Comparing FEA Results to Test Results - Beam Elements

[Astoria]

My company is gearing-up to conduct a structural test on a railcar and I have been charged with locating strain gages for the test. The gage info will be used to determine if yielding occurs and to validate the FEA model. The model is large and complex, made up of beam, plate and solid elements. The beam elements represent rectangular steel tubing members.

The beam stresses reported in our FEA analysis report are combined beam stresses, but we are limited to uni-axial and tri-axial (rosette) strain gages and cannot measure combined stresses. In addition, only one side of the tube is exposed in most cases. My questions are: 1. What information do I need from the FEA results in a beam element in order to compare them to the test results from a strain gage? 2. How do I get that information?

 

[SWComposites]

1) Strain gage measure strain, not stress.
2) Output the beam forces and moments at the strain gage locations, and calculate (by hand) the strains corresponding to the strain gage directions.

 

[OnlineFEASolver]

You need to place three strain gauges 120 degrees apart on the outer cross-section of your tube. There is an equation somewhere where you can calculate the internal bending moments and axial stress through the cross-section based on your 3 strain gauges. Then compare these moments and axial stress to your FEA model beam at that node.

If you need,I can dig up the equation for converting the strain gauge results.

 

[SWComposites]

The three gages will give the axial strain/stress distribution. If you want the shear stresses you need at least a pair of gages at 45° (if you have only torsional loads producing shear); more if you have both torsional and transverse shear loads on the beam.

 

[271828]

You can find all this business about the multiple strain gages in an advanced mechanics of materials book.

 

[prost]

Don't you normally have access to more information with a FE analysis than with experiments? It seems a bit odd that you have access only to 'combined stresses' (do you mean von Mises stress?) with the FE, but have more stress components available with the experiments; usually the reverse is true and the FE gives you more information than you will ever need. If you did the analysis yourself, just redo it and get the stress and/or strain components you need. If not, couldn't you get whoever performed the analysis to redo it?

 

[trainguy]

Check out Roark's Formulas for Stress and Strain, &th ed.

There's a whole section on strain gauges.

tg

 

[rb1957]

rosette strain guages can measure principal strain, allowing you to calculate the principal stress,
which you can compare with your model results.

 

[Astoria]

Thank you for your comments everyone (I was without internet yesterday). After reviewing them and doing a little investigation, I have learned the following:

1. The best place to locate a strain gage on a beam is at an existing node (versus at mid-element) because all the information I need is at the node. In addition, remeshing of the model will not be required in order to compare results.

2. The FEA program I use (FEMAP) places the stress recovery points of a tube at the corners. Obviously I can't place a gage on the corner, so my options are to average the information recovered at the two points on either side of my gage OR move the stress recovery points to the faces of the tube OR I can hand calculate strains based on forces and moments (as SWComposites suggests).

Thanks,
Astoria

 

[rb1957]

personally, i'd place the rosettes in the middle of the panels where the stress gradient is as flat as possible, and away from discontinuities (like edges, jooints, rivets, ...) that upset the stresses and that aren't modelled in your model.

I'd also suggest that you reconsider what your FE is telling you. most elements, because of their displacement assumptions, best predict the centroidal stresses; eg a 2-noded rod is really telling you the stress in the rod at the mid-span, and best way to determine stresses at the ends is to interpolate between adjacent elements.

for your tubes, i'd use 4 axial gauges, anticipating mainly axial load with some bending.

 

[Astoria]

rb1957,

Thanks for the tip. I think using 4 uniaxial gages (one on each tube face) is the best way to go, though I'll have to settle for 2 (on adjacent, perpendicular faces) in many cases because only 2 faces are exposed.

I disagree with your 2-noded rod statement. FEA is giving me information (forces, moments, stresses) at the ends (nodes) of the rod, not at the mid-span. Even when I display the "beam element maximum combined stress" the FEA program simply displays the maximum nodal stress for that element (perhaps not all FEA programs work this way though). However, I think your statement holds true for plate elements (i.e. the stress in the plate is the average of its nodal stresses).

 

[GregLocock]

I suppose it is beyond the wit of man to work out the mid span values from those at the ends.

Cheers

Greg Locock

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

 

[Astoria]

Greg,

Of course it isn't. Just trying to make things as easy as possible when comparing measured results to test results. It's advantageous to do so when you're placing 150 strain gages.

-A

 

[chaboche]

Hi Astoria,

I came across your thread while I was searching for information on beam to shell element connections in FE models.

I wondered whether you had now completed your experiments and how well the results compared to the FE model?

I find that the intersection between beam and shell elements can produce stress singularities and was looking for advise in this area.