Glue strength testing

After doing a lot of testing on the strength of different types of wood joints, I figured it would make some sense to investigate different types of wood glue. I have always just used yellow carpenters glue, on faith that it's good enough. But curiosity got the better of me, and so I launched into testing a whole bunch of different wood glues.

The glues I used in my tests:

LePage Carpenters glue
The glue I usually use. No shelf life issues that I know of. I buy a four liter jug of it every few years. Over time, it gets thick, but stirring in a bit of water fixes it. This glue is not waterproof and can be washed out of clothing.

Titebond 3
I heard that this was supposed to be a really good glue. Waterproof. Couldn't find it in the local stores, but Lee Valley Tools carries it.

Gorilla glue
A polyeurathane based glue. Waterproof. A guy at work swears by it. Many people like it because it expands as it cures.

Weldbond
A water resistant glue, which I had used before for making lawn chairs.
Definitely has shelf life issues - I had a bottle of this glue go completely solid on me one time.

Epoxy
Never used it as a wood glue, but I figured Epoxy has to be strong.

Hot glue
Not really a wood glue, but I figured it might be interesting to see how it stacks up. I used a heat gun to heat my pieces of wood before applying glue, then heated it again before squeezing the parts together to make sure the glue soaked into the wood.

I also experimented a little with the water based varnish I usually use, to see how it would perform as a glue.

I bought small containers of Titebond 3, Gorilla Glue, Weldbond, and LePage epoxy glue specifically for this test. The carpenter's glue was over a year old, as I don't know of shelf life issues with it.

Types of joints I tested with

How a glue performs is probably also a function of what type of joint it's used with. But what joint to use? I wanted the joints to be as simple as possible to cut down on work, and reduce the effect of how I actually made the joints.

Most joints in woodworking either involve two pieces with the grain parallel or at a 90 degree angle. An example of the grain parallel would be gluing up a panel, or stacking pieces to get more thickness. I already knew that the joint is as likely to break beside the joint than at the joint when the grain is parallel, so there was little point in testing with this type of joint.

So joints with the grain running at 90 degrees (call it cross grain) would be a more logical choice. Most mortise and tenon joints are of this type, and in my previous tests the joint always failed mostly at the glue line.

Cross grain joint

This picture shows the geometry of joint I decided on for my tests with spruce lumber. I made one joint with each type of glue and let the glue set for 48 hours. To break them, I then used the same setup as I used for my joint strength tests.

The force of breakage for each of my joints was very close to 140 pounds (64 kg) on my bathroom scale for all the joints except the epoxy glue joint. The broken joint surfaces all looked roughly like the ones at left, with chunks of wood torn out of the post and/or rail. So for the cross grain joints in spruce, the glue was stronger than the wood. I had similar results for the cross grain joints in maple. So I needed to devise other joints, ones where the glue, not the wood, would fail.

Butt joint

One type of joint I was sure would fail at the glue was the butt joint.

To economize on wood, I glued a whole lot of short pieces end-to-end, and then broke off one section at a time with my apparatus. I made a bracket to hold the piece against the post of my test stand, and used another piece of wood to extend the leverage out to 20 cm. Part of my economizing was to produce less scrap, but I also wanted to make sure that all of my test pieces came from the same piece of lumber to reduce variations that might exist between different boards.

I had some consistency problems with the butt joints, but I was getting noticeable variation in maximum breaking force for the different types of glue, so I was pretty certain I was testing the strength of the glue and not the strength of the wood this time. The only catch is that the butt joint isn't used in furniture construction, so perhaps my butt joint test results are not all that relevant.

Scarf joint

I wanted a joint that wasn't a butt joint, but would still fail at the joint. I figured a good compromise would be a scarf joint, so that was the third type of joint I decided to test with.

I cut the lumber at a 20 degree angle and then glued two pieces together. Because the glue really lubricates the joint, it was pretty tricky to clamp the scarf joints without having the pieces slide apart. A staple across the joint helped fix that. I didn't expect the staple to make a difference in the ultimate strength of the joint, but just to be safe I pulled all the staples out of my joints before testing.

The scarf joints required a lot of force to break. Of the hardwood joints, a number of them required more than the 300 pounds to break. My bathroom scale only reads to 300 pounds, so I set up this lever contraption to split the force between my bathroom scale and a fixed pivot. The pivots of my lever are 1/2" steel shafts on the middle and right, and a small strip of wood glued to the bottom of the left end of my lever. I also put a small piece of wood on top of the jack, to act as a pivot there. The idea was to get the whole jack and lever setup to act like pin joined members. That way, I could be sure that the weight was applied exactly to the middle of the lever, and that the weight was split evenly between the scale and the fixed pivot, even if I had some deflection in my setup.

