Author |
Message |
Dominic (dodo)
Member Username: dodo
Post Number: 5 Registered: 09-2005
| Posted on Wednesday, January 04, 2006 - 02:25 am: | |
I am trying to solve the desorption rate equation [ dt(theta)=-v*exp(-E/(R*T))*(theta)^n ] for a laser induced thermal desorption process. This works fine if I assume an isothermal process. If I want to model the desorption with a temperature dependence I run into a problem. I tried different kinds of dependencies (URAMP / Polynomial / Quadratic) but I always get the same error message: Invalid Floating-Point Operation Called from dmatrix: lu_sol / lu_solve – measures: measure_zgalerkin – timesolve: tmeasure I found a way to use the quadratic dependency but I have to multiply the time by a factor of 1000 to avoid the message. This makes no sense to me. Can somebody help? (In the end I would like to use the polynomial temperature dependence.)
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 514 Registered: 06-2003
| Posted on Wednesday, January 04, 2006 - 08:38 pm: | |
Your equaton is dt(theta) = A*theta with A = -6e4*exp(-28/t^2) 1. Evaluation at t=0 immediately causes a divide by zero. (Version 5.0.8 did not halt this way, so apparently it never evaluated at t=0). 2. A grows rapidly to large negative numbers. By t=2.4, A=-500, so the equation is effectively dt(theta)=-500*theta. Errors in theta propagate into dt(theta) multiplied by 500. You have specified a threshold of 5 for theta, so the spatial solver allows errors of 0.005 (ie., ERRLIM*THRESHOLD), creating oscillations of +-2.5 in dt(theta) at a time when theta has fallen to 0.001. 3. If you really want to track theta into the noise, you should model LN(theta) instead. Otherwise, set a stop condition on theta<small. 4. You are modeling a zero-dimensional equation with a 1-dimensional mesh. You should add a div(grad(theta)) on the right to stabilize the spatial coupling. You should also turn off the regridder, since spatial resolution is not an issue, but the large error magnification interacts with the regridder.
|
Dominic (dodo)
Member Username: dodo
Post Number: 6 Registered: 09-2005
| Posted on Thursday, January 05, 2006 - 02:02 am: | |
Thank´s again for your kind help!! |
|