CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Main CFD Forum

How to pause a transient simulation and resume without time step advancement?

Register Blogs Members List Search Today's Posts Mark Forums Read

Like Tree2Likes

Reply
 
LinkBack Thread Tools Display Modes
Old   January 6, 2012, 08:14
Default
  #21
Senior Member
 
Join Date: Aug 2011
Posts: 251
Rep Power: 7
leflix is on a distinguished road
Quote:
Originally Posted by arjun View Post
These Ak's need mass fluxes as you pointed out and the mass fluxes need Rhie chow. Which is what we want to compute from Ak. So it is catch 22. For Ak need Fluxes AND for Fluxes need Ak.

Ok Arjun I see your point but it depends how you proceed in your calculation procedure.

Rhie and chow correction is used to improve the mass flux.
Lets call this improved mass flux: m+
while m would be the mass flux without the rihie and chow correction.

m+ = m + (rhie and chow correction)

m+ could be used
(1) only when you compute the AK,
(2) only when you compute the right hand-side of P' equation,
(3) in the both cases

If you use m+ to compute AK I agree with you there is a catch 22 as you said!

but if you use m to compute AK
and m+ to compute the right hand side of P' equation you do not have any problem anymore.

It has not been demonstrated yet as far as I know, that using m+ to compute AK was superior to only use m+ to compute the right hand-side of P' equation.

what do you think about this ?
leflix is offline   Reply With Quote

Old   January 6, 2012, 09:40
Default
  #22
Senior Member
 
Arjun
Join Date: Mar 2009
Location: Nurenberg, Germany
Posts: 372
Rep Power: 10
arjun is on a distinguished road
Quote:
Originally Posted by leflix View Post

but if you use m to compute AK
and m+ to compute the right hand side of P' equation you do not have any problem anymore.
If you use m to compute Ak and m+ to compute pressure correction equation, this could only be NO PROBLEM if it means that the mass fluxes used in momentum equation is m and the mass fluxes used in pressure euqation is m+.

Now again there are two things:

1. m+ is used in momentum equation, which is the purpose of rhie chow treatment. That is to avoid checker board oscillations. So the mass flux in momentum equation and continuity has to be same. For this reason, to my knowledge all the solvers use m+ in momentum equation. That is AK is calculated from m+ and not from m.

2. Just for fun last month I did try using m and not m+ in momentum equation and used m+ in pressure equation. This method is not stable. It does not converge in more than 95% of cases. It converged for only 1 case out of few i tried. So using m in momentum equation does not work for collocated arrangement.
arjun is offline   Reply With Quote

Old   January 6, 2012, 14:19
Default
  #23
Senior Member
 
Join Date: Aug 2011
Posts: 251
Rep Power: 7
leflix is on a distinguished road
Quote:
Originally Posted by arjun View Post
If you use m to compute Ak and m+ to compute pressure correction equation, this could only be NO PROBLEM if it means that the mass fluxes used in momentum equation is m and the mass fluxes used in pressure euqation is m+.
Yes it is what I mean !


Quote:

1. m+ is used in momentum equation, which is the purpose of rhie chow treatment. That is to avoid checker board oscillations. So the mass flux in momentum equation and continuity has to be same. For this reason, to my knowledge all the solvers use m+ in momentum equation. That is AK is calculated from m+ and not from m.
There are some papers of Date A.W. where it is said that the correction could be applied in mass flux in momentum equation. But the superiority of this approach compared with the one consisting in applying the Rhie and Chow correction in the mass flux of the right hand side of pressure correction equation has never been demonstrated. The both can be done and work fine.
If you read the bible: Computational Methods for Fluid dynamics from Peric you will see that Rhie and chow correction (m+) is applied in the mass flux of the right hand side of pressure correction equation.


Quote:

2. Just for fun last month I did try using m and not m+ in momentum equation and used m+ in pressure equation. This method is not stable. It does not converge in more than 95% of cases. It converged for only 1 case out of few i tried. So using m in momentum equation does not work for collocated arrangement.
So there is a bugg in your program or implementation .
Because in my code, I use m+ in pressure equation only and it works fine for every cases.
In the codes of Peric (comet and others ) m+ is used in pressure equation.
and I could risk my hand that it is the same in Star-CD
leflix is offline   Reply With Quote

Old   January 6, 2012, 18:53
Default
  #24
Senior Member
 
Arjun
Join Date: Mar 2009
Location: Nurenberg, Germany
Posts: 372
Rep Power: 10
arjun is on a distinguished road
Quote:
Originally Posted by leflix View Post
Yes it is what I mean !


If you read the bible: Computational Methods for Fluid dynamics from Peric you will see that Rhie and chow correction (m+) is applied in the mass flux of the right hand side of pressure correction equation.


So there is a bugg in your program or implementation .
Because in my code, I use m+ in pressure equation only and it works fine for every cases.
In the codes of Peric (comet and others ) m+ is used in pressure equation.
and I could risk my hand that it is the same in Star-CD

Interesting you say so, but Mr. Peric himself disagree with you. According to Mr. Peric in starccm+ the m+ is used in momentum equation and anyother place where mass flux is required. (This is based on my personal discussion with him).


