vof fluid sloshing dispersed fluid and continuous fluid

navi · September 2, 2016, 01:42

thank you
ill back with good result

navi · September 2, 2016, 02:19

hey glenn

i have setted adaptive time step, its taking more than 24 hours, can you tell is my adaptive time step setting is good, how long it take for 300000 elements with convergence 1*10^-4 with my adaptive time step (pc specification is 8 gb ram)

ghorrocks · September 2, 2016, 02:30

I see you have set a min and max time step quite close together. Make sure it does not hit these limits - I usually use a min time step size of 1e-10 so it can never hit it. If it hits the limits it defeats the entire point of adaptive time stepping.

You will have to work out how long the simulation is going to take. I do not know if you are running on a super computer or an ipad. And RAM is not the main factor for CFD simulation speed - it is CPU floating point performance.

navi · September 7, 2016, 01:16

hey Glenn

I have used as you say adaptive time step, as you say minimum time step as 1E-10
but its taking more than 5 days but still calculations not completed for 10 sec simulation,
is there any way to check the how long can this program run

and when i am seeing the help in Ansys about adaptive time stepping, there it shows adaptive time stepping is not used for Euler-Euler approach (VOF) as shown in attachments

ghorrocks · September 7, 2016, 01:31

The documentation you quote for adaptive time stepping is for Fluent, not CFX. It is not relevant to CFX.

Multiphase simulations are long and slow. That is why they invented super computers. I said that back in post #2

It is also why you need to do all the setup work first on a simple model to get the mesh, time step and convergence set correctly before you invest a lot of time in a serious simulation. That is why I suggested you do a simplified model back in post #7. You can also do the mesh, time step and convergence sensitivity checks on this simple case to speed things up.

navi · September 7, 2016, 01:42

i done the same with simple model, hex fine mesh, but to relate it is correct with my experiment , i have to run minimum 12 sec to check,
hope ill again scale down and do, thanks glenn

navi · October 4, 2016, 02:32

hey glenn

i have tried all method but its still there is no droplet on freesurface... useded adaptive also with all settings as you said

ghorrocks · October 4, 2016, 05:55

Out of interest - have you done a literature search on spray or jet breakup and droplet formation? It is useful to know what other people have used to successfully model this in the past.

navi · October 6, 2016, 01:48

hey glenn

i am still researching about droplet formation, i got to know even for supercomputers take more than 6 months to converge to solution of N-S equation, and can i know how this ansys solve N-S equation, or is CFX is solving reynolds averaged NS equation instead to general NS equation

and how can i get impact force on walls in my simulation

glenn you said in adaptive time stepping give minimum time step as 1E-10, and you told to make sure timestep is not hitting the limit, but every new time step is taking as 1E-10 only (previous timestep, next timestep),. my question is i guess this timestep is till not accurate to converge solution right

ghorrocks · October 9, 2016, 08:01

CFX can solve the Navier Stokes equation with or without Reynolds/Fauve Averaging.

You can get the wall forces from force_x/y/z()@wall

If adaptive time stepping is just going to time steps of 1e-10s then it shows it is having a hard time converging your simulation. You need to improve the numerical stability of your simulation. Consider whether you have appropriate physics, good enough mesh quality and double precision numerics.

navi · October 13, 2016, 01:57

hey glenn

no droplet still,''
can i know how to refine mesh, as of now my Ansys license say 512k nodes is maximum showing this in solver manager after i put for run, my model mesh node is already 5 lakh nodes and elemnts, i cant decrease element size further. so how to do this mesh to get droplet , how to refine mesh, is there any tutorial or you know how to change mesh quality , what to do now

Is the wall force XYZ @ wall is in output variable, i checked in output variable i couldn't find wall force in output variable, (where can i see this force @ wall)
thanks

ghorrocks · October 13, 2016, 02:05

Have you looked at the sort of meshes required to model droplet formation? Have a look in the literature o find out what other people have used. Have a look at how fine a mesh is required, and estimate how many droplets you are likely to have.

I think you will find you have no chance of modelling it inside 500k nodes. That would mean the simulation you are trying to do is not possible with the resources you have available.

But don't take my word for it. Look in the literature to see what is required and see if the simulation you are proposing is going to be possible with the resources you have available.

navi · October 13, 2016, 02:35

yeah glenn i have seen lot of literature about droplet formation but in every literature i see there are only equations telling in between this range of courant no. and other no. there will be droplet formation .

for my simulation, is moving mesh is better or fixed mesh is better,
now i am not using moving mesh, if i use moving mesh how can I give expression for rolling oscillation

ghorrocks · October 13, 2016, 06:48

Have you seen these tutorial videos?

https://www.youtube.com/watch?v=SwSbXwIKoto
https://www.youtube.com/watch?v=g-PTht2jnOY

This thesis describes the meshes he uses in detail:
http://scholarworks.rit.edu/cgi/view...context=theses

Moving mesh will add complexity to your simulation, something else you need to develop and get right and make the simulation run time longer. But if you need it then you have to use it.

navi · October 13, 2016, 06:57

yeah i have seen that, but i need to do in 3d tank. any how ill do and check in 2d,

is there any way to check mesh element limit in ansys software package

ghorrocks · October 13, 2016, 18:24

But if you look at the meshes they use in 2D and the results they get you can assume that for equivalent results in 3D you will need to have that mesh resolution in the third dimension as well. Hence you can estimate the mesh required.