It took a lot of force to break the scarf joints in hardwood, When they failed, they usually failed without warning and with a deafeningly loud bang. I got into the habit of holding the C clamp attached to my specimen while pumping the jack so that at least nothing went flying!

I made several joints with each type of glue. The photo at left shows most of my joints in spruce clamped up after gluing. I also did the same with maple. The maple I used was not very nice looking grey-ish stuff, probably sugar maple. Definitely not the light colored hard maple. I figured in hardness, that it was a good representation of a middle of the road hardwood in terms of hardness. Plus, I wasn't wasting really good lumber in the process.

Results

The most obvious result of all my testing is a pile of mutilated bits of scrap wood with bits of glue on them. I hate producing this much scrap wood. When I build a project, at least some of the wood gets used in the project. Overall, I built and busted over 60 joints. It was a time-consuming affair.

My hardwood joints were made from slightly smaller pieces of material than the spruce joints. I used wood that was 17x51 mm for the hardwood rails, whereas the spruce rails were 18x53 mm. For the cross grain joints, the posts in spruce were 38 mm wide, but only 36 mm for the hardwood joints. Overall, I would guess the difference in geometry makes about a 10% difference in breaking force - keeping in mind that strength in terms of bending moment goes up more than proportionally with joint size.

Joint type Joint
area
(cm) Breaking force (pounds) at 20 cm from joint
as a function of wood, joint type, and glue type
Hot glue Carpenter's
glue Weldbond Gorilla Glue Titebond 3 Epoxy Varnish
as glue
Spruce
cross grain 5.3 x 3.8 145 140 150 150 135 100 50
Spruce
butt joint 5.3 x 1.8 100,45,70 90,65,85 125,85,135 100,70,155 110,90,100 70,30 25
Spruce
scarf joint 5.3 x 5.2 180 185,170 210,240 245,190 210,250 150,140 55
Maple
cross grain 5.1 x 3.5 90,100 210,210,225 200,250 230,230 220,200 epoxy and
varnish not
tested with
maple
Maple
butt joint 5.1 x 1.7 90,90 110,105,100 110,150 90,85,105 150,100,120
Maple
scarf joint 5.1 x 4.9 180,150 290,280 260,260,270 220,200,280 350,350

Joint type	Joint area (cm)	Breaking force (pounds) at 20 cm from joint as a function of wood, joint type, and glue type
Hot glue	Carpenter's glue	Weldbond	Gorilla Glue	Titebond 3	Epoxy	Varnish as glue
Spruce cross grain	5.3 x 3.8	145	140	150	150	135	100	50
Spruce butt joint	5.3 x 1.8	100,45,70	90,65,85	125,85,135	100,70,155	110,90,100	70,30	25
Spruce scarf joint	5.3 x 5.2	180	185,170	210,240	245,190	210,250	150,140	55
Maple cross grain	5.1 x 3.5	90,100	210,210,225	200,250	230,230	220,200	epoxy and varnish not tested with maple
Maple butt joint	5.1 x 1.7	90,90	110,105,100	110,150	90,85,105	150,100,120
Maple scarf joint	5.1 x 4.9	180,150	290,280	260,260,270	220,200,280	350,350

Observations

My epoxy joints in spruce were significantly weaker than the wood glue joints, and even weaker than the hot glue joints. Based on this, I decided to skip the epoxy glue for my tests with hardwood.
I have since found out that 48 hours at 18 degrees may not be enough time for epoxy to reach its full strength, so my results may not be representative of fully cured epoxy glue. I can't just go back and re-do my epoxy tests because I used up the whole board that I used as material for these tests. re-testing with a different piece of wood, even of the same species, would introduce further bias.
Some of the scarf joints in spruce caused bits of wood from one part to stick to the other side after the joint broke, so the glue joints at 20 degrees to the grain were probably close to the strength of the wood.
The cross grain joints all failed at about the same strength, and all resulted in bits of wood from one part sticking to the other. So these joints failed more with the wood than with the glue. I don't consider the cross grain joints a useful measure of glue strength.
I had some problems with consistency with the butt joints and scarf joints. This could have been caused by slightly different texture on the joint surface, fingerprints, or possibly all the glue getting absorbed into the grain. Some of my butt joints failed really early. I excluded these and made more joints. Butt joints can be surprisingly strong, but you shouldn't count on them. I don't consider the butt joints results to be very useful as a means of judging glue strength.
For the better glues, the variation from joint to joint, even with the same glue, is at least as large as the variation between different types of glue.

Conclusions (part 1)

With the cross grain joints failing in the wood, and the butt joints being unrealistic and unreliable, that leaves the scarf joints as the most useful measure. But even my spruce scarf joints had a tendency to sometimes fail in the wood. That leaves the scarf joint in maple as the most useful indicator.

And the strongest glue is...

Titebond 3

For the maple scarf joint, Titebond 3 looks like the clear winner, and that glue also performed quite well in all the other tests.

