You cut every tube to the number on the sheet. Every notch fitted. You welded it up and the frame came out 4 mm narrower across the back than the model says.
Nothing was wrong with your cut list. Welding shrinks things, and if you don't plan for it, it shrinks them unevenly.
Why a weld pulls
Molten steel at the puddle has expanded. As it solidifies and cools it contracts, and it contracts against material that's already solid and isn't going anywhere. Something has to give, so the surrounding structure moves.
Three movements, and they combine:
Transverse shrinkage pulls the two parts toward each other across the joint. On a notched tube landing on a rail, that's the member getting shorter than you cut it.
Longitudinal shrinkage runs along the weld. On a long seam it bows the part toward the weld side.
Angular distortion is the one that gets you on a frame. Weld one side of a joint and the member rotates toward that side, because the shrinkage is not symmetric about the tube axis. Two or three degrees at a joint becomes centimetres at the far end of a long member.
That last one is why a frame can be dimensionally perfect at every joint and still be out of square overall. The errors are small and they all point the same direction.
Tack like you mean it
Tacking is where distortion gets decided. Everything after is damage limitation.
Tack the whole frame before you run a single finish bead. All of it. The temptation to fully weld a joint while you're in a comfortable position is strong and it's a trap — once a joint is fully welded it isn't moving, and anything that needs to shift to accommodate the next joint can't.
On an individual joint, tack opposite sides in sequence rather than going round. Four tacks on a tube end: twelve o'clock, then six, then three, then nine. Going round the clock in order is the version that walks the member. Opposing tacks pull against each other and roughly cancel; sequential tacks walk the member round progressively.
Make the tacks small and make them properly fused. A fat cold tack is the worst of both — it's rigid enough to lock in error and weak enough to crack when the surrounding metal moves.
Check before you commit
With everything tacked, measure the frame against the model. Diagonals, overall width at three stations, height, and the position of anything that has to bolt to something else later.
This is the last moment anything is cheap. A tacked frame can be persuaded with a clamp and a slotted screwdriver. A welded one needs a cut-off wheel.
Measure diagonals rather than just dimensions. A rectangle with correct side lengths can still be a parallelogram, and that's exactly the error angular distortion produces.
Finish welding, in the least convenient order
The general principle: distribute heat, never concentrate it, and alternate across the structure so each weld's pull is opposed by the next.
In practice that means:
- Skip around. Weld a joint on the left front, then the right rear, then the left rear. Never finish a region before moving on.
- Back-step on anything long. Run the bead in short lengths, each one travelling backwards toward the previous, so longitudinal shrinkage doesn't accumulate in one direction.
- Weld opposing sides of each joint in sequence, the same logic as tacking. Half a bead at six o'clock, then half at twelve.
- Let it cool. If the tube is too hot to rest a hand near, you're putting heat into a structure that's already fighting itself. Go and make coffee. Stacking passes on a hot joint buys you distortion and a coarse heat-affected zone at the same time.
It feels wildly inefficient. You will spend the afternoon walking round the frame instead of settling into a rhythm. That walking is the job.
Pre-set, when you know which way it goes
Once you've built two of something, you'll know which direction it pulls and roughly how far. At which point you can build the next one deliberately wrong by that amount and let the welding bring it true.
This is standard practice in production fabrication and it feels like cheating the first time you do it deliberately. What makes it legitimate is that the effect repeats, and what makes it fragile is the same thing: it only repeats while the material, the joint prep, the sequence and the welder all stay the same. Change the welder and your correction factor is a guess again. Change the heat of steel and it's a guess you've slightly biased. This is why the number is worth writing down along with the conditions it came from, rather than just the number.
Write it down when you find it. A note saying "rear track pulls in 3 mm, set 3 wide" is worth more than most of the tooling in the shop.
What a jig does and doesn't do
A fixture holds things where you put them while you tack. That's real value, and it's most of what you need.
What it doesn't do is prevent distortion. It restrains the structure during welding, and restrained shrinkage doesn't vanish — it goes into residual stress instead of movement. Unclamp a fully welded frame from a rigid jig and some of that stress releases as spring. The frame moves on the bench, after you thought you were finished.
Which is an argument for tacking in the jig, checking, and doing your finish welding with the frame supported but not fully restrained. Let it move a little as you go, and keep measuring, rather than clamping it rigid and hoping the stress stays put.
Comments
Loading comments…