If your license has a node limit it will be in the output file. It tells you which license it is using.

navi · October 19, 2016, 04:51

hey glenn

i am trying in 2d, it was running, mean while can i know interphase transfer in fluid pair models, there are two models free surface and mixture model,as i seen in help it is used for inhomogeneous right, or

if this helps me in visualize droplets formation, i can select mixture model >> in mixture model the interfacial length scale i shoud define (even i seen the equations) how can i calculate this interfacial lenth scale??? is there any standard value for this

thank you

Far · November 2, 2016, 04:24

Quote:

Fluent has a hanging node mesh refinement which keeps mesh quality high. I have used this in fluent and had good results. But to make it dynamically track interfaces is not easy It took me several months of work to get the mesh refinement to dynamically track the interface.

Nice to hear this

Did you use UDF to do this?

ghorrocks · November 2, 2016, 05:14

It requires a UDF, some additional user variables and a scheme file. It is quite tricky to develop (especially for multiprocessor) but once you got it working it works very nicely. I did contact ANSYS and offer it to them to incorporating it in the next release of Fluent but I never heard back from them on this. Shame, because it makes VOF calculations much faster while keeping good resolution of the interface. It could also work well for thing like shock waves and other very steep gradient simulations where local mesh refinement is required.

Far · November 2, 2016, 07:21

Quote:

Originally Posted by ghorrocks

It requires a UDF, some additional user variables and a scheme file. It is quite tricky to develop (especially for multiprocessor) but once you got it working it works very nicely. I did contact ANSYS and offer it to them to incorporating it in the next release of Fluent but I never heard back from them on this. Shame, because it makes VOF calculations much faster while keeping good resolution of the interface. It could also work well for thing like shock waves and other very steep gradient simulations where local mesh refinement is required.

this will make it easier for user and will make less revenue. Do you that UDF still with you? Is it proprietary to some firm?

September 2, 2016, 02:30		#23
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	I see you have set a min and max time step quite close together. Make sure it does not hit these limits - I usually use a min time step size of 1e-10 so it can never hit it. If it hits the limits it defeats the entire point of adaptive time stepping. You will have to work out how long the simulation is going to take. I do not know if you are running on a super computer or an ipad. And RAM is not the main factor for CFD simulation speed - it is CPU floating point performance. Far likes this.

October 4, 2016, 02:32	still no droplets	#27
navi Member naveen kumar s Join Date: Aug 2016 Location: india, bengalore Posts: 51 Rep Power: 10	hey glenn i have tried all method but its still there is no droplet on freesurface... useded adaptive also with all settings as you said

October 6, 2016, 01:48	impact forces on walls	#29
navi Member naveen kumar s Join Date: Aug 2016 Location: india, bengalore Posts: 51 Rep Power: 10	hey glenn i am still researching about droplet formation, i got to know even for supercomputers take more than 6 months to converge to solution of N-S equation, and can i know how this ansys solve N-S equation, or is CFX is solving reynolds averaged NS equation instead to general NS equation and how can i get impact force on walls in my simulation glenn you said in adaptive time stepping give minimum time step as 1E-10, and you told to make sure timestep is not hitting the limit, but every new time step is taking as 1E-10 only (previous timestep, next timestep),. my question is i guess this timestep is till not accurate to converge solution right

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Wrong flow in ratating domain problem	Sanyo	CFX	17	August 15, 2015 07:20
Overflow Error in Multiphase Modelling with Two Continuous Fluids	ashtonJ	CFX	6	August 11, 2014 15:32
Multiphase flow with two continuous fluids and one dispersed fluid	ashtonJ	CFX	5	July 25, 2014 07:31
How to choose the mean diameter value for dispersed fluid?	creddy_trddc	CFX	1	October 30, 2011 05:30
air bubble is disappear increasing time using vof	xujjun	CFX	9	June 9, 2009 08:59

September 2, 2016, 01:42		#21
navi Member naveen kumar s Join Date: Aug 2016 Location: india, bengalore Posts: 51 Rep Power: 10	thank you ill back with good result

September 7, 2016, 01:31		#25
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	The documentation you quote for adaptive time stepping is for Fluent, not CFX. It is not relevant to CFX. Multiphase simulations are long and slow. That is why they invented super computers. I said that back in post #2 It is also why you need to do all the setup work first on a simple model to get the mesh, time step and convergence set correctly before you invest a lot of time in a serious simulation. That is why I suggested you do a simplified model back in post #7. You can also do the mesh, time step and convergence sensitivity checks on this simple case to speed things up.

