Notifications
Clear all

ANSYS Usermat Time Step control

7 Posts
4 Users
0 Likes
395 Views
tuvw080
Posts: 2
Topic starter
(@tuvw080)
New Member
Joined: 14 years ago

Hello,

in ABAQUS/Standard the global time step size can be controlled from within umat using the variable PNEWDT. I wonder, if there is any variable in ANSYS Usermat that do the same.

Thanks

dnoack

6 Replies
Jorgen
Posts: 3964
(@jorgen)
Member
Joined: 3 years ago

Have you looked at the ANSYS user manuals?

An alternative to PNEWDT, that works in all FE codes, is to return a crazy stress. That way equilibrium will not be reached and the increment attempted again.

- Jorgen

6 Replies
bw_composite
Posts: 124
(@bw_composite)
Estimable Member
Joined: 15 years ago

Jorgen,

your last comment about crazy stresses sounds very interesting but regretfully I do not think I understand it at all.

Would you mind extending it with just a couple of lines?

Many thanks

Best Regards

Muzialis

Reply
Jorgen
Posts: 3964
(@jorgen)
Member
Joined: 3 years ago

I have not used the crazy stress approach, but I assume it should work as follows. Implicit FE increments are repeated until equilibrium is reached. If you return bad stresses then equilibrium will not be reached and the solver will be forced to cut time.

This, of course, is only used for implicit simulations.

- Jorgen

Reply
bw_composite
Posts: 124
(@bw_composite)
Estimable Member
Joined: 15 years ago

Jorgen,

many thanks for this clarification

Muzialis

Reply
Jay
Posts: 10
 Jay
(@Jay)
Active Member
Joined: 15 years ago

Keycut variable

Hi,

in ANSYS Usermat you will find the keycut variable. If you set it to 1, ANSYS performs a bisection due to the ANSYS solution logic. A direct control of dt is not (yet) possible.

Armin

Reply
tuvw080
Posts: 2
Topic starter
(@tuvw080)
New Member
Joined: 14 years ago

Hello all,

thank you for the hints and discussions of my problem. I recognized that there is nothing available in ANSYS for controlling the time increment efficiently from usermat.

@Jorgen

Returning a crazy stress is not a clean solution, since it can produce wrong results or crash the job because of numerical problems in the FE-program.

@Armin

The keycut variable seems to be a good option but is limited (yet) to bisectional cut. If the time step size should be reduced with factor eight, the keycut should returned three times.

Unfortunately both possibilities decreases the execution performance.

Many thanks

dnoack

Reply
Share: