Les simulation of a nozzle

AS_Aero · February 6, 2017, 03:16

Hi All

I want to do an internal flow simulation of a nozzle. I am doing LES directly for understanding the unsteadiness. I would like to ask you all, if I should do a RANS first before going to LES and can I give the inlet boundary condition from scratch than using the data from the RANS ? Like I know the mass flow rate or the operating pressure value as per the experiment, so can I use them directly for my LES simulation ? Kindly help me solve this. Thanks in advance.

ghorrocks · February 6, 2017, 04:30

Are you doing pure LES or SAS or DES?

What are you intending to transfer over from the RANS simulation to the LES one? What are you going to do about turbulence as the flow passes through the inlet boundary?

AS_Aero · February 6, 2017, 04:48

Hi ghorrocks

I havnt decided what parameters to transfer from RANS, but as per my knowledge we have to start from RANS to continue them to LES right ? Or can I start an LES simulation from the scratch ? I would try SAS as well as a second option, but first would be pure LES with Smagorinsky or Wale model. At present I am generating my mesh using ansys mesh.

ghorrocks · February 6, 2017, 04:54

The choice depends on what you are modelling. Can you describe more details about what you are modelling?

AS_Aero · February 6, 2017, 05:01

Dear Glenn

Thanks for your abrupt reply.
Well I am doing a simple pipe flow with a nozzle at its end. I just want to simulate the internal flow especially the internal flow in the nozzle. The nozzle geometry is not axisymmetric and has some curves, so to study the flow and its unsteadiness, I planned to do an LES than understanding it with a simple RANS. I have the inlet mass flow rate or the operating pressure condition from experiments and the nozzle exit is given atmospheric pressure 1bar ( Pressure Outlet BC'n ). This is much about the model.

ghorrocks · February 6, 2017, 05:24

You have a fundamental problem in modelling this with LES. You have to describe inlet boundary conditions which include the inlet turbulence with LES scale turbulent eddies in it. This is a very difficult thing to do and there is not a model for it in CFD.

Some options:
* Extend the domain downstream so you can use simple inlet boundaries and by the time the flow gets to your region of interest the turbulence is sufficiently developed. (This is very inefficient as you need to extend the domain size a lot, and that makes the simulation even bigger)
* You can use the SAS or DES models so that you can have RANS flows entering the domain which transition to LES inside the domain. If these models are suitable then they can be very useful.
* You can do a simple 1D pipe flow simulation with a periodic pair as inlets and outlets and do LES on that. Run it until the flow gets to the turbulence level you want. Then "record" the turbulence on this model and use it as a inlet condition for your model. This is a reasonably efficient way of doing it as well, but you will need to develop the LES "recorder" (which will probably require fortran).

AS_Aero · February 6, 2017, 05:55

Dear Glenn

Thanks for your suggestions. But as I said the pipe itself is to make the simulation simpler. I have given an entrance length to the nozzle based on its diameter so that I give the mass flow rate or pressure and the inlet condition at the inlet of the pipe and this method is exactly what you mentioned on the 1st option. With respect to 2nd option SAS can I do it from scratch or does SAS need some RANS input ? In this case my RANS should be atleast accurate right ? And what input should I feed for SAS ? And option 3 sounds bit hard for my experience !!

ghorrocks · February 6, 2017, 17:58

Your question is about how to implement the gross boundary conditions: inlet mass flow rate or pressure?

You should implement which ever better describe the conditions at the boundary interfaces. So the answer depends on the system you are modelling.

The mass flow inlet will be more numerically stable so it run a little better than pressure, but both can work.

My point is that a "detail" about the inlet boundary is often a show-stopper for LES: How do you implement a flow field at the inlet which contains the correct turbulence, given that for LES the larger turbulence scales are resolved in the velocity field? Unless you have a method of doing this you cannot do LES. You will also find a velocity specified inlet much easier to implement something here than a pressure boundary.

Whether RANS is accurate in your case depends on what you are modelling.

As for SAS and DES - these are advanced turbulence models which require a good understanding of the underlying physics. You should read the documentation on these models if you choose to use them.

AS_Aero · February 7, 2017, 03:57

Dear Glenn

Thanks a lot. So as you mentioned, inorder to have the turbulence features I have a lengthy pipe before the nozzle which replicates with the entrance length so that the flow will be fully developed by the time it reaches the nozzle inlet. So I have the inlet data (Mass flow rate of 15m/s as well as Pressure (operating gauge pressure 7 bar ) ) So my question or doubt was which one to use there, so as you mentioned , I must try with both and see. But before all this, my doubt is whether can I start LES from scratch or do I need to input the RANS result to start the LES ?
As of now I havent planned to do SAS or DES , but for sure I will consider that as well.

ghorrocks · February 7, 2017, 04:34

If you have the entrance length set up and it is working then go with that.

