Optimal Joint Method for Making Curved Beams

Woodworkers discuss the best way to fabricate a curved beam, and learn about an interesting historic example of curved laminated timber barn construction. October 19, 2013

(WOODWEB Member):
I have been trying to determine the best method to build curved beams for a ceiling I am working on. I have come to the conclusion that I am not going to attempt to steam bend the 2" stock for this ceiling project. And, the designer would prefer to see 3" thick pieces, which means I have to make (24) 3"x3"x10' beams, curved on a 6' radius.

Rather than steam bending or bent lamination, I am going to miter cut and join several pieces together and then use a bandsaw to rough out the shape, finishing with a router template to get matching curves. The stock material will be dry fir 4x6.

These beams are basically intermediate support pieces between the primary rafters, and serve only to hold up 1x6 t&G paneling with some cellulose insulation on top, no other structural roof loads will be bearing on them. I know a glued butt joint is not particularly strong, but was wondering if it would suffice if I used dowels or biscuits in addition? The miter on each end is about 15 degrees so while it's not completely end grain, it's pretty close.

Do I need to consider more elaborate joinery to hold these pieces together, or will the method described work?

Forum Responses
(Architectural Woodworking Forum)
From contributor B:
My policy is to not to build anything structural. No matter how sound my construction method, I would be opening myself to litigation should anything go wrong. I also feel that a custom structural component should be presented to an engineer for an approval stamp. This is going to require some additional cost.

That said, I would not use your method to make the beams. You would be better suited to make two 1 1/2" thick segmented beams and glue them together face to face. One could be made with three segments, as an example, and the other would be made with four. This would offset the segment joints when the two layers are glued together. This brick laid construction will yield a beam that I often refer to in our shop as "strong as a bridge girder."

From contributor T:

I've had good results using the method described above by contributor B. Also, I use carefully placed pocket screws in the butt joints to speed things along and it makes for nice tight joints.

From contributor K:
I agree with contributor B too. You can make a very accurate template, which can be used to mark the layers for bandsawing about 1/8" outside the line, then use it with a bearing over or under bit to cut a perfect match. Then you can brick-lay as many layers as you want for thickness or length, and end up with a nice finished arc.

From the original questioner:
If I am going to be making 2 glued up segmented layers, then gluing those together, that would basically double the number of joints I'd have to cut and glue, from about 75 to 150 or more depending on how I was able to stagger the segments. At that point I wonder if it would make more sense to forget about all the mitering and end joints, and go with a glued-lamination after all? Clamp and glue 3 pieces of 1x3 fir into the curved form, thus eliminating all the bandsaw and router work. Any tips on springback with this method? Should I worry much about getting that radius with 1" full dimension fir? I had wanted to use dry wood also.

From contributor J:
Certainly bent glue-lams would be best, but lack traditional timber frame appearance (and are expensive to manufacture). A simple pair of butt-jointed band-sawn beams with a bolted horizontal timber span connection would work.

From contributor B:
Before deciding to go with the 1" x 3" built-up bent lamination, try bending a single 1" x 3" x 10' piece of fir. I suspect you are going to find it more difficult than you expect.

Also, you need to watch for straightening at the ends of the bend. It is the nature of straight stock to bend in the middle but try to remain straight at the ends - probably about 24" to 30" in this case. You are going to need a very strong form to overcome this straightening effect on 3 layers of 1" thick material… plus I suspect more pressure than can be pulled with standard bar clamps.

If you are going the laminated route I'd recommend 6 layers of 1/2" thick stock. This will be easier to bend and significantly reduce spring back. Even so you may end up tossing the first one or two beams as you dial in on the spring back equation.

From contributor L:
A 6' radius is going to be really stiff even at 1/2" layers. I'd bet 1" is hopeless. The thicker the parts the more spring back, up to a point. Have you done any of this kind of work before? I know when we've done bent lam, bending one lamination will seem easy, but by the time all the laminations are glued and ready to force to the form, things get much stiffer. I'd do an inner and outer form with the inner form massively stiff and the outer form not very stiff so it will conform to the laminations even if the thicknesses don't quite add up. Do a dry run first! Make notches for your clamps. After the glue has set, remove from form and put some sort of restraint on the parts so they can fully dry and not creep out. If you make too many thin laminations, the shape may pull in tighter as it dries! Somewhere on WOODWEB there is a formula for correct thickness.

From the original questioner:
No, I have not done this kind of work before. I definitely plan to experiment thoroughly to work out the kinks before moving forward. I'm wondering if steam bending 1" thick pieces into the shape, then laminating 3 of them together, would be a possibility? Has anyone ever done a combination of steam bending and lamination like that? I would guess the moisture content would be the biggest problem and I'd have to dry the steam bent pieces down to spec before gluing them up. It would allow me to use bigger sections that way.

From contributor B:
We do what I call steam-lam occasionally. In my case, though, we're ripping a 1" thick x 4" to 5" wide board into 5/8" x 1" strips to steam bend. Once all the strips have been bent, they are then glued back together on an exact radius form.

