artificial turbulence inlet condition, things look downright viscous. Can anyone shine any light on any of these issues with this turbulence model? Thanks!

April 6, 2021, 15:30		#2
chegdan Senior Member Daniel P. Combest Join Date: Mar 2009 Location: St. Louis, USA Posts: 621 Rep Power: 0	Dear KTG, Just a few thoughts... Are you assuming that your turbulence is well-developed when it enters your domain? If so, how? For things like a backward facing step, you perform channel flow simulations and extract a channel exit profile as input into your inlet profile for you backward facing step simulation. For atmospheric boundary layer problems, there are ABL profiles that you can fit to industry accepted values. You can also do a sensitivity study on nutlida to see if it actually dissipates before it reaches an area of interest, if so then zero might be your best inlet condition. Finally, you may have laminar flow entering your domain and in that case, nutilda is zero. The answer is that there is no fixed answer, but there are heuristic approaches that depend on the problem and what you are seeking to answer.

April 6, 2021, 16:33		#4
KTG Senior Member Abe Join Date: May 2016 Posts: 119 Rep Power: 9	Sorry, looks like you posted that while I was writing that last entry! In these simulations I was just applying a laminar log layer inlet. What is strange to me though is that the DES regions are static throughout the simulation when nuTilda=0 at the inlet - even though turbulence is produced around the solid wall boundaries. Also, nut and nuTilda are zero throughout. How can the SA part of the model be functioning properly like this? Thanks for the response-