Controlling distortion on large weldments

A four-metre frame that arrives bowed by 6 mm is not a welding accident. It is the predictable sum of every bead laid into it, in the order they were laid, under whatever restraint happened to be present. Distortion is deterministic. It is just rarely planned.

The frustrating part is that the machining allowance you left — usually 3 mm — will not cover 6 mm of bow, so the part is either scrapped, straightened, or machined into something with uneven wall thickness that will move again later.

Why the metal moves

Every weld bead is a small volume of metal that was molten and is now shrinking against cold parent metal that refuses to shrink with it. Three shrinkage directions come out of that, and they cause different problems:

• Transverse shrinkage pulls the two plates toward each other across the joint. It closes gaps and shortens overall width.
• Longitudinal shrinkage pulls along the weld axis. On a long seam it bows the whole member toward the weld.
• Angular distortion rotates the plates about the joint, because a V-groove or a fillet deposits more metal near the surface than near the root. This is what turns a flat plate into a shallow roof.

All three scale with heat input and with the number of passes. The practical implication runs through everything below: less metal deposited, deposited more symmetrically, means less movement.

Lever one: joint design

The cheapest distortion control happens in CAD, before any shop sees the part.

Do not oversize the weld. A fillet sized 10 mm where 6 mm carries the load deposits nearly three times the metal, because deposited volume goes with the square of leg length. Oversized fillets are the most common and most expensive source of avoidable distortion on machine frames, and they are usually specified out of caution rather than calculation.

Use intermittent welds where the joint allows it. Continuous welds on rib-to-plate joints are frequently unnecessary. A staggered intermittent pattern can carry the same load with a fraction of the heat. Where a continuous weld is genuinely required — sealed sections, fatigue-loaded joints — say so explicitly so it is not value-engineered away.

Prefer double-sided symmetric joints. A double-V groove welded alternately from both sides largely cancels its own angular distortion. A single-V groove cannot.

Place welds near the neutral axis where the structure allows. A weld far from the neutral axis has a long lever arm and bows the member much more for the same shrinkage.

Lever two: sequence

Sequence is the highest-leverage variable the shop controls, and it costs nothing.

Weld from the centre outward, alternating sides. Welding one side of a symmetric structure to completion before starting the other guarantees a bow. Alternating keeps the shrinkage balanced as it accumulates.

Backstep. Deposit short beads travelling opposite to the overall direction of progress. Each short bead’s longitudinal shrinkage is resisted by metal that is already cold, and the cumulative bow is much smaller than a single continuous pass produces.

Skip around. On a large frame with many joints, distributing the welding across distant locations lets each area cool before its neighbour is heated. It is slower in wall-clock terms and far cheaper than straightening.

Balance about the neutral axis. If the structure has welds above and below the centroid, alternate between them rather than completing one region first.

A written weld sequence — a numbered map of which joint is welded when — is a normal deliverable from a competent heavy-fab shop. Asking for it at quote stage tells you a great deal about whether the shop has planned your part or simply priced its weight.

Lever three: restraint and fixturing

Restraint does not eliminate shrinkage. It redirects it: a restrained joint shrinks less dimensionally and retains more residual stress instead. That trade is usually worth making on a machine frame, because the residual stress is going to be removed by stress relief anyway, whereas the distortion would be permanent.

Fixtures hold the assembly in its intended geometry while it is welded and cooled. For a frame that will be built more than a handful of times, a proper fixture pays for itself quickly — it removes fit-up variation between units as well as controlling movement.

Strongbacks are temporary members welded across a joint to resist rotation, removed after cooling. Cheap, effective, and routine on long seams.

Pre-setting — deliberately assembling the part with an offset opposite to the expected distortion, so the shrinkage pulls it into position — works well on joints whose behaviour the shop has seen before. It works badly on new geometry, because it depends on predicting the movement accurately.

One caution: heavy restraint on a high-carbon-equivalent steel in a thick section raises cracking risk. That is a welding-procedure question, and it is one reason grade and thickness choices matter; see choosing a steel grade for a machine frame.

What preheat does and does not do

Preheat reduces cooling rate, which reduces hardness in the heat-affected zone and reduces hydrogen cracking risk. It is a cracking control.

Preheat does not reduce distortion. If anything, a larger preheated volume expands more and can move slightly more on cooling. Shops occasionally offer preheat as a distortion measure; it is the wrong tool, and hearing it offered that way is a signal worth noticing.

Measuring, not hoping

Distortion control that is not measured is a belief system. Three measurements are enough on most frames:

1. After fit-up, before welding. Establishes the starting geometry, so any later argument has a baseline.
2. After welding, before stress relief. Shows what the sequence actually did and whether the machining allowance survived.
3. After stress relief, before machining. Shows what relaxation added, and confirms the part is worth putting on an expensive machine.

Each measurement should reference the drawing’s machined datums rather than whatever surface was convenient. A report measured from features the shop chose cannot be reconciled against your drawing without more work than the inspection saved — the datum question is covered in GD&T and datum strategy for weldments.

What to specify

On the drawing or in the PO:

Weld sequence to be submitted for review before fabrication. Symmetric, balanced sequence required on all longitudinal seams. Fillet sizes as shown — do not increase. Dimensional check after fit-up, after welding, and after stress relief, measured from datums A, B and C, reports supplied. Machining allowance 3 mm minimum to remain on all machined surfaces after stress relief.

That paragraph converts distortion from something that happens to you into something the shop has committed to manage. It costs nothing at quote stage and it is very hard to add after a bowed frame arrives. More ribs mean more welds: see frame mass and price for the arithmetic.

If you have a frame design and want the weld sizing and sequence risk reviewed before it goes out to quote, send the drawing through — the review comes back within 24 business hours. The related question of what happens to the locked-in stress afterwards is covered in stress relieving welded machine bases.

See also: distortion control reference, types and controls compared.