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

USER_FORTRAN solve iteratively a linear equations system

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

Reply
 
LinkBack Thread Tools Display Modes
Old   October 8, 2010, 10:42
Default USER_FORTRAN solve iteratively a linear equations system
  #1
New Member
 
Filippo Pellacani
Join Date: Jul 2010
Location: München
Posts: 13
Rep Power: 7
fipellac is on a distinguished road
Hi,

I implemented already an Additional Variable under the form of a Transport equation. I didn´t get good results and I would like to have full control on the implemented transport equation.
The problem is essentially the presence of the diffusive term in the equation provided by Ansys CFX (I don´t need it) and the presence of the volume fraction in the formulation of each term.

For this reason I decided to implement a new variable linked to User_CEL. Inside the routine then, linearize the system and solve it iteratively.

I would like to receive some information regarding the following aspect:

1) Once obtained a linear equations system I want to solve in an iterative way.
Did someone tried this before in a USER_CEL? Do you have an example of it?

Thank you in advance,

Filippo
fipellac is offline   Reply With Quote

Old   October 9, 2010, 05:48
Default
  #2
Super Moderator
 
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 10,960
Rep Power: 85
ghorrocks has a spectacular aura aboutghorrocks has a spectacular aura aboutghorrocks has a spectacular aura about
If you don't want the diffusive term then can't you just set it to zero? And what are you trying to do anyway?

Before you dive into fortran are you sure you cannot do what you want using additional variables?
ghorrocks is offline   Reply With Quote

Old   October 11, 2010, 04:59
Default
  #3
New Member
 
Filippo Pellacani
Join Date: Jul 2010
Location: München
Posts: 13
Rep Power: 7
fipellac is on a distinguished road
Dear ghorrocks,

"If you don't want the diffusive term then can't you just set it to zero?"

Setting the diffusivity =0 doesn´t help so much. I also set the value to the Schmidt number to a very high value, so that the mu/sc term will be very low.

Even performing this operations the div of such a group will never be zero, but can also have very high value and the software can in some cases become very unstable.

"And what are you trying to do anyway?"

I worked during the last months to implement a one group interfacial area transport equation for the calculation of the interfacial area density. In a gas dispersed liquid continuous simulation. (bubble column D=200mm H= ca. 8 m, experimental data from Research center Dresden Germany).
The MUSIG approach even producing goos results needs really a lot of resources.
Other researcher implemented the same models in their own codes and the results seems to be very good in isothermal and non isothermal condition.

"Before you dive into fortran are you sure you cannot do what you want using additional variables?"

I tried it but this is not producing the expected results. If you want I can send you a document where I explain the implementation of the Additional variable. I documented it already.

Bye,

Filippo
fipellac is offline   Reply With Quote

Old   October 11, 2010, 05:22
Default
  #4
Super Moderator
 
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 10,960
Rep Power: 85
ghorrocks has a spectacular aura aboutghorrocks has a spectacular aura aboutghorrocks has a spectacular aura about
It has been a while since I used additional variables but I seem to remember that one option is "transport equation" where the convective terms are included but the diffusive ones are not. I cannot see how any user fortran is going to have less diffusion than that.

I am also no expert on MUSIG but again it seems like the models you are trying to develop are already in CFX. I guess I have missed something, but I would suggest looking at the full range of options already in CFX before developing new approaches. Fro instance have you looked at the DPMOM approaches? They may be more appropriate for you.

Is this paper something like what you are trying to do? I can only see the abstract but it describes a bit of what they are doing.
http://scitation.aip.org/getabs/serv...ifs=yes&ref=no
ghorrocks is offline   Reply With Quote

Old   October 11, 2010, 05:45
Default
  #5
New Member
 
Filippo Pellacani
Join Date: Jul 2010
Location: München
Posts: 13
Rep Power: 7
fipellac is on a distinguished road
One of the goal of my PhD is the implementation of the Interfacial area transport equation in Ansys CFX. (If interested, You can read something more in the book Thermo-fluid Dynamics of Two-Phase Flow, Hibiki - iShii, Chapter 11).

In the past year I was learning how to use the monodispersed approach and the Musig approach. I already tested them, but they are "STATE OF THE ART". I need to implement something new, in this case the IATE of developed by Hibiki Ishii, with several source terms and cmpare them.

I already tried the additional variable approach and was not bringing the expected results. In user fortran I could implent the equation in the form I need/want.

Thank you for the link, but sorry i can´t open it. I will ask to buy it it seems interesting.
fipellac is offline   Reply With Quote

Old   October 12, 2010, 22:45
Default
  #6
Senior Member
 
Michael P. Owen
Join Date: Mar 2009
Posts: 195
Rep Power: 8
michael_owen is on a distinguished road
You keep claiming that the normal AV transport equation did not give satisfactory results. In what way?
michael_owen is offline   Reply With Quote

Old   May 20, 2011, 05:42
Default
  #7
New Member
 
Filippo Pellacani
Join Date: Jul 2010
Location: München
Posts: 13
Rep Power: 7
fipellac is on a distinguished road
ANSYS CFX calculates the gradients using the central difference scheme. this leads to numerical instabilities and the results of simulation are full of oscillations.

i developed my own set of subroutines for mapping of the points, calculation of gradient and divergencies. and this is lowering a lot the numerical oscillations. the problem now is that the simulation time increased really a lot.

the other thing is that the procedure is working in series and not yet in parallel and i´m wongin just with 2D approximation (vertical pipe 1degree slice).
fipellac is offline   Reply With Quote

Old   May 20, 2011, 10:36
Default
  #8
Senior Member
 
Join Date: Apr 2009
Posts: 517
Rep Power: 12
stumpy is on a distinguished road
Did you ask support if there's an expert parameter to change the scheme used to calculate the gradients?
stumpy is offline   Reply With Quote

Old   May 20, 2011, 10:37
Default
  #9
New Member
 
Filippo Pellacani
Join Date: Jul 2010
Location: München
Posts: 13
Rep Power: 7
fipellac is on a distinguished road
yes, they told me it is not possible.
fipellac 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


Similar Threads
Thread Thread Starter Forum Replies Last Post
On the block tridiagonal linear system zhanglei Main CFD Forum 2 July 24, 2000 06:15
On the block tridiagonal linear system zhanglei Main CFD Forum 0 July 18, 2000 08:55
Good library to solve huge linear system Ricardo Bonon Main CFD Forum 3 May 19, 2000 06:24
More advise on libraries to solve linear system Ricardo Bonon Main CFD Forum 0 May 18, 2000 18:37
Agglomeration multigrid to solve Euler equations Jian Xia Main CFD Forum 3 January 11, 2000 19:07


All times are GMT -4. The time now is 13:30.