If you have both the mass flow rate and pressure data available then choose one and check that the model gives the other as an output. The model will be more stable with mass flow rate inlet boundaries rather than pressure so I would vote for mass flow rate boundary.

Have a look at SAS and DES, but they were really designed for vortex shedding off bluff structures and are not so suitable for pipe flows. If there are sharp bends in your pipe it might work on those but if it is just turbulence from the Re number of the flow in a pipe then I do not think it is applicable.

AS_Aero · February 7, 2017, 04:42

Cool Thanks a lot for your advice. Well my nozzle has some spiral curves and the pipe itself doesnt have any bend or anything. its just a straight section which connects to the nozzle inlet and then there is some bends. Also I have a doubt, can I use the RANS mesh itself for the LES to see if it works or not ?

ghorrocks · February 7, 2017, 05:38

The mesh size is a critical part of the LES model - it is used as part of the turbulence length scale filtering. It is highly unlikely a mesh which is acceptable for RANS is acceptable for LES.

If you don't know what turbulence length scale filtering is then you better do some reading about LES. LES is a MUCH more complex simulation than RANS, it is not just a matter of switching the turbulence model over.

AS_Aero · February 7, 2017, 05:45

Sure I will have a look !! Thanks a lot for your valuable advice !!

February 6, 2017, 03:16	Les simulation of a nozzle	#1
AS_Aero Senior Member Join Date: Mar 2013 Location: Germany Posts: 357 Rep Power: 14	Hi All I want to do an internal flow simulation of a nozzle. I am doing LES directly for understanding the unsteadiness. I would like to ask you all, if I should do a RANS first before going to LES and can I give the inlet boundary condition from scratch than using the data from the RANS ? Like I know the mass flow rate or the operating pressure value as per the experiment, so can I use them directly for my LES simulation ? Kindly help me solve this. Thanks in advance.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Simulation of a Multiphase flow in a Convergent-Divergent Nozzle	RR16988	FLUENT	1	June 9, 2015 06:10
Simulation of a simple De Laval Nozzle	RR16988	FLUENT	1	June 8, 2015 14:23
Large Eddy Simulation - LES	stix	Main CFD Forum	5	April 8, 2013 08:23
outlet profiles of LES simulation as transient inlet b.c. for LES of a bend pipe	Henny	FLUENT	0	March 28, 2013 04:02
[OpenFOAM 1.6] How to run a simulation based on either DES or LES?	imranazizahmed	OpenFOAM Running, Solving & CFD	1	January 29, 2013 01:27

February 6, 2017, 04:30		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,816 Rep Power: 144	Are you doing pure LES or SAS or DES? What are you intending to transfer over from the RANS simulation to the LES one? What are you going to do about turbulence as the flow passes through the inlet boundary?

February 6, 2017, 04:48		#3
AS_Aero Senior Member Join Date: Mar 2013 Location: Germany Posts: 357 Rep Power: 14	Hi ghorrocks I havnt decided what parameters to transfer from RANS, but as per my knowledge we have to start from RANS to continue them to LES right ? Or can I start an LES simulation from the scratch ? I would try SAS as well as a second option, but first would be pure LES with Smagorinsky or Wale model. At present I am generating my mesh using ansys mesh.

February 6, 2017, 04:54		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,816 Rep Power: 144	The choice depends on what you are modelling. Can you describe more details about what you are modelling?

February 6, 2017, 05:01		#5
AS_Aero Senior Member Join Date: Mar 2013 Location: Germany Posts: 357 Rep Power: 14	Dear Glenn Thanks for your abrupt reply. Well I am doing a simple pipe flow with a nozzle at its end. I just want to simulate the internal flow especially the internal flow in the nozzle. The nozzle geometry is not axisymmetric and has some curves, so to study the flow and its unsteadiness, I planned to do an LES than understanding it with a simple RANS. I have the inlet mass flow rate or the operating pressure condition from experiments and the nozzle exit is given atmospheric pressure 1bar ( Pressure Outlet BC'n ). This is much about the model.

