Author |
Message |
Issue Ho (ih1209)
New member Username: ih1209
Post Number: 2 Registered: 12-2004
| Posted on Monday, January 03, 2005 - 03:51 pm: | |
In the esample "harmonic.pde", the equations were broken into real and imaginary parts. FlexPDE required assigning variable to the beginning of each equation. I reversed the order of the variable assignment and the solution would diverge (rms error up to 1e9). Why is that? _____________________________ These work Ok Ur: dx(Srx) + dy(Trxy) + rho*omega^2*Ur + omega*mu*del2(Ui) = 0 Ui: dx(Six) + dy(Tixy) + rho*omega^2*Ui - omega*mu*del2(Ur) = 0 But these don't Ui: dx(Srx) + dy(Trxy) + rho*omega^2*Ur + omega*mu*del2(Ui) = 0 Ur: dx(Six) + dy(Tixy) + rho*omega^2*Ui - omega*mu*del2(Ur) = 0
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 294 Registered: 06-2003
| Posted on Monday, January 03, 2005 - 06:33 pm: | |
The equation identifier links the equation to its boundary conditions. By changing the labels, you have moved the force boundary condition onto the other equation. You also moved the Ur terms off the diagonal of the coupling matrix. This may have an impact on the iterative solver. |
|