Butt joint strength: epoxy versus screw

[rjlittlefield]

Several weeks ago I wrote that I was setting up a butt joint test of a particular epoxy that had been recommended to me. This is to report the results.

First, the joints to be compared:



The material is 0.365" diameter aluminum rod of unknown composition, purchased at my local Ace Hardware store. I used a carbide circular saw to slice the rod into short pieces, which I then fastened back together again using butt joints of two different types.

The epoxy joint at top was done by sanding the outer surface of the rod, cleaning all the cut surfaces with a paper towel soaked in xylene, then agitating the pieces in fresh xylene for 10 minutes, changing the xylene for fresh and agitating for another 10 minutes, then draining the pieces, washing them in acetone that had passed the evaporate-on-glass test for cleanliness, and finally allowing them to airdry. The epoxy was a fresh package of http://www.loctiteproducts.com/p/epxy_w ... mpound.htm, from which I carefully weighed equal parts of resin and hardener (1.67 gm of each, using an Ohaus balance). I mixed the two components on a paper index card using a piece of thick copper wire that had again been cleaned in xylene and acetone. Then I applied a layer of epoxy to both cut surfaces, pressed the two ends together to squeeze out excess epoxy, formed a small bead over the joint, placed the parts in a grooved jig to keep the parts centered and prevent movement, and left the parts to cure.

I'm telling you all this gory detail because I took great care to try to do a proper job forming the epoxy joint. If I have somehow screwed up any of the prep or application, I would be very happy to know how.

The screw joint, at the bottom, was formed by drilling and tapping both pieces (#23 drill, 10x32 thread, hand-tapped), then inserting a 3/4" long piece of 10x32 steel machine screw.

After letting the epoxy cure at room temperature (~72F) for 8 days, I tested the joints by clamping one end in a vice, hanging a 10 pound weight on the other end, and progressively moving the weight farther away from the joint until something interesting happened.

Here are some illustrative pictures of what happened.

First, one of the epoxy joints:



It's holding here at 1.5 inches from joint. It also held at 1.75 inches, but failed catastrophically (brittle break with no warning) at 2 inches. Another one held at 2 but failed at 2.5 inches.

Here are the broken surfaces:



In contrast, here is one of the threaded joints under load. I'll skip the early tests, since those were quite boring. Things started to get interesting around 9-1/2 inches, and here's the view at 10 inches.



It's pretty clear from the above that the joint is overloaded at 10 inches.

Nonetheless, I continued pushing the exercise. Here it is at 12 inches:



Bottom line, the epoxy joint in this test was about 4X weaker than the joint with threaded insert, and the epoxy failed without warning while the threaded insert gave ample indication of impending failure.

When I write that my experiences with epoxy have been less than completely satisfying, this is part of the behavior that I'm talking about.

--Rik

Edit: typo

[Peter De Smidt]

What a great test! The results match my experience with epoxy, which is that it's not reliable for a structural load, at least not with a butt joint.

[rjlittlefield]

...it's not reliable for a structural load...

My take is actually different from that.

I think epoxy can be very reliable.

But what's reliable is that it will behave no better than a brittle plastic with a tensile strength less than 1/10 of what you'd get from a pretty sleazy zinc-aluminum alloy such as "pot metal".

If that matches what you need for a particular engineering application, such as the bonding of brake pads that was mentioned in the other thread, then epoxy is great stuff.

But if you had in mind simply substituting epoxy for metal, or repairing a metal part that has already failed due to exceeding the metal's tensile strength, then I would strongly suggest taking another look at the situation because in my personal opinion there are some pretty obvious risks in that approach.

Loctite's description of this material as being suitable for "Repairing machinery..." strikes me as being highly optimistic, to the point of verging on fraudulent. I apologize if this seems harsh. I am only reporting a personal impression.

--Rik

[frankw]

Rik, great test, with results that verify my experiences with epoxy and aluminum. As a professional boatbuilder for 30+ years, I used epoxy(both West System and System 3) extensively, and in the last 15 years or so, almost exclusively in many projects involving vacume bagging various materials both similar and dis-similar. We spent quite a bit of time trying to bond aluminum (usually 6061). both to itself, as well as to other materials. We tried all of the recommended techniques of etching, prepping, etc, and always had an alarming failure rate, not acceptable for the given applications. Unless there have been some major advances in the formulation of epoxies, I personally would always trust welded and or drilled/tapped joins over epoxy when dealing with aluminum, steel, or any material other than wood to wood, or other various composites.
This is just my "real life" experience with the stuff. Not overly scientific, and admittedly, our shop conditions did not always allow for such meticulous prep work as you did in your test. I have to admit that I was impressed that your aluminum butt joint held as well as it did, a testimony as much to your"shop practices" as it was to the materials! My sense is that if you did not have the "bead" of excess epoxy around the butt joint, it might not have fared so well. Epoxy is an amazing substance when used within the parameters for which it was designed, which,in my experience, requires a certain amount of porosity in the materials involved.
Best,
Frank

