| Author | Message | ||
     f. e. k. (faysal) Junior Member Username: faysal Post Number: 3 Registered: 03-2007 |
When resolving a pde equations : L: div( k*grad(L) - k_i*grad(L_i) ) = 0 L_i: div( k*grad(L_i) + k_i*grad(L) ) = 0 with Natural conditions : Natural(L) =givenvalue1 (x,y,z) Natural(L_i) = givenvalue2 (x,y,z). We have observed that flexpde generated different number of cells ( then meshes) accordingly to the chosen functions givenvalue1(x,y,z) and givenvalue1(x,y,z) values. How to keep the same mesh while changing the givenvalue1(x,y,z) and givenvalue1(x,y,z) values. As we need to test a set of functions over the same mesh. Thank you, F & T. | ||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 780 Registered: 06-2003 |
The principle of FlexPDE is that the mesh should be refined until the solution is within the given error tolerance. Input conditions that create sharper curvatures of the solution will refine the grid more. This is all part of "getting the answer". However, there are several ways you can defeat the default behavior: 1) SELECT REGRID=OFF This will compute a solution on the initial mesh and stop, regardless of the estimated error. Use manual mesh control options NGRID, MESH_SPACING, etc, to create the reference mesh you want to use. 2) Run the problem once with TRANSFER output. Then on subsequent runs import the mesh with TRANSFERMESH and turn of the REGRID as above. Then all subsequent runs will use the final refined mesh of the first run. |