| Author | Message | ||
     quantum (noorbit) New member Username: noorbit Post Number: 1 Registered: 05-2007 |
I want to solve lorenz system,but the following code works correctly only t<40,when>40 the solution is wrong,why? ================================== select ngrid=1 { Since no spatial information is required, use the minimum grid } errlim = 1e-3 variables A V Z initial values A = 1.0 V = 2.0 Z = 3.0 equations A: dt(A)=-10.0*(A-V) V: dt(V) = 28.0*A-V-A*Z Z: dt(Z)=-2.8*Z+A*V boundaries region 1 start (0,0)line to (1,0) to (1,1) to (0,1) to close time 0 to 50 plots for t = 0 by 1 to 50 history(A) versus Z end ==================================== | ||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 842 Registered: 06-2003 |
It is unusual for FlexPDE ptoblems to run 40,000 timesteps, and we try to compress the history data list when the number of stored points exceeds 16000 (Based on the assumption that it is a history list, not a circulating path, and that a display can resolve at best a few thousand points). In your case, this destroys the A,Z relationship and garbles the plot. SELECT HISTORY_LIMIT=100000 will expand the size of the stored history list and preserve all your data. You should also include a diffusion term in each equation, to lock the nodal values together and prevent oscillations (since FlexPDE always thinks that a problem has a spatial component.) Turning off the regrid is a good idea, too. The resulting script is: ----------------------------------------- select ngrid=1 { Since no spatial information is required, use the minimum grid } REGRID=OFF errlim = 1e-3 history_limit=100000 variables A V Z initial values A = 1.0 V = 2.0 Z = 3.0 equations A: dt(A)=-10.0*(A-V) + del2(A) V: dt(V) = 28.0*A-V-A*Z + del2(V) Z: dt(Z)=-2.8*Z+A*V + del2(Z) boundaries region 1 start (0,0)line to (1,0) to (1,1) to (0,1) to close time 0 to 50 plots for t = 0 by 1 to 50 contour(a) contour(v) contour(z) history(A) versus Z end  | ||
| quantum (noorbit) New member Username: noorbit Post Number: 2 Registered: 05-2007 |
thank you very much flexpde is indeed a wonderful work |