How to Sequence Welds to Control Distortion

Why this matters

Sequencing is the cheapest distortion control a shop owns, and it is free only if the plan exists before the first tack. Tacks are the first restraint you install, so by the time an assembly is tacked, most of the sequencing decisions have already been made by whoever was holding the grinder. The related distortion card owns why a weldment moves and in which direction; this one is about choosing when each shrinkage acts and against what, which is a plan you write on the traveller rather than a technique you learn at the arc.

Two hazards in this procedure are created by the procedure itself. Releasing restraint is one: a tacked or strongbacked weldment holds real elastic energy, so support the member, stand clear of the direction the free end will travel, and cut or release one restraint at a time rather than several together. Maintenance heating between passes with a fuel-gas torch is the other: that is combustion in the work area under 29 CFR 1910.253, needing ventilation and a personal carbon monoxide monitor in any enclosed or partly enclosed space, and the member is then hot metal that needs tongs or time. Ordinary arc controls apply throughout: filter shade for the process and current with screens per 29 CFR 1910.133 and 29 CFR 1910.252(b)(2), and local exhaust at the arc plus respiratory protection under a written program meeting 29 CFR 1910.134 for a fume whose constituents follow the base metal and any coating.

The three levers

Everything below is a combination of exactly three things.

• Order: which weld goes first, and what is already stiff when the next one shrinks.
• Direction: which way each bead travels, and whether a long weld is deposited as one bead or as increments.
• Balance: which side of the section the deposit sits on, and how soon the opposing deposit follows.

There is no fourth lever available at the arc. Everything else (groove volume, preheat, restraint level) was decided at the prep and the fixture.

Building the plan, in the order the decisions close

Each decision below has to be made before the one after it, because it constrains it.

1. Restraint strategy. Decide whether the assembly is fixtured, strongbacked or free. Restraint buys straightness and pays in residual stress, and stored tension is the third of the three conditions a hydrogen crack needs, which the hydrogen card owns. On a hardenable steel with a highly restrained joint, the honest answer is often less restraint plus preset, not more restraint. This decision comes first because a rigid fixture makes presetting pointless.

2. Preset or prebend. Only available on a free or lightly restrained assembly. Set the member off in the direction opposite the movement you expect, by the magnitude your own log gives. Direction comes from geometry; magnitude comes from your measurements of previous identical assemblies, and the distortion card owns both halves of that split.

3. Tack plan. Tacks lock in whatever geometry exists when they are made, so their number, size, spacing and location are part of the sequence. Tacks are welds: a short tack on hardenable steel cools fast in three dimensions and is a classic hydrogen crack site, so it gets the same preheat minimum, the same consumable control and the same interpass discipline as the production weld. Grind the ends of tacks that the root pass will run through to a taper so the root can consume them, and where the coating is being ground off, that release is an inhalation route needing tool-mounted capture or respiratory protection rather than eye protection alone.

4. Order of welds across the section. Alternate sides of the neutral axis so opposing shrinkages act close together in time. Two welders on opposite sides simultaneously is the strongest version. One welder alternating is the common version and costs elapsed time.

5. Direction within each weld. For a long seam, backstep. The overall progression runs one way while each increment is welded in the opposite direction, so each increment's longitudinal shrinkage acts on metal that has already shrunk rather than adding to an unbroken run.

Backstep on a 96 in seam, six 16 in increments.
Overall progression runs left to right; every bead
travels right to left, back toward the start of
the bead before it.

 in  0    16   32   48   64   80   96
     |    |    |    |    |    |    |
     [<1 ][<2 ][<3 ][<4 ][<5 ][<6 ]

Bead 1 occupies 0 to 16 in and is welded from the
16 in mark toward 0. Bead 2 occupies 16 to 32 in
and is welded from 32 back toward 16.

6. Interpass management during the waits. Alternating sides means one side sits while the other is welded, and that wait is where the sequence collides with the temperature window. On a job with a preheat and interpass minimum, the sitting side can fall below the floor before the welder returns, so maintenance heat and a contact pyrometer reading before striking belong in the plan rather than in the welder's judgement. On austenitic stainless the interaction runs the other way and the wait is doing useful work against the interpass maximum. The interpass card owns both directions.

7. Release and measure. Release restraint in the order and manner decided in step 1, then measure the same way every time, on the same datum, and write it in the log. A sequencing plan with no measurement record cannot improve, because there is nothing to compare the next assembly against.

Worked example: a fabricated tee, five built the old way

A shop builds a 96 in fabricated tee: a web fillet welded to a flange with one continuous fillet each side, plus four stiffener plates. The first five were welded continuously, one side then the other, each bead run start to finish in the same direction, no preset, no fixture.

Baseline from the log. Straightedge on the flange face over the full 96 in, gap measured at midspan: 1/4, 5/16, 1/4, 3/8, 1/4 in. Sorted, the middle value of the five is 1/4 in, and the range is 1/4 to 3/8 in. Direction: five of five cambered with the welded side concave, which is what the distortion card's rule predicts for a seam offset from the section's neutral axis. All five went to the straightening bay.

