A welded frame assembly kept coming out 1.5 mm too wide, batch after batch. The fixture was a purpose-built weld jig with locating blocks, clamps, and a base plate. The fabricator checked the jig — the blocks were in the right position, the clamps were tight, the base was flat. They remeasured the parts going in — they were within tolerance. The welds still pulled the frame oversize. The jig was doing its job. The design of the jig wasn’t.
The jig holds the parts, it doesn’t fight the weld
Weld shrinkage is predictable. A fillet weld on a 6 mm plate pulls about 0.5 mm across the joint as it cools. A frame with four corners welded pulls in four directions. If the jig locates the parts at their nominal positions and clamps them there, the weld pulls them inward. The frame comes out undersized. This batch came out oversized because the jig had located the parts with spacers that compensated for shrinkage in the wrong direction.
When the frame shrank, it shrank toward the weld. The fabricator added shims to push the parts out, to compensate for the shrinkage. But the frame also warped (bowed upward in the middle). The shims only fixed one dimension. The width stayed wrong because the warpage ate the compensation.
The fix: design for shrinkage, then verify with a test weld
The correct way to build a weld jig:
- Predict the shrinkage. For a 6 mm fillet, allow 0.5 mm per joint across the direction of the weld. A frame with two longitudinal seams shrinks 1.0 mm in length. Build the jig 1.0 mm oversize in that direction.
- Allow for warpage, not just shrinkage. The frame warps because the welds are on one side. Balance the heat by welding opposite sides alternately, or pre-set the jig with a counter-bend. A jig that holds the part flat through the first weld will let it bow after release.
- Test-weld one part before setting production tolerances. Build the jig, weld one frame, measure it, adjust the jig. This is the step that saves the batch. The fabricator skipped it and ran 100 frames before measuring one.
Where the jig actually failed
Three specific issues on this jig:
The locating blocks were on the outside of the frame members. As the weld cooled, the frame shrank away from the blocks — the blocks only locate the part during assembly, they don’t hold it during cooling. The clamps were released before the weld had fully cooled (standard practice is to keep the part clamped until the weld is below 200°C). Releasing early let the frame spring.
The base plate was 12 mm thick steel, but it was only supported at the four corners on the table. Under the weld heat, the base plate bowed up 0.3 mm in the middle. The jig was no longer flat. The frame followed the bow. The fix was a thicker base (20 mm) with a support bar down the middle, and shimming the base plate level before each run.
The datum principle
A weld jig should use the same datum points as the final machining. If the frame is machined from the bottom face, the jig should locate from the bottom face. If the jig locates from the top face and the frame warps 0.3 mm, the machined datum is gone. The fabricator’s jig located from the top face. The final machining located from the bottom. Every frame carried the 0.3 mm warpage into the machining operation. The machining then had to remove 1.5 mm from one side to get the part flat — the “oversize” complaint was the machining operation struggling with a warped frame.
The fix: machine the frame bottom face first, then use the machined bottom as the datum for the second weld and the final machining. Or better, weld the frame, stress-relieve it, then machine it. The stress relief costs $50 per frame but removes the warpage issue entirely.
A jig locates the parts for welding, but it can’t fight shrinkage or warpage. Compensate for the weld, keep the part clamped until cool, use a stiff base plate, and match the jig datum to the machining datum. The 1.5 mm oversize frame wasn’t a fixture problem — it was a fixture that ignored the physics of the weld.
