instruments used in flexural testing of honeycomb sandwich panels 1

[trish129]

Hello.

what exact relevance do strain gauges have in flexural testing of honeycomb sandwich panels ?

what particular property/variable can be calculated from the longitudinal strain data from the top and bottom face skins and how?location of strain gauges ?

we are primarily concerned about the stiffness values of the overal panel when testing as per ASTM D7250..

can strain gauges be done away with in favor of DIC{digital image correlation},given the constraint of having to install several gauges for a large number of test samples?

how much information can a DIC video camera provide for a given flexural test of a sample?be it 3-point or 4point?

 

[RPstress]

Strain gauges are still the most reliable way to assess the strain and also (provided modulus is known reasonably accurately) stress.

Sandwich panels usually have simple sheet skins of a fairly uniform thickness or a thickness which changes in a known way. Because the core also tends to be a consistent thickness the bending moment in the panel can also be assessed from surface gauge readings.

I have assessed four point bending of a sandwich beam with various repairs on it. Gauges were used on the unrepaired area of the central span at least half the beam width away from the end of the repair and the end of the central beam span (it should have been at least a full beam width but there simply wasn't room in that case). They were put on both skins; the skin modulus differed in tension and compression, which should not be ignored (usually analysis crudely uses the average of T and C). We also put gauges near the edge of the beam in one place to check strain and stress distribution over the width. We wanted to check the in-plane failure stress that the repair broke at to compare different repair types. The gauging worked well for assessing this.

A metallic skin may have its thickness adjusted by chem-milling and it's important to avoid gauging a skin above a chem-mill step—I have seen this careless idiocy committed on a flap skin. A laminated skin may have a graduated thickness and while it's normally easy to avoid a rapid change in thickness (might be 10:1 in places), usually it is possible to avoid ply steps in more gradual thickness changes. Small foil gauges can be about 1/16th of an inch long (1.5 mm) although usually for cheapness and accuracy they are about 1/4" (6 mm). Triaxial rosettes are a bit bigger and enable recovery of two direct strains and shear (cheapish ones are 1/2" across—they tend to be three gauges in either an equilateral triangle or a right angle with one in the middle at 45°).

Foil gauges can be used for dynamic measurements at several hundred or even several thousand Hz and can be used in impacts over times of a few milliseconds and give a useful history of the event (subsonic types of impact, not ballistic).

Optical strain measurement is coming on fast. We have used it (from GOM) and it seems it can rival strain gauges for both accuracy and sample rate. Soon it should surpass them if it hasn't already. Setup issues exist for both but are very different. The GOM system seemed to be able to sample an area about 20 mm across but that memory may be unreliable; the system may be improved anyway of course,

http://www.gom.com).

Usually we assume that with a sandwich panel the in-plane strain doesn't vary through the skin thickness but if necessary it can be allowed for with the usual engineer's theory of bending distribution of strain through the panel thickness.

It is possible to put a gauge in the inside of a sandwich skin but this has obvious problems (relieving the core in way of the gauge and its wires and getting the wires out and incorporating the gauge in the panel manufacture) and usually isn't done; well, I have never seen it done, although we've thought about it. Maybe someone else has seen it done.

So, gauges measure strain over an area somewhere between 1.5 mm and 12 mm across and you can usually at least estimate skin endload and panel bending from that, and if it is a panel rather than a beam, full biaxial strain and shear can be recovered.

With composite skins it is assumed with reasonable accuracy that the surface strain isn't affected by the orientation of the surface ply. It may be necessary to prepare a simple (but big) 3D FE model at the ply level to convince people (and indeed yourself) of this.

 

[trish129]

thank you for your insight again RP...

doesnt the ply layup stacking sequence affect the choice of location and orientation of strain gauges on the upper and lower skin?

 

[RPstress]

The nature of strain is that it cannot vary greatly from one layer to the next. It varies a bit, but cannot vary in a step from almost zero to a large value in the very next layer (like stress can). You still have a variation in strain in a laminate in bending; the variation is linear with thickness (or so Cauchy tells us—at least I think it was Cauchy; ETB anyway) but it cannot look like a plot of the stresses with sudden jumps from one layer to the next. It is also in its nature to be relatively insensitive to the ply angles. If a symmetric laminate is in tension in the X direction then it all the laminate will strain in the X direction (coupling like Poisson's may also cause it to strain in other directions and if it's asymmetric it will also bend). And if it's in bending then all the plies will strain in a linear distribution with thickness from +X on top to -X on the bottom. Well, ±X is if it's symmetric; an asymmetric laminate may not have a symmetric distribution, but it will still be a smooth, linear variation—see attached plot of strain for a (+45/0/-45/90) (asymmetric) layup in tension which also puts it in bending.

