Author |
Message |
Philipp Reichenbach (reichenbach)
New member Username: reichenbach
Post Number: 1 Registered: 07-2006
| Posted on Wednesday, July 12, 2006 - 12:40 pm: | |
Hello I have two questions: 1. How can I force the program to keep the timestep Dt above a minimal value? In the Problem I try to calculate there are big time derivations only in small parts of considered region, but the timestep falls down and the calculation gets very long. 2. When I have a x-y-function f(x,y), how can I get a new function g(x,y)=f(x,constant) which is also defined on x-y-plane, but only depends on x? |
Philipp Reichenbach (reichenbach)
New member Username: reichenbach
Post Number: 2 Registered: 07-2006
| Posted on Thursday, July 13, 2006 - 09:27 am: | |
Hello, I'm sorry, but I have some more questions: - When I calculate my problem, the time derivation dP/dt of my dynamical variable P has at first its maximum at 1e15, and the timestep dt goes below 1e-13 (with errlim = 0.1). Later the time derivation falls down to values of below 100, but dt remains at 1e-13. But I need to calculate the problem over a longer time (until 1e-6). How can I increase dt when the dP/dt decreases? - The other question concerns the function g(x,y)=f(x,constant): If you have a function u(x), ís it possible to define a similar function g with g(x,y)=f(x,m(x))? |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 655 Registered: 06-2003
| Posted on Thursday, July 13, 2006 - 02:05 pm: | |
1. The FlexPDE timestep control attempts to find a step size at which the time behavior of the variables is reasonably approximated by a quadratic function in time. The timestep will be cut to tiny values if the temporal behavior cannot be approximated in this way. This usually occurs if you have written inconsistent equations. Boundary conditions that cannot be realized, parameters that jump discontinuously in time, and other similar errors can cause FlexPDE to struggle. An ad hoc override that forces larger timesteps would merely mask the fact that the equations are wrong, and would probably go unstable. The real solution is to discover why the equations are in conflict and fix it. Start by taking MONITORs of all variables and parameters at CYCLE=1 intervals. Plot histories of meaningful quantities. Try to find indicators in these plots of things that are not what you expected or wanted. This should give you clues as to what is amiss. 2. The FlexPDE scripting language allows a high degree of flexibility in defining values. The customary use is merely in cascaded definitions, such as A = x*y B = 12*A You can also define values in a functional form: F(a,b) = a*b+13 {for example} Using this latter form, you could define G(c) = F(c,D) where D has been previously defined. See the
|
Philipp Reichenbach (reichenbach)
Junior Member Username: reichenbach
Post Number: 3 Registered: 07-2006
| Posted on Friday, July 14, 2006 - 08:56 am: | |
Hello, thank you for your answers, but I need the function g(x,constant)=f(x,y) when f is a defined dynamical variable. If I define f as a variable and write then g(x)=f(x,y) in the definitions, then there comes only an error report: "Illegal item encountered". |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 657 Registered: 06-2003
| Posted on Friday, July 14, 2006 - 02:32 pm: | |
You can also use VAL or EVAL (q.v.) e.g. G = EVAL(f,x,const) |
|