[steveminchington]

Rik - Your test demonstrates that this particular epoxy has a lower tensile strength than steel which one would expect but it doesn't indicate that it is useless.
Engineering materials have a wide range of different properties and it is important to match a particular property to a particular application. For example, packing tape has high tensile strength along its length because the polymer chains are arranged that way, so it is very hard to break by pulling it. But it has a low shear strength crossways making it very easy to snap in that direction.
You wouldn't sue a drill company because your drill snapped when it snagged in a hole because you know it is designed to cut steel and with that hardness, it is also very brittle.
Even if you were to weld the rod with a butt joint, you wouldn't do it with squate ends because you know its not going to be very strong. You would chamfer each end and fill the resulting vee with three runs of weld.The load you applied to the rod created a compressive force at the bottom of the joint and a tensile force at the top. Put simply by the diagram, the resulting tensile load at the top of the joint is 24.46 Kg (53.9 lbs).



What you are trying to establish is whether epoxy will hold a focus block and camera in a vertical orientation (worse case scenario). Testing against a bolted joint doesn't prove anything as a bolted joint has a safety factor of about 100:1 in this situation. The aluminium casting will fail before the bolts.
You also have to consider the weakest link in this system which is the 1/4" x 20 screw that attaches your camera to the block. This has a cross sectional area at the root of the thread of 18 sq mm. Which is going to break first, the epoxy or the 1/4" screw?
It would be more appropriate to test the epoxy joint for its shear strength by cementing two pieces of plate together in a lap joint then drill a hole in each end and suspend it from one then hang the weights from the other. It should hold considerably more weight this way. If you replicate the same surface area as the back of the focus block, it will be a more meaningful test.
I did notice too that part of the failure was adhesion to the aluminium and not just fracture of the epoxy. I think if I were using this, I would drill a number of holes in each surface about 3mm dia x 3mm deep to give the epoxy a better key.

[g4lab]

Rik you always post such interesting things.

I notice in your post mortem pictures that that you had failure both in the epoxy itself and also failure of the epoxy aluminum bond. Can you reconstruct after the fact "which end was up".

I am reminded of the great aviation designer Ed Swearingen who designed a bunch of very cool aircraft including a two seat 300 hp hot rod. It was all aluminum and riveted construction because the aluminum crumpled in a crash versus carbon fiber composites snapping explosively. This was in the first heyday of all the composite homebuilts.

I like to use epoxies for repairs when it seems like a reasonable thing to do but I would never trust an epoxy metal bond to hold up my XK$ camera.
Steel might give better bond strength than aluminum but I have knocked epoxy bonds off of that pair too.

My first job out of college was at the Washington University in St. Louis, Sever Institute of Technology, Materials Research Lab. We did a lot of work on epoxies , mostly on fiberglass and graphite composites which is probably where they are the most suitable. The rule of thumb bond strength we used was about 3,200 lb/square inch in tension. We prepped joints in a similar fashion and stirred the epoxy under vacuum to boot.

[ChrisR]

If I had all the numbers for:
tensile, compression & shear strength; modulus, creep, high and low strain fatigue, fracture toughness, brittleness, temperature dependence and a few more,
and:
the consistency of each figure between batches of material,
and:
knew quite a lot about the behaviour of the joint between the materials and its dependence on surface preparation, and temperature changes,
then I could sit down with the mechanical engineering books I put in a box a long time ago, and try to work out whether the stuff would be suitable for an intended application.

Given that I'm not going to do that, I'd only use it if it felt ok, if I could use much more of it than I felt necessary, with an allowance for things I hadn't thought of.

[rjlittlefield]

Rik you always post such interesting things.

Well, I post out things that I find interesting. Apparently you and I have similar tastes!

Can you reconstruct after the fact "which end was up".

Despite my lack of planning for that question, it turns out the answer is yes.

In both cases "up" corresponds to a place where the epoxy peeled off the aluminum.

I have to admit that I was impressed that your aluminum butt joint held as well as it did, a testimony as much to your"shop practices" as it was to the materials!

Thanks. I will happily confess that what's shown above was my second attempt. An earlier attempt in which I was less meticulous about the surface prep was not so successful:



Bear in mind, this was still a freshly machined surface that had been cut with a dry carbide blade. I had expected that it would be clean by construction, but now it seems that expectation was wrong. This trial run snapped in my fingers before I set up a careful measurement. I have pretty strong fingers, though, so perhaps this joint was almost as strong as the ones I measured.

Your test demonstrates that this particular epoxy has a lower tensile strength than steel which one would expect but it doesn't indicate that it is useless.