For a laminate with no bending (either applied or because of asymmetry) it should not matter where through the thickness the strain is measured. If it has bending (as a four point bend test article does) then it matters, and an asymmetric layup in bending will make the strain in one side of the laminate different from the negative of strain on the other side, but the angle of the surface ply should not affect the strain of the surface of the material. The overall behavior of the laminate is dictated by the whole laminate's layup, and it is this which dictates the behavior of the surface. In theory...maybe Cauchy was wrong...

For a sandwich bending specimen gauges should measure strains along the length of the beam, across it if Poisson's effects are of interest, and if the skin has coupling between endload and shear then possibly shear should be measured. There might also be a very small amount of twist is there's coupling between endload and torsion of the skin, but the presence of the core will make this almost negligible (the core will greatly reduce response of the skin to A and D matrixes coupling because out of plane deformation of the skin will be enormously reduced; for a skin to bend or twist the whole sandwich would have to do so and the distance between the two skins makes that hard to do).

 

[RPstress]

There was an issue with the attachment above. Please use the attachment on this post.

 

[adfergusson]

Strain gauges are helpful for determining when, and how badly, your flexure test starts to go wrong (especially in four point bending). Taking RPstress' comment as the basis for an example:

[/skin modulus differed in tension and compression]

And assuming RPstress means his gauges on the compression and tensile skins were showing different absolute strain values (which I've also seen loads of times) then this is indicative of indentation of the rollers into the skins that is sufficient to cause mechanical coupling between the sandwich panel and the test rig. Once this happens (which can be at very low loads for thick cores and/or thin skins) you're no longer measuring the response of the panel; you'r measuring the combined response of your panel and test machine.

For 4 pt bending, put a strain gauge between the two central rollers on both surfaces. What you will see is that the strain measured by both gauges increase in to the same degree at the start (different signs of course). Then, once the rollers start to indent into the panel the strain from the compressive gauge will start to reduce as the test rig starts to contribute to the response of the panel load; the overall stiffness response of your panel will appear to increase. As you keep loading, the stain in your compressive gauge will go to zero and you'll find your panel can appear to be up to almost four times stiffer than you might expect from calculated predicitons. This is because the panel and rig are now effectively joined together and the compressive surface has actually become the neutral axis of flexure for the combined response of the panel and test machine.

You can use the part of your load vs displacement curve where the two strain gauges track each other to back calculate flexural rigidity. Any data taken from after the strain gauges diverge will be give you complete rubbish for a beam.

NB For biaxial flexure, e.g. like this:

)
divergence between strain gauges on tensile and compressive surfaces can be expected due to membrane stiffening and does not indicate that the test is going wrong

 

[adfergusson]

Regarding DIC:

DIC is great for a number of applications but standard flexure testing of honeycomb cored beams would probably not be one of them.

IF you use DIC from side on then you'll have troble getting strain data through the panel due to the geometry of the honeycomb. From a side on view, you also won't be able to get in plane strain data right at the surface of you samples as you'll have regions/facets in your images that are not occupied by your sample. Macro lenses and a much smaller field of view (FOV) would help resolve this but you'd the have the problem of your sample moving out of your FOV as the test proceeds.

To visualise strain in the skins from above or below you would need a 3D DIC setup which is more expensive and requires a calibration procedure before you can use it. This will also have problems as displacements to failure in flexure test are typically quite large. To get around this you could stop down your lens to give a greater depth of field (DOF) but then you'd need powerful, uniform light sources to avoid high iso settings/grainy images. Also, once you start getting to F.8 and beyond, optical aberrations and drops in image sharpness tart to come into play.

