What Distortion Is and Why It Is Predictable

Why this matters

Shops treat distortion as bad luck, which is why they pay for it twice: once welding the assembly and again straightening it. It is not luck. The direction a weldment moves is set by geometry and can be written down before anyone strikes an arc. The magnitude is not calculable in a fabrication shop, but it is repeatable, which means your own measurements of the last six identical assemblies predict the seventh better than any formula would. Knowing which half of that is derivable and which half is empirical is the whole skill, and it is the difference between presetting a member correctly and guessing at it.

Before anyone unclamps anything: a restrained weldment holds real elastic energy, and the parts move when the restraint comes off. Support the assembly, stand clear of the direction the free end will travel, and release strongbacks and clamps one at a time rather than cutting several at once. If straightening by flame heating is on the table later, that is fuel gas and combustion under 29 CFR 1910.253, with ventilation and a personal carbon monoxide monitor in any enclosed or partly enclosed area, and the maximum temperature permitted on a quenched and tempered or high-strength steel is owned by the steel producer's data and the engineer of record, not by the operator with the torch. A weldment out of the fixture is also hot metal, so handle it with tongs or wait.

The mechanism, in one paragraph

Heat expands the metal at the joint. The cold metal around it will not move, so the hot metal yields in compression instead of expanding freely. Then it cools from that yielded state, and it contracts as if it had never expanded, so it ends up shorter than it started and pulls everything attached to it. The joint region is left in tension with the surroundings in balancing compression, whether or not the structure is ever loaded. The related card on what a welded joint changes about the parent metal owns the residual stress consequence; what matters here is that the contraction is a fixed volumetric fact and the only question is which way the section lets it act.

The three modes, and their directions

Transverse shrinkage acts across the weld and pulls the two members together. It scales roughly with the deposited cross-sectional area and falls with increasing plate thickness at constant groove volume, because a thicker section has more metal resisting the same pull. The joint preparation card owns groove area, and it is the single largest lever on this mode.

Longitudinal shrinkage acts along the weld. Per inch it is small. Over a long seam it is not, and it is the mode that becomes camber: if the weld sits away from the member's bending neutral axis, the shortening of that fibre bends the whole member. Direction: the welded side becomes the concave side. Lay the member on a flat surface with the welded side up and the ends lift; run a straightedge along the welded surface end to end and the gap opens at midspan.

Angular distortion comes from shrinkage that is not uniform through the thickness. A single-V groove welded from one side puts more metal, and therefore more contraction, at the wide top of the groove than at the narrow root, so the plates rotate toward the weld face and the joint closes upward. On a tee with fillet welds each side of the web, the same non-uniformity through the flange thickness lifts the flange edges toward the welds.

Each direction has a check at the other end. Balance the same groove about the mid-thickness as a double-V and the two rotations oppose, so angular distortion approaches zero for the same total thickness. Put the same long seam on the neutral axis instead of off it and the camber goes away while the transverse shrinkage does not.

Buckling is a fourth thing and it does not behave like the other three. In thin sheet, the longitudinal compression that balances the weld's tension can exceed what the panel can carry as a flat plate, and it snaps out of plane. That is a stability failure, so it appears suddenly below a thickness threshold rather than growing in proportion to heat input, and it does not respond to presetting. Below that threshold the answers are different: less heat, intermittent welding where the designer permits it, and support that carries the panel flat while it cools.

What scales the magnitude

• Deposited volume, first and by a wide margin. Everything that reduces the groove reduces all three modes.
• Number of passes, at constant total groove volume. More and smaller passes generally increase angular distortion in a V groove, because each later pass contracts against metal already frozen beneath it.
• Restraint. Restraint converts movement into residual stress. It does not remove the contraction, it stores it, and stored tension is the third of the three conditions a hydrogen crack needs. The hydrogen card owns that model, and this is the trade at the centre of every fixture decision: a rigidly restrained assembly comes out straight and carries higher stress and higher cracking risk than a free one.
• Preheat, at constant heat input, heats a larger volume and therefore usually increases distortion. That is a real cost of a control taken for a different reason, and it is one the interpass card's floor does not price for you.

One rule, two outcomes

The rule: shrinkage is resolved by where the deposited metal sits relative to the section, and by the restraint present while it cools. Same weld size, same consumable, same energy, two assemblies.

