| Author | Message | |||
     scorpion Member Username: scorpion Post Number: 21 Registered: 03-2007 |
Hello everyone, could you please tell me - can I somehow restrict minimum cell size? I don't want too small cells because then I get too many variables and it takes too long to solve my equations. Could you also please see my file and tell me why my plots start at t=40 not t=0. Thank you very much.
| |||
| rgnelson Moderator Username: rgnelson Post Number: 1337 Registered: 06-2003 |
1. FlexPDE refines the mesh in order to represent the shape of the solution as accurately as possible. In your case, you have very sharp spikes in the P distribution, so FlexPDE puts a dense mesh in this area to resolve the shape of P. (Your very large threshold on vx and vy has effectively disabled error checking on these variables). If you inhibit the mesh refinement, you are degrading the accuracy of the computation. You can do that if you want to, in several ways: a) SELECT REGRID=OFF This disables the mesh refinement and retains the initial mesh. You may see oscillations in the solution. b) SELECT NODELIMIT=<number> You can restrict the size of the mesh by controlling the total number of nodes. If the NODELIMIT is larger than the initial mesh size, the remaining free nodes will be distributed in the areas of highest error. c) If the strong spikes in the solution are not physical, you can try to discover errors in your boundary condition specifications. Correcting them might remove the necessity for dense meshing. 2. If you do not specify an initial timestep, FlexPDE will try 1e-5 times the total requested time range. It will then iterate on the initial step size, trying to find a value for which the temporal behavior of the solution is polynomial. You have not specified an initial timestep, and you have an extremely large end time. As a result, the initial timestep is 40. Initial iteration apparently thinks the solution is polynomial at this value, so it starts off with a stepsize of 40. The first plot at t=0 is presented, but all the variables are uniform value. The next plot is then at the end of the first step, or t=40. Try specifying a reasonable initial timestep, like TIME 0 TO 2000/omega BY 1 Once the problem is running, FlexPDE will constrain the timestep to match plot times. But due to an oversight, this restriction is not imposed on the initial step (I guess we thought no one would ask for plots inside the first step). Your plot requests at intervals of 5 prevent the timestep from growing larger than this, even if FlexPDE thinks it could use larger timesteps. |
