A frame that racks has not sagged. Its sides are still straight and its members are still the length they were cut. What has changed is the angle at the corners, and because there are four of them, the whole rectangle has shifted into a parallelogram. Measure the diagonals and they are no longer equal.
That distinction matters because the fixes are different. Bending is fought with sequence, restraint and pre-set along the member. Racking is fought at the corners, because the corners are where the angle changed, and it is fought with the geometry of the frame rather than with the torch.
Most fabricators meet racking as a surprise. The frame was square when it was tacked, it was square after the first two corners, and it was out by the time the last corner was welded. The movement happened at the end, which is exactly when the frame had the least freedom to resist it.
Four members pinned at the corners form a mechanism. Nothing in the shape resists a change of angle, so the only thing holding it square is the stiffness of the corner joints themselves. A welded corner is stiff, but it is stiff in a way that concentrates all the restraint in a small region of hot metal.
That is the reason a frame can be strong in bending and still rack. Load applied across the frame bends the members, and the members are stiff. Load applied diagonally rotates the corners, and the corners are the weak link.
Add a diagonal and the mechanism disappears, because the triangle cannot change shape without changing the length of a member. This is why frames that must stay square carry a diagonal or a shear panel, and why frames without one rely entirely on corner stiffness.
Most racking starts with a weld on one face only. A corner welded on the outside shrinks on that side and rotates both members toward the weld. Do that at all four corners in the same rotational direction and the frame becomes a parallelogram with no member out of straight.
If the four corners are welded in inconsistent directions, the result is worse, because some corners rotate one way and some the other. The frame then racks and twists at the same time, and the diagonals differ in both length and plane.
The remedy is either to balance the deposit on both faces or to control the direction so all corners rotate the same way and the movement becomes predictable. Predictable movement can be pre-set. Random movement cannot.
A corner is a heat trap. Two members meet, the section is thicker than either member alone, and the weld has to fill a joint that is often the largest deposit on the part. That concentration of heat sits at the point where the angle is most sensitive.
Welding all four corners in one pass around the frame makes the last corner hotter than the first, because the frame has been warmed by the previous three. The last corner therefore shrinks more, and the frame closes toward it.
Cooling between corners, or alternating diagonally opposite corners, spreads that accumulation. It costs time, and it is the cheapest distortion control available on a frame.
A mitred corner has a long weld along the joint line and shrinks along the mitre, which pulls the corner closed. A butt corner with a cope has a shorter weld and behaves differently. A corner with a gusset or a plate has its weld distributed away from the angle itself.
The gusset is the most useful of the three when racking is the concern, because it moves the shrinking metal away from the corner point and adds stiffness exactly where the mechanism was. It also gives the welder somewhere to put deposit that is not on the visible corner.
Where a gusset is not possible, a small internal fillet on the opposite face from the main weld balances much of the rotation. It is often omitted because it is not visible and not structural in the analysis, yet it is the difference between a square frame and a racked one.
Squaring a frame after welding means pulling it back with a diagonal clamp or a jack, and it works because the corners have enough ductility to accept a small permanent change. It also leaves residual stress in the corners, which will redistribute the first time the frame is loaded or heated.
Squaring before welding, and pre-setting the diagonals, produces a frame that finishes square with far less locked-in stress. The difference shows in service: a frame straightened cold tends to move again under vibration, while a frame that finished square tends to stay there.
Either way the check is the same. Diagonals measured corner to corner, both directions, with the frame resting free and cool. A frame measured while clamped reads whatever the clamp wants it to read.
Titanium frames rack more visibly than stainless for the same corner detail, because the material is stiff for its weight and thin sections carry the load with less margin. A thin titanium frame with one-sided corner welds will show a diagonal difference that a heavier stainless frame would absorb.
Titanium also resists correction. Pulling a racked titanium frame back into square cold works against a material with significant spring-back, and each attempt adds cold work at the corner. The pre-set has to be right at the start.
Stainless is more forgiving of a small correction but more prone to heat tint and distortion along the member from the corner heat. The corner sequence that keeps a stainless frame square also keeps its colour even, which is a useful side effect.
Frame geometry is usually specified as an outline dimension, which says nothing about racking. A few extra lines on the drawing remove most of the risk.
Fabricated frames, brackets and plate work appear under titanium fabrication components, and general welded assemblies under other parts. Tube-based structures are listed on exhaust pipe, with welded titanium details on TIG titanium welding parts. Custom drawing and sampling is covered on customization, and the production site on about us.