I'm a little disappointed with Gorilla Glue. Although some joints were quite strong, I didn't get good consistency with it. Gorilla Glue does foam up and expand as it cures, and I suspect the glue got pushed out of my joints by the foaming up. The expansion and foaming up was quite messy. I guess it's a good glue to expand into inaccurate joints, but I prefer to just make my joints so that they are accurate.

I got to liking the Weldbond glue. It didn't perform as well as Titebond 3, but of the waterproof glues, it was the least messy. Also, that glue dries transparent and retains some flexibility. I think that may give it some toughness and reduce the likelihood of the glue joint cracking over time. I didn't like how the Titebond 3 dries to a dark brown. Any time there is a gap that is filled with glue, it really shows. I think Weldbond will be my waterproof glue of choice.

More testing with hard maple

Because I gave all my joints 48 hours to cure before testing them, my experiments stretched over many days. I had already written the above conclusions but then decided to do another cross grain test in hard maple (sometimes also called rock maple). Hard maple is the hardest of the commercially available domestic hardwoods, and I figured a cross grain joint in that would be both demanding and realistic. The cross grain joint is most like typical furniture joinery, and the extra hardness of the maple would put more stress on the glue before the wood would give way.

My joint area was just 4.1 x 4.1 cm. Again, I applied force at 20 cm, causing the joint to be twisted off in a shearing motion. I used fresh planed smooth surfaces for all the joints, and was careful not to get any fingerprints on the joint area.

Breaking force (pounds) at 20 cm from joint
for a 4.1 x 4.1 cm hard maple cross grain joint
Glue type: Carpenter's
glue Weldbond Gorilla Glue Titebond 3
Individual
readings: 130,135,145
150,155,135 175,160,170
195,195 165,170,165
135 165,165,165
175,165
Average: 141 179 159 165

	Breaking force (pounds) at 20 cm from joint for a 4.1 x 4.1 cm hard maple cross grain joint
Glue type:	Carpenter's glue	Weldbond	Gorilla Glue	Titebond 3
Individual readings:	130,135,145 150,155,135	175,160,170 195,195	165,170,165 135	165,165,165 175,165
Average:	141	179	159	165

With the cross grain joint in hard maple, it looks like the Carpenter's glue is out of its league. It was the only glue that didn't cause wood fibers to tear out of the wood, so the carpenter's glue was probably not as strong as the wood in this test. I had some consistency problems with Gorilla glue, and excluded one sample that failed at just 105 pounds. I think the foaming up of the glue may have pushed some of the glue out of the joint. The foamed up part isn't nearly as strong as solid glue.

What's interesting is that Weldbond wins this round by a narrow margin. My guess is that it's at a slight advantage with the very hard wood because it retains a little bit of flexibility after drying. This would make the joint less likely to crack open gradually from the side, giving Weldbond an edge over the other glues that become hard. But overall, other than yellow carpenter's glue, all the glues I tried were tearing small chunks of wood out of the wood, so mostly we are looking at a failure of the wood.

Conclusions (part 2)

I was hoping I could find a clear winner among the glues, but I'm starting to think that the strength of a glue joint is a function of many variables, and the type of glue used is just one of many. One company's glue chemists may not necessarily be smarter than those of another company. Instead, they have optimized for a different set of circumstances than the other resulting in different glues that perform best in different applications. It would be interesting which glue does best with some dense and oily African hardwoods, but that sort of test could get expensive.

I have since taken some macro photographs of the joint surfaces for my hard maple tests.

Does the glue matter?

I found myself questioning this. Considering how difficult it was to devise a test that showed a difference in strength in the glues, I'd say that for the majority of applications, any good quality wood glue is going to be stronger than the wood. As such, which specific wood glue you use probably won't matter that much. I'll continue to use carpenter's glue for most of my work. It dries to a nice light color, so if there's a glue filled gap in my work, it doesn't show much. It keeps forever, and can be washed off clothing, and I always have some around. The other glues all have shelf-life issues, which means I won't be buying a four liter jug to always have some around.

Glue shelf life

Much time has passed since I did these tests. Only a few months after doing these tests, the remainder of the WeldBond glue became unusable. The gorilla glue stayed usable for the better part of a year. I have not noticed the Titebond 3 glue go bad. And I know from experience that the plain non-waterproof yellow glue is still usable after years, and if it thickens up too much, just add water.

Weldbond and low temeratures

A guy wrote to me "something you should know about weldbond", and continued: "Weldbond wood glue fails completely below freezing. Thought you should know. I build bows - found this out one chilly morning just before delivering a birthday longbow to a bud. I switched to epoxy (G2).".

The bow in question was at -1 degrees, and stringing it up made the handle fall off. (not drawn, just strung). I guess the staying flexible aspect of weldbond doesn't deal well with cold.