Learning Forum

[Miguel]

How can it be that the same setup works successfully in a first run...:

...but fails in a second attempt with no changes at all?!!:

I first wanted to just introduce a small modification but since it kept failing I ran it exactly the same as the 1st successful run and, after working exactly and fine in the 1st loadstep, starts bisections and fails in the 2nd one!

Any explanation/solution? Thanks

[peteroznewman]

Something changed in the second attempt relative to the first attempt.

After I have a successful run like you did in the first attempt, I save the project and use File, Archive to keep a copy of the project in a .wbpz file that I can open if things don't go well in the second attempt.

[Miguel]

Thanks @peteroznewman. You´re right and I´m sorry: I double checked again and realized I had forgotten to suppress an object (joint) as it was in the first run...:

Still, I cannot see how to make it work with this additional condition activated, which is what I really need. Any idea would be welcome and appreciated.

[peteroznewman]

Are you saying the successful first run had this joint suppressed and with it unsuppressed, the convergence stalls?

In the Solution Information folder Details window, type a 3 into the Newton-Raphson Residual Plot box. This shows you which elements are out of balance. I expect it is the elements that this joint is scoped to.

You could edit the joint and change the Mobile end to Deformable instead of Rigid.

Check that the Coordinates for that joint are near to the scoped elements.

[Miguel]

As far as I can see yes, that is the only difference. Now I am running another try with some changes I introduced in the joint definition (I changed stops to locks and set a broader range to avoid interference with the contact scoped to the same part) but it takes a while to see if it works... I´ll let you know when it´s finished (if it does) and will also try your suggestions. Thanks again for them

[Miguel]

Hi again Peter,

it seems to be working so far although it´s taking even longer than I expected:

Hopefully, it´ll finish tomorrow (there are 10 steps in total) and I´ll be able to review the results. I´ll let you know.

Good night 

[Miguel]

Good morning Peter,

Unfortunately (and unexpectedly because it was going well), soon after my last message the run failed before reaching step 5:

I have the 3 NR-residuals I configured like you recommended and the error/warning messages but I can´t see how to use them to diagnose the problem... any suggestion? Thanks

[Miguel]

Too good to be true 😩..: I managed to add the additional step to an existing and working 5-step analysis, made it work successfully as well but..., when I compare the results in the first 5 steps of this last analysis with those of the original one ... they are completely different although the setup for those 5 steps remained unchanged! How can that be? I can´t understand it.

[peteroznewman]

It looks like the joint is scoped to a singe edge on the face of a solid mesh. The problem is a line has no area so the stress is infinite. The corrective action is to split the face with two lines to create a non-zero area to scope the joint to and that will reduce the peak stress.

[Miguel]

 

 

Well, in that visualization the crossbars appear solid (thick-shells&beams style on) but in the model only the corner joints are solid bodies and they are connected through beams from their holes (faces) with crossbars end-points (vertex), modelled as line-bodies (beam elements) and with brackets holes (edges) as surface-bodies (shell elements):

On top of that, bonded-contacts are defined between solid-joints, surface-brackets and board and crossbar-beams...

 

 

[Miguel]

I am definitely stuck and adrift. Would you be so kind to have a look at my files? Thanks.

[Miguel]

 

Hi again. I think I found the problem: somehow I had enforced displacements way too large and the structure deformed enormously until it could not go on and crashed…:

I changed the displacements for the actual forces (I started estimating the former trying to aid convergence…) and increased the nr. of initial substep and everything went smooth and faster:

Now I have to use these results to setup a low-cycle fatigue analysis, which is one of the original intended goals. Being the other one just 20 cycles of a similar load test. In relation to this, I started this other post (->). Do you know how to go about that?

Once more, thanks