And I don't believe I said that it is. Despite what I've written up here in the style of relentless critique, I actually use quite a bit of epoxy. In my glue box at this very moment are long-cure clear, 5-minute clear, 5-minute gel, long-cure for plastics, and this Loctite Weld. While I was waiting for the butt joints to finish curing, I repaired a screen door lock handle that I had broken by pushing it the wrong way. I expect that repair to hold well, but that's because I cleaned off the paint down to bare metal, then sandwiched the broken part between two splints made of 0.010" steel plate that I carved from an old feeler gauge, with Loctite Weld being used to hold the sandwich together. Most of the stresses in that case will be shear, and over a fairly large area. If I screw up and push the handle the wrong way again, I have no idea if the repair will be the part that breaks, and if it does break, nothing very bad happens. It's the perfect use.

Put simply by the diagram, the resulting tensile load at the top of the joint is 24.46 Kg (53.9 lbs).

I'm curious -- can you compute tensile psi at the maximally stressed point, taking into account the circular cross section of the rod? I'm curious to know how the breaking strength in these tests compares with the "3355 psi" that's written on the package.

It would be more appropriate to test the epoxy joint for its shear strength by cementing two pieces of plate together in a lap joint then drill a hole in each end and suspend it from one then hang the weights from the other. It should hold considerably more weight this way. If you replicate the same surface area as the back of the focus block, it will be a more meaningful test.

Sure, I have no disagreement with any of that. Mocking up the final design using the final materials and then testing to failure is always a good way to get reliable information.

In the case of the focus blocks, there's a surprising amount of variation in design aspects that will matter for mounting. At http://www.photomacrography.net/forum/v ... 984#162984, ChrisR and Chris S. collaborate to show us a CH2 focus block in which the back of the block is apparently a solid piece of metal, integral with the rest of the block. Lots of good gluing surface there. But on eBay, there was recently another "CH-2" block being sold that has a very different construction, with what appears to be a removable back plate and thus perhaps 5 or 10 times less integral metal to glue to. How does that play into the options? Beats me -- I haven't laid hands on the piece yet.



Anyway, in the "harvest" thread that I started, my main purpose is to gather information that other people can use to make their own decisions about how to accomplish their own goals. Part of that is to have some awareness for the capabilities and limitations of materials and techniques that make reasonable alternatives. Using a brittle plastic that sticks to things may make perfect sense in the end. I'd just like people to think it through first.

--Rik

[steveminchington]

I'm curious -- can you compute tensile psi at the maximally stressed point, taking into account the circular cross section of the rod?

Probably but it would be a lot more complicated than my simplified version. I was just showing how the load has substantially increased at the top of the joint.

I'm curious to know how the breaking strength in these tests compares with the "3355 psi" that's written on the package.

Yes, I would like to know under what conditions they achieved that figure.

I'd just like people to think it through first.

And you certainly get people thinking with your willingness to do these tests and share your findings in an interesting way. It's nice to see it stimulates some good debate.

[ChrisR]

can you compute tensile psi at the maximally stressed point, taking into account the circular cross section of the rod?

"One" could, by integrating across the circular area given a figure for the joint thickness. The load is spread across the joint, though some of it would be in compression, with a neutral axis I would guess somewhere between the lower edge and the middle.
Even then the joint interface characteristics would probably not be simple, which would throw things out.

This work on a sandwich rod shows a fairly wide range of figures for multiple for bend vs straight tensile strength, though with different materials:
www.jmst.org/fileup/PDF/pass189.pdf

A tensile test is done with a dumbbell shaped specimen, fat at each end and narrow centrally, made to a precise size, so the failure point is predictable and with known dimensions. The machine spits out the modulus, elastic limit, elongation to fracture, yield point, and Ultimate Tensile Strength.

For brittle materials, Fracture Toughness is often more important, where a specimen with a carefully made notch is set in a machine then snapped, often by a weight swinging on a pendulum. The energy absorbed in breaking it is recorded. (K1C, iirc)

The thermal expansion coefficient of epoxies I found is 2 to 3 times that of aluminium, ~ 45 to 60 vs 22 x 10^-6 per K. Elongation to fracture is a small number of percent, so it would take a large specimen for temperature to split the joint, but at small scale I expect there would be accumulating separations over a period with many temperature cycles.

I think we're safe to assume the interface is the weak area, so as Steve says, anything one can do to spread the load and increase the "key", would be good.

--

I don't recognise the CH stand in the picture. I have a couple, and they're solid backed, apart from a small "hatch " near the top. The focus knobs are also different. The stand machined in the pictures taken by Chris S, above, is a BH stand I think, though I'm not sure which version.

[rjlittlefield]

can you compute tensile psi at the maximally stressed point, taking into account the circular cross section of the rod?