Further I do not think there is a bug in my program because it is as stable as any commercial solver out there. I have run it with very very difficult or bad meshes still able to converge with using m+ in momentum.

Could you point me to any paper or study where somebody has used m only in momentum equation for collocated grid. I tried this approach because I felt it would be more stable but it turned out to be opposite of what I thought.
arjun is offline   Reply With Quote

Old   January 7, 2012, 14:46
Default
  #25
Senior Member
 
Join Date: Aug 2011
Posts: 251
Rep Power: 7
leflix is on a distinguished road
Quote:
Originally Posted by arjun View Post
Interesting you say so, but Mr. Peric himself disagree with you. According to Mr. Peric in starccm+ the m+ is used in momentum equation and anyother place where mass flux is required. (This is based on my personal discussion with him).

I just follow what is written in his book, and in his codes too...
But everybody has right to change his mind... ;-)

Quote:
Further I do not think there is a bug in my program because it is as stable as any commercial solver out there. I have run it with very very difficult or bad meshes still able to converge with using m+ in momentum.
No I'm sure your code is fine.. and I'm sure that using m+ in momentum is right and ok. In the original paper of Rhie and Chow it is done that way. So no problem with it.
I just mean that it's not the only alternative..
In my code I use m+ only in right hand side of pressure correction equation and it works fine.
But it is also possible that if I add m+ in momentum it works too and perhaps better...I'm just surprised that when you implemented m+ in pressure correction like me it didn't work.

Quote:
Could you point me to any paper or study where somebody has used m only in momentum equation for collocated grid.
You mean in momentum or in pressure correction ?
I will check the papers of Date and send you the right references


Quote:
I tried this approach because I felt it would be more stable but it turned out to be opposite of what I thought.
No it's not what I said ! I think your way is ok and perfectly stable! I just mean that the other alternative works fine too.
Now I have in mind a paper where these different ways of implementing the Rhie and Chow correction have been analysed in terms of accuracy and convergence performance... As far as I remember the conclusion was that the three ways (m+ in momentum only, m+ in pressure correction only, m+ in the both) work more or less the same way.
I have to check it and I will come back to you.
leflix is offline   Reply With Quote

Old   January 7, 2012, 20:55
Default
  #26
Senior Member
 
Arjun
Join Date: Mar 2009
Location: Nurenberg, Germany
Posts: 372
Rep Power: 10
arjun is on a distinguished road
Quote:
Originally Posted by leflix View Post
I just follow what is written in his book, and in his codes too...
But everybody has right to change his mind... ;-)
I am not sure about he changed his mind or not. The point he pointed out that where-ever fluxes are needed it contained Rhie Chow terms to so that it follows the continuity tightly. This was the approach taken in starccm+.

But since starccm+ is made by many people and it is not he alone that decides.

Quote:
Originally Posted by leflix View Post
No I'm sure your code is fine.. and I'm sure that using m+ in momentum is right and ok. In the original paper of Rhie and Chow it is done that way. So no problem with it.
I just mean that it's not the only alternative..
In my code I use m+ only in right hand side of pressure correction equation and it works fine.
But it is also possible that if I add m+ in momentum it works too and perhaps better...I'm just surprised that when you implemented m+ in pressure correction like me it didn't work.

You mean in momentum or in pressure correction ?
I will check the papers of Date and send you the right references


No it's not what I said ! I think your way is ok and perfectly stable! I just mean that the other alternative works fine too.
Now I have in mind a paper where these different ways of implementing the Rhie and Chow correction have been analysed in terms of accuracy and convergence performance... As far as I remember the conclusion was that the three ways (m+ in momentum only, m+ in pressure correction only, m+ in the both) work more or less the same way.
I have to check it and I will come back to you.

Thank you, that paper might point out what I did that was wrong or may be what mistake I made. It will be good read, ideas are always welcome.
arjun is offline   Reply With Quote

Old   January 11, 2012, 11:34
Default
  #27
Senior Member
 
Join Date: Aug 2011
Posts: 251
Rep Power: 7
leflix is on a distinguished road
Quote:
Originally Posted by arjun View Post

Thank you, that paper might point out what I did that was wrong or may be what mistake I made. It will be good read, ideas are always welcome.
Hi Arjun,

I'm back.. here are the references of the papers I was speaking about.

A.W. Date, "Fluid dynamical view of pressure chechboarding problem and smoothing pressure correction on meshes with colocated variables", International journal of heat and mass transfer 46, (2003) 4885-4898

A.W. Date, "Complete pressure correction algorithm for solution of incompressible Navier-Stokes equations on a nonstaggered grid", Numerical Heat Transfer, Part B, 29, (1996), 441-458

B. Yu, W-Q Tao, J-J Wei, Y. Kawaguchi, T. Tagawa, H. Ozoe, "Discussion on momentum interpolation method for collocated grids of incompressible flow", Numerical Heat Transfer, Part B, 42, (2002), 141-166


Papers of Date implement Rhie and chow interpolation in the mass conservation i.e in the right hand side of pressure correction equation as I usually do.