By steam bending first, you are taking the spring back out of the boards. Even if the steam bent radius is around 50% larger than final radius, you are no longer bending from straight to the curve. As such the final part does not spring back.

Hopefully you have some experience steam bending, as the first few attempts can be frustrating.

From contributor W:
In a Dutch style barn down the road from me they took 4 layers of 1" x 4" rough ash and nailed them together around an arched pattern on the floor of the hay loft. I would guess they used green material.

Looks like they used 16 commons and let them stick through the first layer by an inch or so and then clinched them. These rafters were about 30 feet long and they just staggered the joints using nominal lengths. They have lasted a hundred years and still solid - no sign of coming apart.

From contributor B:
That's very interesting. Did someone back at the time of the construction keep a record that is now on display… or something similar? If the source of that information is available to the public it would be fun and educational to read.

From the original questioner:
In this part of the country farmers still build Quonset buildings framed with 1x4's or 1x6's by driving stakes in the ground to the curve and then clinch nailing and gluing the pine boards around them. Big difference, though, from a 6' radius. They are usually 40 or 50 feet wide and used for machine sheds. Green or air dried lumber steams a lot better than kiln dried.

From contributor W:
No, sorry - that is just my observation after asking the homeowner to let me inspect his barn with the arch rafters. I was awestruck at the simple solution. I have 40 years in the carpentry trade - the method was apparent. I've used the same method laying out and raising gambrel roofs.

From contributor A:
So when I read this I read curved beam as curving in plan versus an arched beam that would curve in elevation. Is this beam curved in plan or elevation?

From contributor B:
I've read this conversation as the beam being constructed while lying flat on the ground in plan view but being installed in elevation view.

From contributor W:
That whole barn was a work of art. Still plum and level as the day it was built. The workmanship throughout was the very best. It was obvious that the carpenters were masters of the trade. I have never seen this technique documented anywhere. That's why I shared it. If I am ever in that area I will stop again and take pictures.

From the original questioner:
To clarify, the beams will be curved in plan! The structure is circular, 18' in radius, with (12) 6x6 fir struts holding up a center compression ring. The pitch of the struts/beams is about 2/12. They are 3' apart at the compression ring, and spread to about 9' apart at the tension ring on top of the walls. These curved pieces are intermediate supports for the 1x6 T&G paneling that will sit on top of them for the ceiling, to reduce the free span of the 1x material. Thus they are not very structural, just holding some 1x pine and cellulose insulation, no roof load.

At this point, having done test beams with the brick method (using 1.5" material), the end joint method, using a lap joint and 3" material, and the glue lam method, using 1/2" lams, I think I am going to use the glue lam method… that which I was so adverse to in the beginning. The time to bandsaw, then glue up, then finish the curves with a router seems like it will be a bit much for 24 of these things. The lam method is more work up front, but much less finishing.

My plan is to get 6x6's and resaw them, then glue them up in the forms, and once dry, rip them in half on the table saw to make my 3x3 beams. That reduces the number of glue ups to 12 instead of 24, and I can face the cut edge to the ceiling if need be.

Going to get the 6x6 stock tomorrow and do some test runs with a male-female clamping pattern to figure out spring back. Any advice on springback, glues to use, or any other considerations is welcome!

From contributor M:
Contributor W, I'm also interested in the barn that you refer to. Can you remember roughly the span of these rafters, or guesstimates of the radius or center height? Many of us would probably like to be able to picture the structure of this barn given your description so far.

To the original questioner: Glad that you chose not to do the butt joints with this. Plan on at least staggering the butt joints in the installation, maybe even scarfing these purlins. Also, with laminations, when you machine the top to the compound curve, they will get all twisty on you if you have a lot of tension in the laminate. Use thinner laminates than you think you can get away with or it'll come back to haunt you.

Who knows, maybe I don't have this shape right, and this makes no sense. I'm picturing a conical ceiling, and you're making circular purlins that sit on top of radial rafters, and with lengths of tapered t&g forming the developed surface; fun!

For what it's worth, I would have steamed this in full thickness if it could have been in white oak. Otherwise, it would be 1/4" laminates.

From contributor W:
This particular barn was sort of small by hay barn standards, with a total width of about 40 feet. The rafters were about 30 feet long, so this hay loft was about 24 feet tall to the ridge. They used the arch design because they could hold more hay and it eliminated the top set of beams in the traditional gambrel design.

Anyway, these 4 x 4 arched rafters were set into mortises in the center of a 8 x 8 top plate. Then they added a 2' overhang with about 7/12 pitch nailed on the side of the arched rafter and the side of the barn.

These barns were built to be full of hay and as city folks buy these properties with old barns, they eventually empty all the hay out and the barn starts to rack and loosen the joints. In my case the people that owned my place before I bought it were burning the barn braces for firewood.

From contributor D:
While not commenting on the joinery method, why don't you get smaller sections of the arch cut on a CNC? They'll all be dead nuts accurate and you could have dowel holes put in the sides to line everything up. Once assembled and glue, all you'd need is a quick sanding rather than cutting on a band saw.