September 7, 2016, 01:42		#26
navi Member naveen kumar s Join Date: Aug 2016 Location: india, bengalore Posts: 51 Rep Power: 10	i done the same with simple model, hex fine mesh, but to relate it is correct with my experiment , i have to run minimum 12 sec to check, hope ill again scale down and do, thanks glenn

October 4, 2016, 05:55		#28
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	Out of interest - have you done a literature search on spray or jet breakup and droplet formation? It is useful to know what other people have used to successfully model this in the past.

October 9, 2016, 08:01		#30
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	CFX can solve the Navier Stokes equation with or without Reynolds/Fauve Averaging. You can get the wall forces from force_x/y/z()@wall If adaptive time stepping is just going to time steps of 1e-10s then it shows it is having a hard time converging your simulation. You need to improve the numerical stability of your simulation. Consider whether you have appropriate physics, good enough mesh quality and double precision numerics.

October 13, 2016, 01:57		#31
navi Member naveen kumar s Join Date: Aug 2016 Location: india, bengalore Posts: 51 Rep Power: 10	hey glenn no droplet still,'' can i know how to refine mesh, as of now my Ansys license say 512k nodes is maximum showing this in solver manager after i put for run, my model mesh node is already 5 lakh nodes and elemnts, i cant decrease element size further. so how to do this mesh to get droplet , how to refine mesh, is there any tutorial or you know how to change mesh quality , what to do now Is the wall force XYZ @ wall is in output variable, i checked in output variable i couldn't find wall force in output variable, (where can i see this force @ wall) thanks

October 13, 2016, 02:05		#32
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	Have you looked at the sort of meshes required to model droplet formation? Have a look in the literature o find out what other people have used. Have a look at how fine a mesh is required, and estimate how many droplets you are likely to have. I think you will find you have no chance of modelling it inside 500k nodes. That would mean the simulation you are trying to do is not possible with the resources you have available. But don't take my word for it. Look in the literature to see what is required and see if the simulation you are proposing is going to be possible with the resources you have available.

October 13, 2016, 02:35		#33
navi Member naveen kumar s Join Date: Aug 2016 Location: india, bengalore Posts: 51 Rep Power: 10	yeah glenn i have seen lot of literature about droplet formation but in every literature i see there are only equations telling in between this range of courant no. and other no. there will be droplet formation . for my simulation, is moving mesh is better or fixed mesh is better, now i am not using moving mesh, if i use moving mesh how can I give expression for rolling oscillation

October 13, 2016, 06:48		#34
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	Have you seen these tutorial videos? https://www.youtube.com/watch?v=SwSbXwIKoto https://www.youtube.com/watch?v=g-PTht2jnOY This thesis describes the meshes he uses in detail: http://scholarworks.rit.edu/cgi/view...context=theses Moving mesh will add complexity to your simulation, something else you need to develop and get right and make the simulation run time longer. But if you need it then you have to use it.

October 13, 2016, 06:57		#35
navi Member naveen kumar s Join Date: Aug 2016 Location: india, bengalore Posts: 51 Rep Power: 10	yeah i have seen that, but i need to do in 3d tank. any how ill do and check in 2d, is there any way to check mesh element limit in ansys software package

October 13, 2016, 18:24		#36
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	But if you look at the meshes they use in 2D and the results they get you can assume that for equivalent results in 3D you will need to have that mesh resolution in the third dimension as well. Hence you can estimate the mesh required. If your license has a node limit it will be in the output file. It tells you which license it is using.

October 19, 2016, 04:51		#37
navi Member naveen kumar s Join Date: Aug 2016 Location: india, bengalore Posts: 51 Rep Power: 10	hey glenn i am trying in 2d, it was running, mean while can i know interphase transfer in fluid pair models, there are two models free surface and mixture model,as i seen in help it is used for inhomogeneous right, or if this helps me in visualize droplets formation, i can select mixture model >> in mixture model the interfacial length scale i shoud define (even i seen the equations) how can i calculate this interfacial lenth scale??? is there any standard value for this thank you

November 2, 2016, 05:14		#39
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	It requires a UDF, some additional user variables and a scheme file. It is quite tricky to develop (especially for multiprocessor) but once you got it working it works very nicely. I did contact ANSYS and offer it to them to incorporating it in the next release of Fluent but I never heard back from them on this. Shame, because it makes VOF calculations much faster while keeping good resolution of the interface. It could also work well for thing like shock waves and other very steep gradient simulations where local mesh refinement is required.