Coupling Steady and Time-Dependent Pr... Log Out | Topics | Search
Moderators | Register | Edit Profile

FlexPDE User's Forum » User Postings » Coupling Steady and Time-Dependent Problems « Previous Next »

Author Message
Top of pagePrevious messageNext messageBottom of page Link to this message

Matt Munson (mmunson)
New member
Username: mmunson

Post Number: 1
Registered: 09-2005
Posted on Friday, September 02, 2005 - 12:55 pm:   

In version 5 is there a way to pass the solution to a steady PDE to a transient PDE so that I don't have to re-solve the steady pde at each time step?

I have a set of scripts written by a colleague that run on v3, I cannot get them to run in version 5.

The first script solves for a steady electrostatic field: dx(sig*dx(phi)) and works perfectly. This script writes out the potential (phi) using the function "TRANSFER".

The second script tries to solve a time dependent convection-diffusion equation where the convection depends on the field.

dt(c1) = dx(D1*dx(c1)) - dx(velocity1*c1)

where velocity1 is defined (in the definitions section) as follows:

Transfer('phi.dat', phi)
velocity1 = uB + V0*mu1*dx(phi)

(uB, V0, mu1 are all constants)

Again, these scripts run perfectly in v3, but in version 5, I get an error that says "evaluation point (2689.02,7.97427,0) is outside TRANSFER mesh"

I see in the v5 manual that transfer can't be used with time-dependent, eigenvalue, or staged simulations, but there is no such prohibition stated in the v3 manual.

In version 5 is there a way to pass the solution to a steady PDE to a transient PDE so that I don't have to re-solve the steady pde at each time step?
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 446
Registered: 06-2003
Posted on Friday, September 02, 2005 - 04:00 pm:   

TRANSFER is the standard way to pass data from one FlexPDE run to another.

There is no restriction preventing its use in time-dependent, eigenvalue or staged simulations (tell me where you found that in the manual, so I can take it out).

There is no mechanism for reading different TRANSFERS at different times, so maybe that's what you were referring to. If you want your phi to be a function of time as well as space, then you will have to calculate it instead of importing it. (The disk space required to store an entire time history of every node would be prohibitive).

The implication of your remark about evaluation points is that "evaluation point (2689.02,7.97427,0)" is inside the transfer mesh. Is it?
Is the transfer 3D? If not, give a 2D reference point.
Is the point on the boundary of the transfer mesh? Perhaps roundoff error is making it look like it is outside.



Top of pagePrevious messageNext messageBottom of page Link to this message

Matt Munson (mmunson)
New member
Username: mmunson

Post Number: 2
Registered: 09-2005
Posted on Friday, September 02, 2005 - 05:02 pm:   

The statement that Transfer is not support for time dependent, eignevalue of staged prolems can be found on page 96 of the FlexPDE 5.0 reference manual.

My domain is 2D a rectange with corners at 0,0 and 3000,10: so with the exception of the reference to a z coordinate that does not exist, the point is within the domain.

I am noticing that the dimensionality of the problem was not specified in the old scripts. Having specified that at the begining of each script, I still get the same error.

It is not apparent to me how to control the position of this reference point that gives an error. I am not asking for the value at that point, I simply import the variable using transfer and then attempt to evlaute another quantity (dx(phi)) over the entire domain, and that is one point it decides to look for.

I am attaching my scripts. TGF.parameters.pde is just a bunch of constants (more or less), field.pde is the script for the field and should be run first to create phi.dat, and TGF.1.pde is the script for the time dependent problem.

application/octet-stream
TGF.parameters.pde (1.2 k)

application/octet-stream
TGF.1.pde (1.9 k)

application/octet-stream
field.1.pde (1.2 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 449
Registered: 06-2003
Posted on Sunday, September 04, 2005 - 07:20 pm:   

FlexPDE uses a quadtree of rectangular boxes to search for points in a TRANSFER mesh. Each rectangle contains a list of mesh nodes in or near the box, which provides a base for cell inclusion testing.

In your case, the extremely thin domain caused some of the boxes to have no nodes listed. This resulted in an erroneous report of "Point is Outside Transfer Mesh".

I have corrected this error in FlexPDE version 5.0.7, which is available for download from www.pdesolutions.com/sdmenu5.html


Incidentally, I urge you not to use fixed timesteps. Any fixed step size will be inaccurate when the solution is changing rapidly and/or too slow when things are not. This guarantees the least accuracy for the greatest cost. The automatic stepsize control gives a much better balance of accuracy and cost.

Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action:

Topics | Last Day | Last Week | Tree View | Search | Help/Instructions | Program Credits Administration