Case A, a butt weld in 1/2 in plate, single-V, 60 degrees included, welded from one side, unrestrained. The deposit sits above the mid-thickness of a section that is symmetric in plan, so the mode available is rotation. Predicted before welding: the plates rotate toward the weld face. Measured after, with a digital level on each plate: 2.5 degrees on one plate and 2 degrees on the other, both closing upward, giving about 4.5 degrees of included rotation across the joint. No measurable camber, because the seam is not offset from a bending axis in any direction that matters on a flat plate.

Case B, an 8 ft stiffener fillet welded along one flange of a beam, unrestrained. The seam is 96 in long and sits well away from the member's bending neutral axis, so the mode available is longitudinal bow. Predicted before welding: the welded side becomes concave and a straightedge on that side gaps at midspan. Measured after, straightedge over the full 96 in: 3/16 in of gap at midspan, opening toward the welded side.

Same shrinkage, opposite consequence, because the geometry resolved it differently. A shop that has only ever built Case A concludes that distortion means angular movement and will not see Case B coming.

Where the magnitude number came from

The 3/16 in was not calculated. It was read out of the shop's own log of six previous identical assemblies, measured the same way with the same straightedge over the same 96 in:

• Measured gaps: 1/8, 3/16, 3/16, 1/4, 3/16, 1/8 in. Sorted, the middle two are both 3/16 in, so the median of the six is 3/16 in and the range is 1/8 to 1/4 in.
• Direction: six of six opened toward the welded side. Not five of six.

That is the honest shape of the prediction. Direction was right six times out of six because it comes from geometry. Magnitude spanned a factor of two across nominally identical assemblies, which is why the shop presets to the median and expects to be within about a sixteenth of an inch rather than exact. Anyone quoting a distortion magnitude to three figures on a fabrication drawing is quoting a number that assembly-to-assembly variation will swamp.

Correction, printed. The 3/16 in median describes assemblies built the way those six were built: continuous welding in one direction, no preset, no restraint. It is not a property of the design, it is a property of the design plus that procedure. Change the sequence and the median moves, so the log has to be reset when the method changes rather than carried forward. The related sequencing card owns the method and the new baseline it produces.

What would change the prediction's direction, not just its size. Weld the Case B stiffener on both sides of the web symmetric about the same axis and the two longitudinal shrinkages act at nearly the same distance from the neutral axis in opposite senses, so the camber largely cancels and what is left is the flange's angular movement. Move the Case A butt weld to a double-V prepared from both sides and the angular rotation cancels while the transverse pull remains. In both cases the mode you have designed out is gone and the one you have not is unchanged, which is the useful way to think about every geometry change.

Check it against the rules the siblings state

• Direction stated with its measurement method, both cases: Case A rotation reported as 2.5 and 2 degrees closing toward the weld face, about 4.5 degrees included; Case B reported as 3/16 in of straightedge gap over 96 in opening toward the welded side.
• Count word verified against the printed series: the log lists six gaps (1/8, 3/16, 3/16, 1/4, 3/16, 1/8 in), the median of those six is 3/16 in, the range is 1/8 to 1/4 in which is a factor of two, and the direction was consistent in six of six.
• Every relationship carries what it is held constant against: transverse shrinkage falls with thickness at constant groove volume; more passes increase angular distortion at constant total groove volume; preheat increases distortion at constant heat input.
• The convention on the groove angle matches the joint preparation card: 60 degrees is stated as the included angle, not the bevel per member.
• Restraint is priced against the hydrogen card's third factor rather than presented as free: restraint stores the contraction as tension and raises cracking risk, and that appears where the fixture decision is made.

References

• AWS D1.1, structural welding code for steel, in the edition invoked by your project specification or the adopted building code, for tolerances on straightness, camber and distortion in welded members and for permitted correction methods
• Steel producer data and the engineer of record for maximum permitted temperature when straightening quenched and tempered or high-strength steel by heating
• 29 CFR 1910.253 for fuel gas used in flame straightening, with ventilation and carbon monoxide monitoring in enclosed areas
• See related: Why Joint Preparation Decides More Than Technique; How to Sequence Welds to Control Distortion; What a Welded Joint Changes About the Parent Metal