XANSYS Message: 38712 [Go back to message list]
[bookmark on del.icio.us]
No rating yet
Rate item:

Subject: Re: [struct] Sparse solver failure
Author: Bogdan Balasa
Date: 2002-06-12 11:32:00

Ricardo,

We have noticed a very similar behavior in ANSYS 5.6 and 5.7, for modal
analysis with Block Lanczos solver activated. When you model contains a
mixture of SOLID 45/95 and SHELL 43/93, the amount of memory required to
solve the problem was really huge. At that point in time, the only solution
was to switch to a way larger UNIX box.
Later on, we have noticed that replacing the bricks (SOLID 45/95) with tets
(SOLID 92), also in combination with shells, leaded to way smaller amount of
memory required, so the corresponding problem could be solved on a PC also.
More ANSYS 5.7 offers dynamic memory allocation, for INTEL also, so you
might try with smaller values for the workspace (256 workspace and 64 for
DB).
ANSYS claims that newer versions (6.0 and 6.1) requires less amount of
memory for Sparse solver than 5.6 and 5.7.
Another option will be, if you have access to the ANSYS 6.0 with Parallel
Performance option, the new AMG solver should perform better than PCG for
ill conditioned problems.

Hope this helps.

Bogdan Balasa
General Manager
INAS SA
1 Tehnicii Street, 1100 Craiova, Dolj, Romania
Phone: ++40 51 438 789
Fax: ++40 51 438 906
Email: bbalasa@i...
WWW: www.inas.ro

----- Original Message -----
To:
Sent: Wednesday, June 12, 2002 5:48 PM

> Hello everybody.
> My model's features are:

> element types:
> SOLID95
> SHELL93
> TARGE170
> CONTA174
> number of elements: 46993
> number of nodes: 168161
> solver: sparse

> The solver crashes. The output and error files at the crashing moment look
> like this:

> OUTPUT FILE

> *** ELEMENT MATRIX FORMULATION TIMES
> TYPE NUMBER ENAME TOTAL CP AVE CP

> 1 36400 SOLID95 40.500 0.001113
> 2 1 TARGE170 0.000 0.000000
> 3 4210 CONTA174 3.109 0.000739
> 4 1832 SHELL93 1.969 0.001075
> Time at end of element matrix formulation CP= 81.21875.
> FORCE CONVERGENCE VALUE = 116.3 CRITERION= 1.170
> MOMENT CONVERGENCE VALUE = 0.000 CRITERION= 1.389

> SPARSE MATRIX DIRECT SOLVER.
> Number of equations = 506236, Maximum wavefront = 31274
> Memory available for solver = 643.82 MB
> Memory required for in-core = 4263.52 MB
> Optimal memory required for out-of-core = 183.21 MB
> Minimum memory required for out-of-core = 46.00 MB

> *** ERROR *** CP= 973.672 TIME= 16:30:19
> Sparse Matrix Solver failure. Please check the error file (file.err)
> for possible causes.

> *** ERROR *** CP= 973.672 TIME= 16:30:19
> Bisection may be activated if it is a nonlinear analysis.

> ERROR FILE

> *** WARNING *** CP= 21.875 TIME= 16:11:59
> Previous testing revealed that 1179 of the 46993 selected elements
> violate shape warning limits. To review warning messages, please see
> the output or error file, or issue the CHECK command.

> *** WARNING *** CP= 28.734 TIME= 16:12:06
> An initial contact closure factor (ICONT) of 2.7471381E-04 has been
> selected for real constant set 2.
> 0***** process error reported by subroutine xdslfa
> see xdslfa abstract (ier = -505)

> 0***** process error reported by subroutine xdslsv
> see xdslsv abstract (ier = -702)

> 0***** process error reported by subroutine xdslsl
> see xdslsl abstract (ier = -600)

> *** ERROR *** CP= 973.672 TIME= 16:30:19
> Sparse Matrix Solver failure. Please check the error file (file.err)
> for possible causes.

> Since strange subroutines are getting angry, I'm thinking everything's due
> to insufficient workspace. I've asked for 1200 megabytes total and 300 for
> database: NT doesn't allow me to ask for more, even if I've got 2 GB RAM.
> Can anybody please help me?
> P.S. = PGC runs (VERY slowly).

> Thank you very much

> Riccardo Testi
> R&D and Operations 2 Wheels/Engine Design
> Piaggio & C. spa
> v.le R. Piaggio, 25 - 56025 Pontedera (PI) - ITALY

Posts possibly associated with message #38712AuthorDateScore
38704[struct] Sparse solver failureRiccardo Testi2002/06/12 
38707Re: [struct] Sparse solver failureYan, Hong (Bolton) hyan@2002/06/12 
38708Re: [struct] Sparse solver failurePeter Attar2002/06/12 
38711Re: [struct] Sparse solver failureGene Poole2002/06/12 
38712Re: [struct] Sparse solver failureBogdan Balasa2002/06/12 
38754Re: [struct] Sparse solver failureRiccardo Testi2002/06/13