In paper 3, the authors examined 3 different practices of implementing Rhie and Chow interpolation.
A refers to what you do i.e. interpolation in the mass flux of momentum equation,
B refers to interpolation in the right hand side of pressure correction equation,
C is the same as B including divergence of velocity in the AP coefficient while in A and B it is not.

Happy reading..

I believe this discussion should be switched in Rhie and Chow specific discussion in this forum
leflix is offline   Reply With Quote

Old   January 11, 2012, 14:56
Default
  #28
Senior Member
 
Arjun
Join Date: Mar 2009
Location: Nurenberg, Germany
Posts: 372
Rep Power: 10
arjun is on a distinguished road
I have pdf form of book from Anil Date so I will look into it as I get time.

I downloaded this paper though.

Quote:
Originally Posted by leflix View Post
B. Yu, W-Q Tao, J-J Wei, Y. Kawaguchi, T. Tagawa, H. Ozoe, "Discussion on momentum interpolation method for collocated grids of incompressible flow", Numerical Heat Transfer, Part B, 42, (2002), 141-166
At quick glance on this paper says that it also uses rhie chow based velocity momentum equation. Look at equation (6) , there are two things 1. phi and 2. u. The phi_e is calculated as QUICK or any other scheme but u_e is calculated as some scheme he developed BUT based on Rhie Chow terms. In fact most of this paper is how to calculate this u_e using Rhie Chow type scheme.

I will look into Anil Date explanation too when I get time.


-----------

Edited to add:


I looked into Date's book. At least in the book , the word Rhie appears only once and that is in bibliography. The procedure that he explains in the book is some what very similar to what CFX4 used to do and later on CFX5 also does. Which is different than Fluent starCCM approach.
(I read CFX4 manual in year 2000 so I do not remember it very well now but I am sure they both are same approach).

This method is stable as demostrated by various CFX versions, but I am not sure it is much relevant to discussion at hand because this is not the most popular way of doing things.

PS: I do have some issues how Date explained in his book but that is not much use discussing.

Last edited by arjun; January 11, 2012 at 16:43.
arjun is offline   Reply With Quote

Old   January 12, 2012, 08:30
Default
  #29
Senior Member
 
Arjun
Join Date: Mar 2009
Location: Nurenberg, Germany
Posts: 372
Rep Power: 10
arjun is on a distinguished road
Quote:
Originally Posted by leflix View Post

B. Yu, W-Q Tao, J-J Wei, Y. Kawaguchi, T. Tagawa, H. Ozoe, "Discussion on momentum interpolation method for collocated grids of incompressible flow", Numerical Heat Transfer, Part B, 42, (2002), 141-166


I read this paper carefully in the evening and it seems you are correct, in approach B and C they do interpolate velocity with out rhie chow terms for momentum.

I would try this approach again when i get time.
arjun is offline   Reply With Quote

Old   June 20, 2014, 01:43
Question Fluent Resume/Restart. How to do it?
  #30
New Member
 
Sarthak Nag
Join Date: Jun 2014
Location: Himachal Pradesh
Posts: 2
Rep Power: 0
iamsarthaknag is on a distinguished road
If only the initial case file and all the data files are saved and suddenly your system shuts down, can you restart or resume the program from the time step you have left, means from the time step the computer shut down.?? My computer shut down after about 80%time steps and now I have to do the simulation again.
iamsarthaknag is offline   Reply With Quote

Old   June 20, 2014, 02:36
Default
  #31
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 918
Rep Power: 15
ghost82 will become famous soon enough
Open the cas, then open the last dat file and your simulation will resume from the latest time.

Daniele
ghost82 is offline   Reply With Quote

Old   June 20, 2014, 07:50
Default
  #32
New Member
 
Sarthak Nag
Join Date: Jun 2014
Location: Himachal Pradesh
Posts: 2
Rep Power: 0
iamsarthaknag is on a distinguished road
Quote:
Originally Posted by ghost82 View Post
Open the cas, then open the last dat file and your simulation will resume from the latest time.

Daniele
Thank you... It worked.. But will the history files of drag and lift will resume recording data too.??
iamsarthaknag is offline   Reply With Quote

Old   June 27, 2014, 18:40
Default
  #33
New Member
 
Manuel
Join Date: Jan 2014
Posts: 18
Rep Power: 3
86lolo is on a distinguished road
I got lost and I have to say I didn't read through all the posts...
But to answer the 1st question which gives title to this thread: It is possible (in FLUENT, at least), to resume calculations in the same time-step where you stopped.
TUI command:
solve/iterate [number_of_iterations]

This, in steady state, just iterates. In transient, it adds iterations to the time step and will never step into the following time-step (it will stop when the number of iterations is reached, or upon convergence criteria).
This will NOT resume the transient calculations, this just iterates in the time step you are. Therefore, if you want to resume your calculations completely, it is a 2-step task:
solve/iterate (to finish your last time step where you paused)
solve/dti (to continue your unsteady calculation)

Hope it helps
86lolo is offline   Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On



All times are GMT -4. The time now is 07:05.