Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project!

## RE: [Libmesh-users] Nonlinear hanging nodes

 RE: [Libmesh-users] Nonlinear hanging nodes From: Roy Stogner - 2005-06-27 19:13:12 ```On Mon, 13 Jun 2005, KIRK, BENJAMIN (JSC-EG) (NASA) wrote: > As I mentioned, the other missing piece is the ability to take an arbitrary > global vector and extract the element solution vector from it, accounting > for constraints. I'll implement that shortly and check it it. Did this ever make it into CVS? I was actually a little confused by this discussion - I thought that hanging node DoF values were correctly set by the solver, and that (in serial) the correct values could thus be pulled straight out of the solution vector. Certainly my own nonlinear adaptive code seems to be converging correctly. What I have run into is a problem in parallel - nonlinear matrix assembly in general needs coefficient values from neighboring processors, and mine hits that bounds-checking assert() when trying to get them. I could do the "cache everything in a serial vector" hack again, but it would be nice if there was a solution that only had to communicate inter-processor boundary DoFs rather than every DoF. --- Roy ```

 RE: [Libmesh-users] Nonlinear hanging nodes From: KIRK, BENJAMIN (JSC-EG) (NASA) - 2005-06-10 13:50:07 Attachments: Message as HTML ```I plan to add such a method to the DofMap after 0.5.0 is released (which should be tonight). The interface is something like this: DofMap::build_elem_solution (global_solution, dof_indices, local_solution); For non-constrained Dofs this will simply set local_solution[i] = global_solution[dof_indices[i]] And for constrained Dofs the constraint matrix will be built and applied to the solution vector. Thanks for your patience... -Ben -----Original Message----- From: libmesh-users-admin@... [mailto:libmesh-users-admin@...] On Behalf Of Michael Povolotskyi Sent: Friday, June 10, 2005 6:27 AM To: libmesh-users@... Subject: [Libmesh-users] Nonlinear hanging nodes Dear Libmesh developers, I solve nonlinear equation systems using the NonlinearImplicitSystem class. I'd like to use the adapative mesh refinement for the nonlinear system. Does libmesh have a tool that can handle the hanging nodes problem for a nonlinear system? Thank you, Michael. ------------------------------------------------------- This SF.Net email is sponsored by: NEC IT Guy Games. How far can you shotput a projector? How fast can you ride your desk chair down the office luge track? If you want to score the big prize, get to know the little guy. Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20 _______________________________________________ Libmesh-users mailing list Libmesh-users@... https://lists.sourceforge.net/lists/listinfo/libmesh-users ```
 RE: [Libmesh-users] Nonlinear hanging nodes From: KIRK, BENJAMIN (JSC-EG) (NASA) - 2005-06-13 18:20:58 Attachments: Message as HTML ```Sure... The constraints are applied to an *element* residual vector or Jacobian matrix much in the same way they are applied for the linear case in ex10. So, if you assemble the residual and Jacobian element-by-element you simply need to call dof_map.constrain_element_matrix(...) and dof_map.constrain_element_vector(...) before inserting them into the global residual/Jacobian. As I mentioned, the other missing piece is the ability to take an arbitrary global vector and extract the element solution vector from it, accounting for constraints. I'll implement that shortly and check it it. I've got a modified version of example 13 which illustrates the use of the nonlinear solver, but it does not address adaptivity. I'm leaning instead to creating a new example altogether... -Ben -----Original Message----- From: Michael Povolotskyi [mailto:povolotskyi@...] Sent: Friday, June 10, 2005 9:02 AM To: KIRK, BENJAMIN (JSC-EG) (NASA) Subject: Re: [Libmesh-users] Nonlinear hanging nodes Thank you for the answer! Could you, please, tell more details on how to apply these constraints to the Jacobian matrix and to the residual vector. Michael. KIRK, BENJAMIN (JSC-EG) (NASA) wrote: I plan to add such a method to the DofMap after 0.5.0 is released (which should be tonight). The interface is something like this: DofMap::build_elem_solution (global_solution, dof_indices, local_solution); For non-constrained Dofs this will simply set local_solution[i] = global_solution[dof_indices[i]] And for constrained Dofs the constraint matrix will be built and applied to the solution vector. Thanks for your patience... -Ben -----Original Message----- From: libmesh-users-admin@... [mailto:libmesh-users-admin@... ] On Behalf Of Michael Povolotskyi Sent: Friday, June 10, 2005 6:27 AM To: libmesh-users@... Subject: [Libmesh-users] Nonlinear hanging nodes Dear Libmesh developers, I solve nonlinear equation systems using the NonlinearImplicitSystem class . I'd like to use the adapative mesh refinement for the nonlinear system. Does libmesh have a tool that can handle the hanging nodes problem for a nonlinear system? Thank you, Michael. ------------------------------------------------------- This SF.Net email is sponsored by: NEC IT Guy Games. How far can you shotput a projector? How fast can you ride your desk chair down the office luge track? If you want to score the big prize, get to know the little guy. Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20 ; _______________________________________________ Libmesh-users mailing list Libmesh-users@... https://lists.sourceforge.net/lists/listinfo/libmesh-users ; -- ------------------------------------------------------------ Michael Povolotskyi, Ph.D. University of Rome "Tor Vergata" Department of Electronic Engineering Viale Politecnico, 1 - 00133 Rome - Italy Phone + 39 06 72597367 Fax + 39 06 2020519 http://www.optolab.uniroma2.it/pages/moshe/moshe.html ; ------------------------------------------------------------- ```
 RE: [Libmesh-users] Nonlinear hanging nodes From: Roy Stogner - 2005-06-27 19:13:12 ```On Mon, 13 Jun 2005, KIRK, BENJAMIN (JSC-EG) (NASA) wrote: > As I mentioned, the other missing piece is the ability to take an arbitrary > global vector and extract the element solution vector from it, accounting > for constraints. I'll implement that shortly and check it it. Did this ever make it into CVS? I was actually a little confused by this discussion - I thought that hanging node DoF values were correctly set by the solver, and that (in serial) the correct values could thus be pulled straight out of the solution vector. Certainly my own nonlinear adaptive code seems to be converging correctly. What I have run into is a problem in parallel - nonlinear matrix assembly in general needs coefficient values from neighboring processors, and mine hits that bounds-checking assert() when trying to get them. I could do the "cache everything in a serial vector" hack again, but it would be nice if there was a solution that only had to communicate inter-processor boundary DoFs rather than every DoF. --- Roy ```