FEA 64 bit problem

[ing]

We have recently switch from XP to win7. Basically 32 to 64 bit. The simulation results in the 32bit vertion of inventor has always been second to none. But the new (new to us) 64 bit vertion has yeiled some false results. more worriingly the results indicate that the model is in fact good enough to do the job. I have attached a document outlining my findings.

 

I have contacted my software supplier and they reported it to autodesk. Autodesk has in turn said that the problem is noted but the case has been closed.  I have no problem accepting a minor deviation, even 5% but the results can have an error of 0% to 80%.

 

If anyone knows a cure or has had this type of trouble, please let me know.

[Ray_Feiler]

Autodesk 360 link to simulation data.

http://a360.co/1irYgJ3

[blair]

received this in my mail this morning: Dear Blair, Thanks for your e-mail. We have created new Case # 08977274 for the below request and we have transferred it to the Product Support team, who should contact you shortly. Best regards, Simona Golisciano Global Customer Service & Operations

[ing]

Thank you Blair, my autodesk reseller has also contacted me to say that a case has been opened for this particular issue

[ing]

All of the features in the parts are rolled back in your project. Can you correct it and resend it?

[Ray_Feiler]

You can do that by opening each IPT and pulling down the End of Part marker.

[LT.Rusty]

I'm having fun and games with this as well.

I've got a set of parts where I've done the real-world testing and found that it takes X lbf to create a .25" deflection, and the FEA is telling me that it takes X*6 or so lbf to get that deflection, when using Inventor 2013 ... but only about X*3 to get that deflection on 2014.  Oh, and it's also telling me that LDPE is actually about .75% stronger than HDPE.  Haven't run the parts through Sim Mech yet.

I'm using 64-bit software in both cases.  I may download a 32-bit trial version of 2013 and see if there's a difference.

(Data sets are proprietary.  Can't post 'em here, but can share with Autodesk.)

[LT.Rusty]

Lovely. Can't install 32-bit Inventor on a 64-bit machine, and I don't have a 32-bit machine handy to work with. That's spectacular.

[ing]

I have first spotted the problem with inventor 2013 64 bit. Lol I figured that someone else was going to bring it up. I'm not suprised that you are getting skewed results.

[LT.Rusty]

Real world results using a semi-generic HDPE gave me a reaction force of 230-250 lbf to cause a .25" displacement.  2014 said 635 lbf.  2013 said 1200 lbf.  2013 says that LDPE should give me about 1300 lbf, too, which is even more interesting.

[ChrisMitchell01 · Accepted solution]

All,

A fix for this issue is being worked on for inclusion in the next update for 2014. That should be released early January. The changes to the FEA results are due to the way that contacts are being calculated differently in 2014, compared to how they were in 2013.

There is a workaround to return to previous behaviour by creating a registry setting:

[HKEY_CURRENT_USER\Software\Autodesk\Inventor\RegistryVersion18.0\Assembly Stress Analysis]
"WizOutputContactMaxPoints"=dword:00000002

Sample Reg file also attached.

Setting this now will give the same end-result as you will get after installing the upcoming update.

Hope this helps.

Chris

[ing]

You guys Rock!! Thank you for this quick fix. This finally puts our minds at ease.

[ljordan]

I just added this registry key to my machine.  I ran into something interesting myself.  I ran this prior to adding this key, generated a report, and then repeated this after the registry key was added.  The scariest difference that I see is I had a reaction moment on my fixed contraints, as expected, prior to the key, but then after the update I do not have any reaction moment at the constraints!

Please note that "Stress Analysis Report 11_22_2013.mht" is prior to adding key, and "Stress Analysis Report 11_22_2013-revised.mht" is after.

Again, all I did was hit Simulate, close file and Inventor, add key, reopen file, and hit Simulate again. 

[ljordan]

Here is jpegs if nobody wants to download the actual files.

[ChrisMitchell01]

Hi,

Try to ‘dirty’ the simulation first in order to force a re-solve:

1. RMB on the Simulation in the broswer => Edit Simulation Properties
2. Change the contact tolerance value very slightly from default (e.g. 0.1mm to 0.1001mm or 0.004in to 0.004001in)
3. OK the changes => this will dirty the contacts, mesh and results
4. Re-solve

If you want to dirty only the results, change a load value slightly and re-solve.

This info will be included in an upcoming Technical Solution about this issue.

Thanks
Chris

[ljordan]

Thanks! I created a new Simulation altogether thinking that would fix it, but it didn't. I will be honest, I thought you were nuts, but it worked!

[QuasiMojo]

Are you saying that the solutions are somehow cached?  I cannot just recalculate without little weird tweaks like that?  Re-solve doesn't always re-solve?

[Anonymous]

THIS IS HORRIBLE!

Now at this moment, thousands of users around the world are performing analyses with not accurate results without knowing anything about it. I do not even dare to think about what can happen around the world now, with designs based on not accurate calculations. The only reason why I am aware of this problem is that a colleague of mine recommended me to take a look at this page.

The only way to handle this, is that Autodesk immediately deliver an update and send out warnings to all the users.

WHY THIS SCILENCE AUTODESK?

Dan

[pball]

Well FEA analysis should always be taken with a grain of salt. I always try doing a simplified calculation by hand to compare with my FEA results.

[ChrisMitchell01]

The "official" fix for is planned for inclusion in the next update for 2014, (early January). That will essentially do the same as setting the registry value documented below.

Thanks,
Chris

[RANDYBLOSSER3231]

Chris,

Is there a simple validation problem you can provide that we can run to check that the quick fix and upcoming service pack have indeed fixed this issue?

Randy