I started using DIC a decade ago with ARAMIS and DIC is becoming a progressively more capable and powerful tool. However, I think you'd be disappointed by it for your requirements that you've stated so far. Under other circumstances, e.g. if you were going to look at a real component undergoing a scale test too destruction, or your panels had foam cores then I'd say give DIC a look. If your skins are NCF or woven composite material then 2D DIC is definitely worth a look for regular tension shear, etc...tests; I couldn't imagine calibrating damage models for such composites without it. For 2D DIC you also need not shell out tens of thousands of dollars on software from the likes of GOM or LAVision if you've got some hardware and software programming skills as there are a few people/groups who've published some details on how they've done it (mostly in Matlab, but there are examples of the various functions needed out there in numerous other languages that you could port from or code with). We now use our own DIC system where data acquisition is done by a LabView programmed, FPGA controlled system (normally very expensive, unless you buy them off eBay!) and the analysis is done in a Matlab based program but with some functions based in other languages (for speed and because I don't feel the need to shell out hundreds or thousands of dollars to Mathowrks for a few extra functions/toolboxes). The learning curve is pretty steep but once you get the hang of it you'll have a fully customisbale and flexible tool (e.e. our system works much better with synchronising strain gauges, accelereometers, etc.. with DIC data than Aramis does) tool that can optimised and automated for your applications.

 

[trish129]

thank you Fergusson for building on RP's argument and elucidating on a "latent" dimension of this testing...

i am trying to absorb what you guys shed light on...this is what i call "getting the feel of mechanics"...

doesnt the 25mm wide steel blocks on the loading jaws enunciated by ASTM c393 and 7249 address this mechanical coupling due to indenting rollers?

i am using 600mm length for 4point bending as per d7249 and 200mm length for standard 3point bending as per C393/D7250.as per the guidelines for applying strain gauges i have no problem attaching a strain gauge in the geometric centre of the 600 mm specimen as the loading blocks(25 mm wide each) have a load span of 100 mm and the gauge just fits in un-hindered..but the problem comes with attaching the gauge on the 200mm sample as the single loading pad is directly incident on the centrally placed gauge .

 

[adfergusson]

Steel blocks, pressure pads, etc... all sorts of things you can try and do to prevent local contact problems. Whether they will suitably resolve the inherent problems for what you are doing (your panel has a fairly thick core and thin skins I seem to recall) is best determined by simply doing the tests, just put some strain gauges on some samples and do some tests.

On the three point bend just offset the strain gauge a measured amount from the middle.

 

[trish129]

tested a sample of 600mm length..there came a loud cracking noise and i halted my test there..skin de-bonded from the middle..betwen the loading pads...strain reading registered for the top skin but 0 reading for lower skin..why?

 

[RPstress]

The only likely mechanism is that the compression skin underwent a wrinkling failure although the curvature of the deformed shape would have lessened the chance of that. Check out NASA CR 912 Shell Analysis Manual §3.50 (

http://ntrs.nasa.gov/search.jsp?R=19680015330)

or NASA CR 1457 Manual For Structural Stability Analysis of Sandwich Plates and Shells (

http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA310684).

Just Googling the names should give those URLs.

I take it that the compression skin debonded if it happened between the loading pads.

Did the entire length of the skin between the loading pads debond? A wrinkling failure may just have debonded a shorter length (I have seen pictures of this but never seen it happen).

What was the stress in the compression skin at debond? This may tie in vaguely with the wrinkling allowable.

I also take it that the gauge on the tension skin didn't register strain at any point in the test so 0 for the lower skin either means the skin didn't strain (no possible way for that to be true that I can think of—how much did the test piece deflect when the debond happened?) or that the gauge was faulty. Either it disbonded from the skin or the electrical connection was defective. It pays to have several gauges even if they're measuring different aspects of the part (a gauge monitoring strain at the edge would give a reading even if the primary gauge in the middle failed).

With a simple bend test the deflection data alone will give almost enough data for calculating skin stresses and preliminary allowables.

 

[adfergusson]

[/I also take it that the gauge on the tension skin didn't register strain at any point in the test so 0 for the lower skin either means the skin didn't strain (no possible way for that to be true that I can think of—how much did the test piece deflect when the debond happened?) or that the gauge was faulty. Either it disbonded from the skin or the electrical connection was defective. It pays to have several gauges even if they're measuring different aspects of the part (a gauge monitoring strain at the edge would give a reading even if the primary gauge in the middle failed).]

^This seems the most likely situation to me to. Did you measure the resistance across your gauges just prior to your test to check that the connections were working?