FEA Verification/Validation

[gurmeet2003]

What accuracy can be expected from FEA stress results as compared to experimental test results? Recently I have heard from experts that agreement within 10% is achievable for both linear and non linear analysis.

In machine parts high stresses usually occur in areas of stress concentrations (fillets etc.). Stress gradients at these locations are also very high. A slight error in placement of strain gage at these locations could lead to significant error in results. To avoid this should one measure stresses only in low gradient areas?

I would appreciate it if you would share your experience.

Thanks,
Gurmeet

 

[rb1957]

you're right, i doubt you can accurately determine the stress peak (of a stress concentration) with strain gauges. i have worked with strip gauges (a row of axial gauges), but i remember this as a qualitative assessment.

sure your FE results may be 10% out, but this is still the best way to experiment with your design; and finalise on the one that works best for you. you can analyze the fatigue performance either by hand calc or by packaged code (ncode for example). if fatigue is critical for you, i'd suggest testing the final part.

good luck !

 

[GregLocock]

For frequencies I'd be glad to see 5% for say the first 6 modes of interest. For global static deflections I'd also say 5% is OK, but I wouldn't be wild about it as a result.

Both of these would be second iterations, typically.

The implied error of the above for local strains is >>5%, and in practice I think errors of 100% are common and not unworkable (eg near welds where local yielding takes place during the first test cycle).







Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[gwolf]

I routinely get well under 5% on stress and static and dynamic stiffness. This is plenty good enough because it would be a brave engineer who claimed to know their loads to within 5%.

If you want to correlate the stress at stress concentration features you must measure the strain in a low stress gradient location, align your FE model with the strain gauges and then use the FE model to predict the stresses at the point of interest.

This type of activity is highly dependent on the quality of the FE model. Garbage in, Garbage out.