February 6, 2017, 05:24		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,816 Rep Power: 144	You have a fundamental problem in modelling this with LES. You have to describe inlet boundary conditions which include the inlet turbulence with LES scale turbulent eddies in it. This is a very difficult thing to do and there is not a model for it in CFD. Some options: * Extend the domain downstream so you can use simple inlet boundaries and by the time the flow gets to your region of interest the turbulence is sufficiently developed. (This is very inefficient as you need to extend the domain size a lot, and that makes the simulation even bigger) * You can use the SAS or DES models so that you can have RANS flows entering the domain which transition to LES inside the domain. If these models are suitable then they can be very useful. * You can do a simple 1D pipe flow simulation with a periodic pair as inlets and outlets and do LES on that. Run it until the flow gets to the turbulence level you want. Then "record" the turbulence on this model and use it as a inlet condition for your model. This is a reasonably efficient way of doing it as well, but you will need to develop the LES "recorder" (which will probably require fortran).

February 6, 2017, 05:55		#7
AS_Aero Senior Member Join Date: Mar 2013 Location: Germany Posts: 357 Rep Power: 14	Dear Glenn Thanks for your suggestions. But as I said the pipe itself is to make the simulation simpler. I have given an entrance length to the nozzle based on its diameter so that I give the mass flow rate or pressure and the inlet condition at the inlet of the pipe and this method is exactly what you mentioned on the 1st option. With respect to 2nd option SAS can I do it from scratch or does SAS need some RANS input ? In this case my RANS should be atleast accurate right ? And what input should I feed for SAS ? And option 3 sounds bit hard for my experience !!

February 6, 2017, 17:58		#8
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,816 Rep Power: 144	Your question is about how to implement the gross boundary conditions: inlet mass flow rate or pressure? You should implement which ever better describe the conditions at the boundary interfaces. So the answer depends on the system you are modelling. The mass flow inlet will be more numerically stable so it run a little better than pressure, but both can work. My point is that a "detail" about the inlet boundary is often a show-stopper for LES: How do you implement a flow field at the inlet which contains the correct turbulence, given that for LES the larger turbulence scales are resolved in the velocity field? Unless you have a method of doing this you cannot do LES. You will also find a velocity specified inlet much easier to implement something here than a pressure boundary. Whether RANS is accurate in your case depends on what you are modelling. As for SAS and DES - these are advanced turbulence models which require a good understanding of the underlying physics. You should read the documentation on these models if you choose to use them.

February 7, 2017, 03:57		#9
AS_Aero Senior Member Join Date: Mar 2013 Location: Germany Posts: 357 Rep Power: 14	Dear Glenn Thanks a lot. So as you mentioned, inorder to have the turbulence features I have a lengthy pipe before the nozzle which replicates with the entrance length so that the flow will be fully developed by the time it reaches the nozzle inlet. So I have the inlet data (Mass flow rate of 15m/s as well as Pressure (operating gauge pressure 7 bar ) ) So my question or doubt was which one to use there, so as you mentioned , I must try with both and see. But before all this, my doubt is whether can I start LES from scratch or do I need to input the RANS result to start the LES ? As of now I havent planned to do SAS or DES , but for sure I will consider that as well.

February 7, 2017, 04:34		#10
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,816 Rep Power: 144	If you have the entrance length set up and it is working then go with that. If you have both the mass flow rate and pressure data available then choose one and check that the model gives the other as an output. The model will be more stable with mass flow rate inlet boundaries rather than pressure so I would vote for mass flow rate boundary. Have a look at SAS and DES, but they were really designed for vortex shedding off bluff structures and are not so suitable for pipe flows. If there are sharp bends in your pipe it might work on those but if it is just turbulence from the Re number of the flow in a pipe then I do not think it is applicable.

February 7, 2017, 04:42		#11
AS_Aero Senior Member Join Date: Mar 2013 Location: Germany Posts: 357 Rep Power: 14	Cool Thanks a lot for your advice. Well my nozzle has some spiral curves and the pipe itself doesnt have any bend or anything. its just a straight section which connects to the nozzle inlet and then there is some bends. Also I have a doubt, can I use the RANS mesh itself for the LES to see if it works or not ?

February 7, 2017, 05:38		#12
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,816 Rep Power: 144	The mesh size is a critical part of the LES model - it is used as part of the turbulence length scale filtering. It is highly unlikely a mesh which is acceptable for RANS is acceptable for LES. If you don't know what turbulence length scale filtering is then you better do some reading about LES. LES is a MUCH more complex simulation than RANS, it is not just a matter of switching the turbulence model over.

February 7, 2017, 05:45		#13
AS_Aero Senior Member Join Date: Mar 2013 Location: Germany Posts: 357 Rep Power: 14	Sure I will have a look !! Thanks a lot for your valuable advice !!