Leaving Analysis after nonconvergence

[DanJordy]

I'm running an analysis in ANSYS to determine when pull-in occurs when a parallel plate capacitor is subjected to a voltage and a mechanical shock. When pull in occurs, ANSYS says that the solution does not converge. I am keeping the shock constant and running transient analyses at increasing voltage levels to determine pull in. Is there a way to exit the run completely (get all the way back to BEGIN) when the first pull-in occurs so that I can extract the voltage at which this occurs?
Thanks

 

[Drej]

Couple of questions.

I am keeping the shock constant and running transient analyses at increasing voltage levels to determine pull in.

I think by definition a shock cannot be constant (it is a stepped load), so I'm not sure what your thinking is on this.

Is there a way to exit the run completely (get all the way back to BEGIN) when the first pull-in occurs so that I can extract the voltage at which this occurs?

Why do you need to go to the BEGIN level? This sounds very much like a non-linear buckling analysis which we mechies carry out. If this is the case, it should be exactly the same procedure.
[ul]
[li]Run your transient analysis until non-convergence occurs (make sure you save all of the results for each substep using OUTRES,ALL,ALL).[/li]

[li]When the analysis fails to converge, ANSYS will create a results file with the set number 999999. You'll then need to go into the time-history postprocessor (/post26) and examine your results.[/li]

[li]Review the voltage-deflection curve in the /post26 at the point in your model where you expect pull-in to occur. Be aware that an unconverged solution does not necessarily mean that the structure has reached its maximum load![/li]

[li]The curve will have a characteristic shape, something like a function x^0.5. On this curve, the pull-in voltage should be the last but one result (the last result will be the unconverged solution and will be meaningless).[/li]
[/ul]

Let me know how you get on with it.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[DanJordy]

I'm applying both a (1) electric shock (stepped voltage) and (2) base acceleration (ACEL). The base acceleration is a half sine pulse, which does not vary in duration and magnitude.

The *do loop that I am running increases the magnitude of the voltage every loop of a transient analysis. The problem that occurs is that when non-convergence occurs, it says "Solution not converged at time XXXXX (load step X substep X) Run Terminated". Clicking Proceed, it then prompts "During file base input (/input or macro) X warnings occurred. Should input processing be suspended?" (X warnings refer to the solution commands that I am running, but are not recognized in BEGIN.) Sometimes clicking YES does not stop the analysis from running (the voltage is incremented further and the transient analysis is restarted). Other times it is ok.

I can get the results, but I just need to check to make sure that the voltage that it is reporting is the lowest voltage that pull-in occurs at.
Thanks

 

[Drej]

I have a few issues:

1) Have to be careful with terminology. A base acceleration is not something applied with the ACEL command. This is a body (inertial) acceleration, such as that induced by the force of gravity. It acts on the entire body hence its name. A true base acceleration is something different. The best example of this is a seismic acceleration applied to the foundation of a structure.

2) You say "The base acceleration is a half sine pulse, which does not vary in duration and magnitude.". This is contradictory. How can a sine wave pulse, being a function:

Amplitude = function(t)

not vary in time or amplitude? By your definition it has to.

The *do loop that I am running increases the magnitude of the voltage every loop of a transient analysis.

I think this is central to your problem. Why the need to do this? In a transient analysis you can vary the voltage wrt time. That's one of the reasons for running a transient analysis. (In fact, why do you need to run this as a transient analysis? Do you have properties that vary wrt time, or something else? Are inertial effects important?) You can then post-process your results in /post26 easily to obtain the voltage. I think the need to go back to the BEGIN level is then removed, as this is very unconventional.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[DanJordy]

Ok, for (1), I am mistaken. Body acceleration is what I meant.

For (2), I was trying to find the pull-in voltage for a given shock input. My original intention was to automate this process by running a transient analysis multiple times with a *do loop. Every time the *do loop runs, the voltage increases, but the shock pulse does not change. This was causing me too many problems, and it was easier to advance the voltage manually. I was able to get it this way, and may try to automate it later.