Assembly Stress Analysis; simulation completes with no results

[greg.szkilnyk]

Hello all,

Having some difficulties with Inventor Stress Analysis on some assemblies I am working on (gearboxes).  I have successfully run simulations on about 6 assemblies of varying complexity (30 parts to 200 parts).  Those assemblies that have a higher part count only do so because of roller bearings or bolts in the assembly (i.e. each roller/bolt is an individual part).  

However, for 3 other assemblies that are VERY similar in number and types of parts I cannot seem to get any results.  The contacts (manual and automatic) compute, the mesh will generate, the simulation will complete but NO results are present in the browser.  I have re-simulated several times with the same outcome.

I am using adaptive refinement with 5 iterations, a 5% convergence rate on VM stress (preferred), 0.75 refinement threshold.  0.1 average element size, 0.2 min element size, 1.5 grading factor, 60 deg max turn angle, Create Curved Mesh Elements and Use Part-Based Measure For Assembly Mesh are CHECKED.  (These settings worked for the 6 successful sims).  I have toyed with convergence from 10% to 2% with no success.

I am using Inventor Professional 2017, 64-bit, Build:196, Release 2017 SP1 - Date: 7/13/2016

I have Windows 10 Pro on a multi-core (10) computer; Intel Xeon CPU E5-2640 v4 @2.4 GHz

Any insights as to what might be happening?  The entire assembly takes 12 hours - 24 hours to run so probably not worth attaching.  I will continue troubleshooting but thought somebody might have an idea?

Thanks,

Greg 

[TheCADWhisperer]

I recommend reading Vince Adam's book on FEA.

I have never found a reason (except for Frame Analysis) to have that many parts in an analysis.

I suspect that this process could be simplified to fewer parts in the analysis.

[LT.Rusty]

What CAD Whisperer said. No need whatsoever to have that much going on in your assembly for FEA.

If you're doing analysis on a top-level assembly, simplify it as much as possible. You should be doing detail work on bolts / rollers / etc at a micro level, not macro.

[johnsonshiue]

Hi Greg,

This is a very interesting case. When you said you tried other similar models and they worked fine, are they geometrically similar? Does this problematic one contain more detail geometry? You said you did not get the result. Was there any error message?

Many thanks!

[greg.szkilnyk]

Hello,



They are geometrically similar (though not identical and have some different
features) and I would say they have roughly the same level of detail.
Strangely, no error message comes up and nothing seems to show up in the
log; the progress bar shows that the simulation has finished but no results
are present in the browser.



I have since been able to make each faulty assembly run now by excluding
several of the parts; but haven't identified if one part in particular is
causing the problem.



Thanks,



Greg Szkilnyk

Design Engineer

Ker-Train Research Inc.

497 Discovery Ave.

Kingston, Ontario,

Canada, K7K 7E8

email: greg.szkilnyk@kertrain.com

main office: 613.531.3155 ext. 108

www.kertrain.com



This transmission is intended only for each correct addressee. It may
contain confidential, privileged, proprietary and commercially sensitive
information("Information"). Any unauthorized disclosure, copying, or
retransmission of this message or any Information is strictly prohibited. If
you have received this transmission or any Information in error, please so
confirm by sending us a return email copy immediately, then delete the
original message.



By responding by email to a communication sent by us, or by sending us an
email, you expressly confirm your continuing agreement with and consent to
the exchange of Information by unencrypted email, and expressly accept the
risk of interception or unauthorized receipt of email transmissions and
Information by other parties. We will have the right to continue
communicating with you, your advisors, agents and employees, by unencrypted
email and to transmit Information using unencrypted email to any of them,
until this consent is withdrawn by written notice to us.

[mord_g]

I routinely run the "unrecommended" static simulation on assemblies with more than 350 parts.

There are several conditions which can cause trouble in a simulation.

1)  Re-mesh.  Force it if needed.  Change mesh parameters by a very small value (like one degree difference in min. turn angle)

2) Verify material definitions have not changed- also, when a material is used in one library shares the same name in another library, there can be issues.

This also applies to any custom material definitions which are in the individual part, but are not defined in a material library or have different values.

3) force model update before entering stress environment.

4) try creating a copy of the simulation- delete the original, review convergence settings, and re-run the simulation.

5) UPDATES- make sure you have all inventor updates (2017-R4), update your GPU drivers, operating system, etc.

Above is a short list, there are also problematic welds, impossible meshing from slivers of a solid, face drafts, etc., running out of swap space on C: drive, issues with anti-virus software...

I've had luck with difficult parts by specifying a custom local mesh control for them. 

Some parts had to have a larger mesh specified to allow simulation, others needed a tighter mesh.

  Seemed counter-intuitive, but I've had complex lofted plastic parts require a local mesh value of 0.4.

[greg.szkilnyk]

Thanks for all the recommendations!  I am working through the list now to see if anything helps.

Just to note:  the majority of the time that I have encountered these problems, the software has allowed me to create the INITIAL mesh (i.e. clicking on Mesh View) successfully before running the actual simulation that did not provide results.  Does this eliminate any of the items from your list?

[greg.szkilnyk]

For each problematic assembly, I have eliminated bolts, rollers, etc. and reduced the part complexity down to roughly 8-10 parts (4 parts in one case).  Unfortunately, same outcome (no results).

[m.minajev]

I have the same problem now.. did you manage to solve it and understand the cause?

[johnsonshiue]

Hi! I am not sure if your issue is the same as this one. This thread is specific to certain assemblies with detail geometry. Your problem is that FEA does not even generate a result for a simple block.

Many thanks!

[m.minajev]

You've got a point there, yet my simulation is complete but no results are present in the browser.. The assemblies/models are different, but result is the same..