Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Iterative Solver Not Terminating After Converging?
Posted 28 avr. 2017, 15:58 UTC−4 3 Replies
Please login with a confirmed email address before reporting spam
I am attempting to model acoustic scattering. I've started with a fluid sphere so that I can compare the results to an analytic solution to get an idea of the mesh sizing/memory I would need for my actual models. When using the direct solver, I get results that compare well with the analytic solution, but my actual models will require too much memory using the direct solver, so I've started trying to get the iterative solver to work on the spherical scatterer.
I am using the GMRES solver with a multigrid preconditioner and default SOR smoothers.
My problem is that the iterative solver doesn't seem to terminate after reaching the designated error level. I've set the relative tolerance to be .01. The error on the convergence plot gets to that level and then seems to jump to where it was at the beginning, it continues and again jumps once it has hit the tolerance level. Is there a reason for this seemingly large discontinuity in the error level? Some sort of instability in the solution? Here is a picture of the convergence plot:
I am using the GMRES solver with a multigrid preconditioner and default SOR smoothers.
My problem is that the iterative solver doesn't seem to terminate after reaching the designated error level. I've set the relative tolerance to be .01. The error on the convergence plot gets to that level and then seems to jump to where it was at the beginning, it continues and again jumps once it has hit the tolerance level. Is there a reason for this seemingly large discontinuity in the error level? Some sort of instability in the solution? Here is a picture of the convergence plot:
Attachments:
3 Replies Last Post 11 août 2017, 03:07 UTC−4
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
When using the direct solver, I get results that compare well with the analytic solution, but my actual models will require too much memory using the direct solver, so I've started trying to get the iterative solver to work on the spherical scatterer.
When you say that you're trying to get the iterative solver to work on the spherical scatterer are you talking about using the exact same fluid sphere model that gave you a convergent solution with the direct solver? Same geometry, physics, and mesh?
--
Nick
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
When you say that you're trying to get the iterative solver to work on the spherical scatterer are you talking about using the exact same fluid sphere model that gave you a convergent solution with the direct solver? Same geometry, physics, and mesh?
Yes, the same geometry physics and mesh. Just changing the solver from direct to iterative.
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
Hi Scott,
It's really hard to help you find out the solution to your problem without detailed information, like what model did you use, boundary conditions, discretization of your varaibles and so forth.
While based on the info you provided so far, there are something you may want to try if you haven't done yet.
1. If you couple with CFD module, did you add a point constrain for pressure otherwise your solution won't be unique which is usually not the issue for direct solver but the convergence will fail in iterative solvers.
2. Check your mesh quality.
3. Keep the stabilization technique enabled.
Could be more things to cause your issue but we really lack of enough background
Yi
It's really hard to help you find out the solution to your problem without detailed information, like what model did you use, boundary conditions, discretization of your varaibles and so forth.
While based on the info you provided so far, there are something you may want to try if you haven't done yet.
1. If you couple with CFD module, did you add a point constrain for pressure otherwise your solution won't be unique which is usually not the issue for direct solver but the convergence will fail in iterative solvers.
2. Check your mesh quality.
3. Keep the stabilization technique enabled.
Could be more things to cause your issue but we really lack of enough background
Yi