Originally Posted by
Steve Baumgartner
I've seen many examples of this kind of tests over the years and found that they are rarely consistent. One says joint A is strongest, another joint B... From reading all these tests, I've come to some skeptical conclusions. People don't want an engineering analysis, they want a single number or an A,B,C ranking. So the tests are oversimplified. People who like a particular joinery technique tend to pounce on any report that favors it and reject ones that favor another.
- The tests rarely include enough samples to have any statistical significance and the reports don't talk about consistency from sample to sample. They only test a single species of wood and kind of glue, but interpret the outcome as applicable to all.
The study referred to above used 5 samples of each type of joint, which is enough to be statistically valid. I don't know whether the test was done, but an "ANOVA" calculation would have predicted whether the variation was due to sample variation or method variation or the percentage of each. In view of the credentials of the researcher, I would guess he did that or something similar. Although I would be interested in such an analysis, I would estimate that 99.95% of all the readers of this study would not. It is understandable why such information would be left out.
It isn't reasonable to claim that if you only test one species of wood, one type of glue, etc. that you can't infer generally applicable information. In this case, the length, cross sectional area and spacing geometry of the connecting pieces, among other things, was what was being measured. Using multiple species of wood and multiple glue types would have dictated that several hundreds or thousands of samples would be needed and little additional information would have been gained. The important thing about this type of study is to define what it is you are measuring and eliminate all other variables.
- The loads at failure are nearly always greater than what you can expect in real life, so the differences are somewhat academic unless you design very spindly stuff. Other factors such as cost of the tooling, precision (consistency), and time spent making each joint may actually matter more than raw strength of a new joint. This is part of the point the OP was making.
Yes!! I often use pocket screws for non critical and non appearance applications.
- The tests nearly always involve loading the joint in one plane parallel to its parts. Most often they make an "L" and squeeze the ends of the legs together or clamp one leg and push the end of the other one parallel to it, lever-fashion. That kind of test may be relevant for the seat frame to back leg of a chair when someone tilts back. But even for a chair, squirming around in the seat is more common and involves multi-directional loads including twisting.
By definition, this particular study set out to measure the ability of various joints to withstand racking. That is the most important property you can measure. Racking resistance in one joint is what prevents the twisting forces you describe in other joints.
- The tests are always of a single, isolated joint. Real furniture is an assembly in which multiple joints often reinforce each other. For example, a rectangular frame with a joint at each corner is much stronger than you would conclude by testing one corner without the others. The stretchers between chair legs are added for a reason!
Refer to previous comment.
- The greatest strength factor revealed by these tests seems to be the area of glued surface. With most joint systems you can increase the area by using more tenons, larger tenons, etc. But the tests assume that you are just using a particular system "out of the box" without analyzing the loads and adjusting the joinery. Comparing a 6mm domino with a 1" wide beadlock tilts the playing field!
No. The greatest strength factor in these tests is the cross sectional area, the length and the spacing geometry of the connecting material. Surface area is a measure of glue strength and except in 2 or 3 predictable cases, (a ridiculous but joint being one) the glue didn't fail. The "playing field" as you call it, is defined by the joint the woodworker chooses. That is precisely what we want to know. In this particular study, the Beadlock joint was stronger than the Domino joint because the Domino limited the size and shape of the connecting material. That is valuable information.
- A large fraction of the joints fail by the wood breaking adjacent to the joint. The amount of "meat" left in critical stress areas is therefore nearly as important as glue area.
In nearly all cases, wood breakage near the joint is the root cause of the failure. The glue joint didn't fail and isn't relevant. At the expense of repeating myself, the greatest strength factor in these tests is the cross sectional area, the length and the spacing geometry of the connecting material. Obviously, that is related to the joinery method used. That is, in fact, the exact information we are looking for in this type of study.
And after all that, my bottom line is that the tests don't tell you much about the way that real furniture joints fail. After years of aging, seasonal cycling of expansion and contraction, compression due to loads, etc., the joints become loose. Then they either just fall apart or break because the looseness puts all the stress in a small area.
In the short term, studies like the one Jim mentioned will tell you a lot about whether a particular joint is adequate for a particular application and are, therefore, very useful in my opinion. In the long run (decades or centuries), the glue material becomes more and more important. There are tests which will artificially age assemblies to test for long term material failure. They are typically outside the scope of what a typical woodworking magazine can do.