The plan written for the sixth.

• Restraint: free, no fixture, because the steel is hardenable and the joints are already restrained by the section itself.
• Preset: prebend the flange 1/4 in in the opposite direction, the median from the five.
• Tacks: 12 in centres, which on a 96 in seam is nine tacks including one at each end, each 1 in long, ends ground to a taper, each preheated to the WPS minimum before striking.
• Order: alternate sides after each increment rather than after each full-length weld.
• Direction: backstep, six 16 in increments per side, as drawn above.
• Interpass: contact pyrometer at the joint before every increment, maintenance heat applied whenever the reading falls under the WPS minimum.
• Stiffeners: welded last and alternating diagonally across the member, so no two adjacent stiffener welds shrink consecutively.

Measured results, four assemblies with the sequence and the preset. Gaps of 1/16, 1/32, 1/16 and 1/16 in, so a median of 1/16 in against the baseline median of 1/4 in. That is 4/16 in down to 1/16 in, a reduction of three quarters.

Isolating the two changes. Two things moved at once, so the shop ran the next four with the sequence and no preset: 1/8, 1/8, 5/32 and 1/8 in, median 1/8 in.

• Baseline 4/16 in to sequence-only 2/16 in: the sequence accounts for 2/16 in of improvement.
• Sequence-only 2/16 in to sequence-plus-preset 1/16 in: the preset accounts for a further 1/16 in.
• Total improvement 3/16 in, of which the sequence is two thirds and the preset one third.

Stated the other way, the comparator is honest in both directions: the 1/4 in baseline is uncorrected, meaning it was measured on assemblies that had neither control, and the two controls are separated by measurement rather than by assumption.

The time trade, in one currency. Sequencing added about 0.5 shop labour hours per assembly in waiting, repositioning and maintenance heat. Straightening the earlier assemblies averaged about 1.5 shop labour hours each. Both figures are shop labour hours on the same crew, so the comparison is like for like: roughly 3 hours of straightening avoided for every 1 hour of sequencing added, on this assembly.

What the trade does not include. Straightening is a heat treatment nobody specified. The maximum temperature permitted during flame straightening on this hardenable steel is owned by the steel producer's data and the engineer of record, and a bay working by colour rather than by a temperature-indicating crayon has no evidence it stayed under it, so the 3 to 1 hours comparison prices only the labour.

What would change the plan. If the flange were thin enough to buckle rather than camber, presetting is the wrong tool: buckling is a stability failure that arrives suddenly rather than in proportion, and the answers become less heat, intermittent welds where the designer permits them, and support holding the panel flat while it cools. If only one welder is qualified for the positions involved, the balanced order stretches out and step 6 becomes the binding constraint rather than a detail.

Keeping the log honest

Measure on the same datum every time: a straightedge on the flange face at midspan over the full 96 in is a definition, and a log mixing that with a reading taken 12 in from the end is not a log. Reset the baseline whenever the method changes, because the 1/4 in median describes the old method and stops being a valid comparator the moment the sequence moves. And watch the tacks after the fact: tack cracks are small, they turn up at the root of the first pass, and they are the cheapest evidence that the preheat discipline is being skipped on the part of the job nobody thinks of as welding.

Check it against the rules the siblings state

• Every improvement claim prints both figures: baseline median 1/4 in from five assemblies (1/4, 5/16, 1/4, 3/8, 1/4), sequence-only median 1/8 in from four, sequence-plus-preset median 1/16 in from four.
• Attribution arithmetic closes: 4/16 to 2/16 is 2/16 from the sequence, 2/16 to 1/16 is 1/16 from the preset, totalling the 3/16 improvement claimed, two thirds and one third.
• Comparator declared: the 1/4 in baseline is stated as uncorrected, measured on assemblies carrying neither control, and the two controls were separated by running four assemblies without the preset rather than by assumption.
• Ratio uses one currency, as the ratio rule requires: 0.5 against 1.5 shop labour hours, both on the same crew, stated as hours avoided per hour spent rather than as a return.
• Interpass rule from the sibling card appears as an action in the plan, not a mention: a contact pyrometer reading at the joint before every increment, with maintenance heat whenever the reading is under the WPS minimum.
• Count checked against the diagram and the list: six 16 in increments span 96 in, and nine tacks at 12 in centres including one at each end is what 96 divided by 12 plus one gives.

References

• AWS D1.1, structural welding code for steel, in the edition invoked by your project specification or the adopted building code, for permitted distortion correction, straightness tolerances and restrictions on heat straightening
• Steel producer data and the engineer of record for the maximum temperature permitted when straightening quenched and tempered or high-strength steel by heating
• 29 CFR 1910.253 for fuel gas used in preheating and straightening, 29 CFR 1910.134 for respiratory protection, 29 CFR 1910.133 with 1910.252(b)(2) for arc radiation
• See related: What Distortion Is and Why It Is Predictable; What Interpass Temperature Is Protecting; What a Welder Qualification Covers and What It Does Not