Restarting Mechanica analysis after losing network license

[ninjaneer777]

Does anyone know if there is a way to get a Mechanica analysis to resume once it has lost it's network license half way through the run?

I've had Mechanica stop mid run, due to the license servers temporarily going down, a few times, and even though Pro/E quickly picks the license back up Msengine.exe can't seem to do the same. Is there a way to get msengine.exe to realize that the license is back and continue on with the analysis?

It would be nice to not have to restart a 2 or 3 day mechanica study from the start. I'm using Single Pass Adaptive, in case you wondering, because I think you can restart if using multi-pass adaptive, but that would not be very practical for the studies with lots of contacts, which these are.

Thanks!

 

[TheTEMP]

Multi-pass is the recommended method for Analysis with Contacts. Load stepping may help by opening or closing contacts. That may make your run faster or able to finish.

How do you Resume a multi-pass adaptive run?

 

[seymours2571]

Temp,

Care to elaborate on MPA being the preferred convergence method for simulations containing contacts? This is the first I have heard of this. I find your statement hard to believe.

Also, I think you are a little off base with the concept of load stepping.

Ninja,

I don't know of a way to restart a simulation. There used to be a method, but I never had much sucess with it.

2-3 days is a pretty long runtime. Is the element count high, number of contacts high, number of link elements (not rigid links) high? Also, is simulation being pushing into swapspace due to lack of physical ram? Also, is the hard drive that you are using to crunch the simulation a slow drive? All of these will dramatically affect performance.

The best way to tackle your problem might be to simplify the simulation to get the overall runtime down.

Steve


Stephen Seymour, PE
Seymour Engineering & Consulting Group

http://www.seymourecg.com

 

[TheTEMP]

Seymour
From Pro/E help files- Definition of -Single Pass Adaptive:
"While the single-pass adaptive option is available for contact analysis, we do not normally recommend its use because it can increase run time"

So if Single Pass is not reccommended then Multi-Pass must be recommended. Unless PTC doesn't have a recommended way to solve contact problem
I have been running some analysis with contacts that won't converge unless load stepped.

 

[seymours2571]

That is strange they would say that, given that a few paragraphs below that statment they have:

"In general, use single-pass adaptive convergence when it is available. Larger models run with single-pass adaptive convergence typically require less disk space. Also, single-pass adaptive convergence yields comparable results to multi-pass adaptive convergence of 10% with generally shorter run times."

Since the recommnedation they provide is based only on runtime and not accuracy, I would run it with SPA. I run contact simulations all the time using SPA sucessfully, however I do have a high powered server that does the number crunching.

Either way in my experience models which cannot converge the non-linear iteration during a contact analysis have situations where large changes in the required contact stiffness exists. You will find that other FEA programs will return error messages in events like this that say "the problem is ill-conditioned". Therefore, look for aspects of your contact model that are just physically difficult to solve. Load stepping is an option to assist with iterative convergence for problems with large changes in contact stiffness between iterations and it does help...sometimes.

Not sure what version you are using, but in your contact interfaces is infinite friction specified? I have not had great success with inifinite friction in any FEA package. It seems as though I always have iterative convergence issues using this type of friction specification.

Also, are your contact surfaces overly large? Meaing, contact between a ball bearing and a flat surface takes place in a very small area. However, many people will by default select the bottom half of the ball bearing and the entire flat surface as the contact interface. Sometimes you can use the split surface option in the contact interface, however it not always available depending on the referenced geometry.



Stephen Seymour, PE
Seymour Engineering & Consulting Group

http://www.seymourecg.com