CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Numerical problem (http://www.cfd-online.com/Forums/cfx/97161-numerical-problem.html)

 CrisCFD February 10, 2012 05:40

Numerical problem

Hello! I have inserted a constant in my model (dielectric constant), which is very low (2,21 E-10) and the simulation does not converge.
What can I do?

 ghorrocks February 10, 2012 06:23

You question is meaningless. What does the constant do? How is it coupled to the simulation? What are you modelling?

 CrisCFD February 10, 2012 07:27

Quote:
 Originally Posted by ghorrocks (Post 343755) You question is meaningless. What does the constant do? How is it coupled to the simulation? What are you modelling?
I previously described my case, in another thread...
I'm modelling a reverse electrodialysis system...so I have Na+ (cations) and Cl- (anions).
So I have two transport equation (cations and anions)
I have to consider the electrical interactions in the transport equation as an additional term, as a source term.
I just found an expression like this: 1) S+ = - σ / ε * (C-A) (cations)
and
2) S- = σ / ε * (C-A) (anions)
Where A and C are the concentrations of anions and cations, respectively
But the very low value of the costant "ε " (dielectric constant = 2,21 E-10) is a problem for the solver, a numerical problem, I think.
Also theese source terns which are dependent on each other could be a problem...I know it is a particular issue...
What do you think about it ?

 ghorrocks February 11, 2012 06:24

Just because a constant term is small does not tell you it has a big effect or small. It is how it interacts with the flow which is important. There will be some form of non dimensional number to describe these effects, pesumably the ratio of electrical forces to fluid flow inertial forces or something like that. That sort of number starts telling you how significant your constant term is.

 CrisCFD February 15, 2012 13:20

Quote:
 Originally Posted by ghorrocks (Post 343905) Just because a constant term is small does not tell you it has a big effect or small. It is how it interacts with the flow which is important. There will be some form of non dimensional number to describe these effects, pesumably the ratio of electrical forces to fluid flow inertial forces or something like that. That sort of number starts telling you how significant your constant term is.

 ghorrocks February 15, 2012 18:13

Starting simulation without knowing which effects are important is a reciepe for disaster. You need to do some back of the envelope calculations to work out what is significant and what is not before starting modelling.

 CrisCFD February 18, 2012 10:06

Quote:
 Originally Posted by ghorrocks (Post 344681) Starting simulation without knowing which effects are important is a reciepe for disaster. You need to do some back of the envelope calculations to work out what is significant and what is not before starting modelling.
My supervisor ordered me (:D) to consider these electrical interactions

 ghorrocks February 18, 2012 21:07

Do you know how many times it has come back to "my supervisor told me to"...

Regardless of what your supervisor says, do some back of the envelope calculations to check it is significant. And the degree of significance will tell you what sort of approach you need to take.

 CrisCFD March 6, 2012 12:06

Quote:
 Originally Posted by ghorrocks (Post 345147) Do you know how many times it has come back to "my supervisor told me to"... Regardless of what your supervisor says, do some back of the envelope calculations to check it is significant. And the degree of significance will tell you what sort of approach you need to take.
Experimental measurements showed that these terms are relevant. What can I do?

 ghorrocks March 6, 2012 17:06

My previous post explains what you must do. If the electrical terms are relevant, then what about the viscous terms? The inertial terms? If any of these are irrelevant then you can select the correct options to proceed. Otherwise you are just guessing.

But what is the physics of what you are modelling anyway? I seem to remember you asking about imposing a constant drift velocity on top the fluid velocity - is this the same issue?

 CrisCFD March 8, 2012 14:26

Quote:
 Originally Posted by ghorrocks (Post 347983) My previous post explains what you must do. If the electrical terms are relevant, then what about the viscous terms? The inertial terms? If any of these are irrelevant then you can select the correct options to proceed. Otherwise you are just guessing. But what is the physics of what you are modelling anyway? I seem to remember you asking about imposing a constant drift velocity on top the fluid velocity - is this the same issue?
I'm modelling the flux of cations and anions through electrodialysis membranes.
So I have to consider the electric interations as source terms in the transport equations (cations and anions). We have a laminar flow in the channel

 ghorrocks March 8, 2012 18:26

My question is what is the physics of what you are modelling. To model it you need to know exactly how each important force works because you need to describe an equation which accurately describes it.

Questions include:
* How is charge balance maintained? Does it have a conservation equation?
* How is motion of the charges coupled to the fluid flow?
* Is the charge motion independant of the fluid flow?

But my point about working out the significant forces needs to be considered first. It will tell you what approach you need to take and what things can be ignored. It is quite likely that CFX is not appropriate for what you are modelling at all - you will only know once you know what forces are significant.

 All times are GMT -4. The time now is 09:13.