XANSYS Message: 42273 [Go back to message list]
Average Rating: 0 (2 votes)
Rate item:

Subject: Re: STRUC Combin14 spring-damper
Author: Sheldon Imaoka
Date: 2002-09-28 21:21:00

Hi Asha,

Is your problem taking 7-8 hours because the iteration is taking long
or because it is going into a nonlinear solution? I wasn't sure
from your email if you had one or both problems.

If you have KEYOPT(1)=1 and CV2=0, I don't believe that COMBIN14
should behave in a nonlinear manner. You can verify whether it is
doing so if the Convergence Monitor graphical tracking is being
displayed (L2 norms and CRIT). Checking the jobname.mntr file after
a solution will also confirm how many total iterations were performed.

On the other hand, if the solution with COMBIN14 is taking a long
time because the iterations is taking long, I would suggest using the
sparse direct solver (EQSLV,SPARSE) to speed up your solution. If
the COMBIN14 spring stiffness is too small, the matrix may become ill-
conditioned. The PCG solver (and other iterative solvers) will have
trouble if the matrix is too ill-conditioned. Look in the
jobname.PCS file for information on the PCG solver. If you run two
cases - with a large and small value for the spring stiffness,
compare the number of iterations required by the PCG solver in the
jobname.PCS file, and you'll probably see a drastic difference (order
of magnitude or so). Hence, for ill-conditioned matrices (i.e.,
cases where the spring stiffness is very small), use of the sparse
direct solver is preferred over the PCG solver, and the jobname.PCS
file will help you determine this.

Best regards,
Sheldon
ANSYS, Inc.

http://www.ansys.com/ (Official Site)
http://ansys.net/ansys/ (Cheap imitation)
"This email is not being provided in my capacity as an
ANSYS employee. I am solely responsible for the content."

--- In xansys@y..., "Asha Rani" wrote:
> Hi,
> I am conductiing a static structural analysis(only interia load) of
> a piping system where I am using combin14 as spring pipe supports
> hanging from the roof (ie. tension springs).
> Here are the details of the combin14 keyoptions used.

> 1. It is created with I-J nodes in the Z direction of the global
> cartesian system.
> 2. keyopt 1 = 0
> keyopt 2 = 3
> keyopt 3 = 0
> 3. only k value as real constant, around 360.

> Now the problem is it takes 7 to 8 hours to solve this problem,
when
> it should not take maximum of 2-3 hours.
> If I make the springs very stiff, it solves in an hour.Even when I
> have set the keyopt 1 to 0, which means linear analysis, it is
going
> into non-linear. I do not have any other kinds of non-linearity in
> the model.

> I am using Ansys 5.7, PCG solver with 1e-3 tolerance.

> Is there anything which I should be careful about Combin14. Can
> anybody suggest a way to come out of it?

> Thanks,
> Asha Rani

> Tata Consultancy Services
> Schenectady, NY
> Phone (518) 385 3489

Posts possibly associated with message #42273AuthorDateScore
42232STRUC Combin14 spring-damperAsha Rani2002/09/2710
42273Re: STRUC Combin14 spring-damperSheldon Imaoka2002/09/280