XANSYS Message: 34719 [Go back to message list] [bookmark on del.icio.us]
No rating yet Subject: Re: Regarding substructuring Author: Herbert Liu Date: 2002-03-08 10:12:00I have not used substructuring for quite some time. At least not after the iterative solver became available. I remember the substructure requires the generation of FULL stiffness matrix at the size of master dof's instead of the diagonal matrices in regular FEA structure models. The rule of thumb (at least back then) was limiting your master dof's to no more than few hundreds. Does anyone know the approximate value of "# of master DOF's per Gb RAM" in running substructure ?
Herbert Liu CAE Team Leader Index Development Engineering HUSKY INJECTION MOLDING SYSTEMS (905) 951-5000 x2483 hliu@h...
-----Original Message----- Sent: Friday, March 08, 2002 3:16 PM To: xansys
Again we have a confusion. I have a big volume having 90000 elements. I have selected only boundary nodes those are around 3000. I have around 9000master dof's. But the system is creating a huge page file and its not solving in the gereration pass.
Thanks
Regards
Somesh
At 09:04 AM 3/7/02 -0500, you wrote: > I don't understand "the frontal solver can solve only 50000 DOFs" - >there >is no such limit. Perhaps you mean what the solver can only handle 50000 >master DOFs? If so, this sounds like an extremely large number. Is it >possible >that you have set all DOFs to be masters rather than just some boundary or >loading nodes? >John Swanson >Swanson Analysis Services, Inc.
>> -----Original Message----- >> From: Somesh J. Ghunakikar [SMTP:someshjg@i...] >> Sent: Thursday, March 07, 2002 9:11 PM >> To: xansys >> Subject: RE: [xansys] Regarding substructuring
>> I have a single volume meshed with 50000 elements, I can not solve in >> generation pass, its coming out by giving watson error. Since the frontal >> solver can solve only 50000 DOF's. The problem is I can not divide the >> volume also. >> Again in the use pass, while solving its taking only frontal solver. But I >> need atleast sparse. I need an iterative solver in the use pass. How I can >> do it.
>> Thanks
>> Regards
>> Somesh
>> IEL, Hyd. >> India
>> At 08:01 AM 07/03/2002 -0500, you wrote: >> > A slightly longer answer is that substructure generation is a partial >> >factoring of the set >> >of equations, down to the set of Master DOFS. The other solvers in >> ANSYS, >> >both iterative >> >and direct, do not have this capability, which leaves you with the >> frontal >> >solver. >> > Direct solvers might be modified to have this capability (Sparse, for >> >example), but I have >> >not seen any research on how this might be done with an iterative solver. >> >John Swanson >> >Swanson Analysis Services, Inc.
>> >> -----Original Message----- >> >> From: Bohlen, Dan (GEAE) [SMTP:dan.bohlen@a...] >> >> Sent: Wednesday, March 06, 2002 11:33 AM >> >> To: 'xansys ' >> >> Subject: RE: [xansys] Regarding substructuring
>> >> I believe the short answer is: you can't.
>> >> -----Original Message----- >> >> From: Somesh J. Ghunakikar [mailto:someshjg@i...] >> >> Sent: Tuesday, March 05, 2002 6:14 PM >> >> To: xansys >> >> Subject: [xansys] Regarding substructuring
>> >> Dear Users >> >> In substructuring , the generation pass has only frontal solver? How >> we >> >> can change the solver instead of frontal in generation pass.