"One" could, by integrating across the circular area given a figure for the joint thickness. The load is spread across the joint, though some of it would be in compression, with a neutral axis I would guess somewhere between the lower edge and the middle.
Even then the joint interface characteristics would probably not be simple, which would throw things out.

That last sentence turns out to be a notable understatement.

I took a stab at modeling the joint, using what was essentially a finite differences technique. At first glance this seemed simple enough. We just model the adhesive layer as a bunch of little springs, and calculate how much the springs have to stretch or squish as the joint is deformed. Spread the springs uniformly over the cross-section, however oddly shaped it might be, and see what happens.

Well, the first thing that happens is exactly what we'd expect from the geometry of the circular cross section. The springs in the middle end up just sitting there, not stretching or squishing very much at all, while the ones closer to the top and bottom get exercised in proportion to their distance from the pivot line. Most of the load ends up getting carried by a couple of horizontal bands of adhesive about 90% of the way from center to top and bottom, where the band is far enough above/below center to have some good leverage while still being a sizable fraction of the rod width. Very near the top and bottom of the rod, the adhesive band gets weak just because it gets narrow.

But then things get complicated. Whilst thinking about those little springs, it occurred to me that because of irregularities in the metal surface, some of the springs will be longer than others. Being made of the same material, the short springs will necessarily be stiffer, and that means they end up carrying more of the load. As we approach the material's breaking strength, the short stiff springs go first, and the problem ripples out from there. Off to the literature...

http://www.masterbond.com/techtips/unde ... -thickness has what may be a good summary of the situation:

In a bond joint, it is very important to have a uniform bond line thickness for optimal adhesive performance. Shim spacers can be used for keeping the bond line thickness uniform. In many cases, glass micro-beads can be added to the adhesive to create a uniform bond line thickness.

A quick perusal of the literature shows that thin bondlines are generally stronger than thick ones, so apparently it's a matter of "thin, but not too thin". Bit of a mess, really!

For what it's worth, the springs model reproduces a joint failure at 19 inch-pounds of torque when the bond tensile strength is 3000 psi and the pivot point is midway between center and bottom. There's a certain amount of numerology in this result, but it looks like the quoted number for bond strength is in the right ballpark for this experiment.

--Rik

[rjlittlefield]

I don't recognise the CH stand in the picture. I have a couple, and they're solid backed, apart from a small "hatch " near the top.

Likewise for the CHA that I have. But I see on eBay several other units that have the same style of frame with the open back: items 261759801186 (CHT), 171699959008 (CHT), 201292357344 (CHS), 271301014604 (CHT).

--Rik

[g4lab]

https://www.youtube.com/watch?v=YBQp04glQqc

Ben Krasnow has a YouTube channel called Applied Science. He built his own scanning electron microscope. In the above video he shows a few adhesives he like including a 3M concoction that is two part and is good for sticking to plastic. They guy has lots of really wonderful videos.

[TheLostVertex]

https://www.youtube.com/watch?v=YBQp04glQqc

Ben Krasnow has a YouTube channel called Applied Science. He built his own scanning electron microscope. In the above video he shows a few adhesives he like including a 3M concoction that is two part and is good for sticking to plastic. They guy has lots of really wonderful videos.

Rik's reply also made me think of one of Ben's videos, but for a different reason. As Rik described thinking about the epoxy as springs that are compressing or tensioning, I was reminded of this video on chemically strengthening glass https://www.youtube.com/watch?v=y02AXdec1sE In it he talks about imperfections in the glass surface and how they do not respond well to tension, and how the strengthening process helps alleviate the problem. One of my favorite youtube channels.

[lothman]

I took a stab at modeling the joint, using what was essentially a finite differences technique. At first glance this seemed simple enough. We just model the adhesive layer as a bunch of little springs, and calculate how much the springs have to stretch or squish as the joint is deformed. Spread the springs uniformly over the cross-section, however oddly shaped it might be, and see what happens.

Rik,

it is very complicated to model adhesive layers interacting with the flanges. But from an engineering education standpoint (where I'm from) an adhesive joint gives highest strenght when loaded on shear stress equally distributet over the whole adhesive area. On your test you did the worst scenario for any adhesive. Bending moment in the butt joint means the highest stress on only a line on the top (I'm lacking the translation of the German word Schälbelastung). So high tension on only the top line, beeing also the shortest due to round shape. If the top line fails the "second" line comes in work seeing an even higher load and fails also... that's why it cracks all at once. And that is the reason why adhesives strenght is not tested in bending moment but with a doubled layer arangement on shear stress. These (high) values of strenght you will find in the catalogue ;-)

So with a tube you could distribute the load to all the adhesive area and also with shear stress. This would lead to much higher strength of the joint because of the load suport to the whole adhesive area.

regards
Lothar


just googled this paper what describes it much better than me:
http://www.henkelna.com/us/content_data ... 404796.pdf