FEA partners ready?

pete_yodis

Ive got a poor man's cloud with 3 machines here at work running FEA on an enclosure being designed to resist a 621 psi hydrotstatic overpressure test.  It's painfully slow, but really necessary - so we soldier on.  Are any partners ready for folks to give their system a try?  I'd love to use my Onshape account as a gateway to FEA.  I think that would be a great driver of business to Onshape as well.  It would be awesome to run the analyses I've been running and will continue to run this week in a much more capable platform.  We often stop considering geometry variances because we run out of time in analyzing.  There are plenty of more scenarios I would like to run to also question the assumptions we make in FEA studies.  Running more scenarios has the advantage of playing one's own devil's advocate and understanding when the FEA model is sensitive to certain assumptions you are making.  Are you partners willing to let a hack get an early taste?

sergio_p_

Hi Pete,

Why is taking so long to solve? Is just an internal pressure and lineal analisis with contacts? Should no take more than a few minuts to solve, maybe a good meshing of the part will be more painfull than solving. SW is not good at solving big models, you should try another solver. If you want some advice (or maybe professional work), just get in contact.

Regards

pete_yodis

Thanks Sergio.  I just may.

pete_yodis

Today, my IT forced Windows update is messing with me...

pete_yodis

Sergio_P_ said:

Hi Pete,

Why is taking so long to solve? Is just an internal pressure and lineal analisis with contacts? Should no take more than a few minuts to solve, maybe a good meshing of the part will be more painfull than solving. SW is not good at solving big models, you should try another solver. If you want some advice (or maybe professional work), just get in contact.

Regards

No penetration contact on large lid and main body is driving the solution time up.  I need to understand what is happening in the joints for this particular study as well as the rest of the enclosure (flameproof/explosion proof enclosure).

sergio_p_

Some basic advices after 15 years of doing FEA :-)

You need a model that can be solved very quick and be enough precise (not necessary 100% accurate), in order to allow you to made several desing iterations, try different configurations, materials or boundary conditions in razonable time. The more quick the model solve, the more time you have to make mistakes and learn from that. So...

1) If your part is simmetric, then use in your simulation. You can reduce the size of your model to a half, and in terms of solving time maybe 5-10 times less.
2) Remove all the small/finishing radius of your model prior to mesh. You must model your part having the FEA in mind (leaving the small radius, simmetries as the last modeling features, so you can roll back easily). This will reduce the quantity of elements and increase his quaility (shape). After the first runs you must check where are the stress hotspots and just add the radius in that areas, the same for mesh improvement (you don't need a perfect mesh in all the part, but must be accurate at least in the stressed areas)
3) Contact is iterative process very time demainding, so you must try to avoid it and replace for another kind of idealizations in the FEA model, at least for the initial runs. 

Regards, and we keep in touch

pete_yodis

Thanks Sergio.  The parts are not symmetric enough.  Numerous other runs have been made at more simplistic levels.  Now I am onto simulating the contact areas - since these are critical.  I must use no penetration contact, so I must pay the penalty of long solutions times.  I could reduce some of the radii outside of the contacting area, but I think the number of elements in the contact area is really a large factor in what's driving the long solution times.  I wonder if a much more powerful platform to compute on would make a difference in solution times anyway.

Regards.

sergio_p_

Maybe an outsource contractor for the FEA when you reach the limit of your actual software solution?

Regards.

By the way, if you are interest in the contact areas then all the radius can be removed. Can you show a picture of your mesh maybe?

pete_yodis

@sergio_p_ Check out these folks.  Impressive webinar today....

http://www.simsolid.com/

sergio_p_

@pete_yodis, thanks for the advice. I had scheduled the presentation but due to another meeting I almost lost it, just caught a glimpse of the last slide of the ppt with the Simsolid link. I set up an account on Friday and now I'm waiting to receive his confirmation. Even worst, I was answering you on the forum and my computer hangs up in the middle :-(

I recognize that meshing (and preparing the geometry) is the most demanding task in the FEA process, now I have started to work with open source tools like Salome-Plataform, Netgen and Calculix and I can really understand the advantage of comercial codes in terms of features, support and productivity. I'm very anxious to see this "mesh less" process, have heard on the past but never sought on action.

Do you know if is possible to review the recorded webinar in some place?

Regards.

pete_yodis

@sergio_p_ I do not know if this webinar was recorded.  Maybe @joe_dunne can tell us.

michael3424

I'd like to see a recording of the webinar as well - signed up but something else came up and I missed it.

dan_1

@michael3424 @pete_yodis @sergio_p_


the recording link: https://www.onshape.com/videos/simsolid-partner-webinar

sergio_p_

Thanks @dan_1!

As @pete_yodis says, very impressive. Doing FEA on assemblies is a very time consuming task, and very often at the end you consider "everithing is glued" to get at least a first results and then focus on the stressed areas.

Regards

michael3424

dan_1 said:

@michael3424 @pete_yodis @sergio_p_


the recording link: https://www.onshape.com/videos/simsolid-partner-webinar

Thanks!