|
[Sponsors] |
August 26, 2005, 02:53 |
About risiduals
|
#1 |
Guest
Posts: n/a
|
Hi,
Two questions about residuals: 1. Ones it did a simulation with 4500 iterations. The residuals went to a minimum after ab. 1500 iterations. But right after that, they are increasing again... what happened? 2. I have noticed, that the number of iterations depends on the number of cells in my geometry. For simple 2D geometry it takes maybe a few hundred iterations until residuals go flat, for the same 3D case it takes a few thousand. Is there a direct relation? It would be fine, to estimate the number of iterations before starting the simulation.. Thanks for your ideas! Ralf |
|
August 26, 2005, 04:15 |
Re: About risiduals
|
#2 |
Guest
Posts: n/a
|
1) Maybe you grid and/or solution setup is not that good?
2) In 3d more equations have to be solved. So it takes longer than 2d. I have not heard of a direct relation yet. The better the grid ist the faster the solver converges. |
|
August 26, 2005, 04:59 |
Re: About risiduals
|
#3 |
Guest
Posts: n/a
|
1) How can I increase the quality of my grid? And how to change my solution or setup?
2) There are two points: first: the complex 3D simulation takes more time for one iteration than for the 2D case (much more) and second: it needs more iterations for convergence. So, because there are more equations in the 3D model, each iteration will take more time, but why does it needs more iterations for convergence?? |
|
August 26, 2005, 05:25 |
Re: About risiduals
|
#4 |
Guest
Posts: n/a
|
1) Grid quality can be checked in gambit via examing mesh. Check EquiAngleSkew. The value should not be too high. Use a regular mesh (hexahedrons) or refine the mesh, if possible. You can check e.g. for large gradients via adapt gradient in Fluent. Did you see any inconsistencies in the solution due to the mesh? Did you choose the right boundary conditions, solver settings, models for heat transfer, turbulence, ... if necessary.
2)Due to more equations, which are to be solved and a possible coupling (e.g. pressure-velocity) between them I think the solution needs more iterations to converge. |
|
|
|