Weld shrinkage

An ideal free 200 mm bar with constant expansion coefficient 12 × 10⁻⁶ K⁻¹ contracts 1.20 mm under a uniform 500 K temperature drop, equivalent to 0.6% strain magnitude. The model excludes restraint, solidification, phase changes and nonuniform heating. A weld assembly can respond with translation, rotation, residual stress and shape change. Free contraction does not establish a weld allowance.

TWI explains that surrounding cooler metal restrains weld-region contraction during solidification and cooling. Local yielding can leave permanent movement. TWI, Distortion: Types and causes

Contraction, shape change and residual stress are separate outcomes. Holding the assembly flat can hide movement until release.

Directions and resulting shapes

The direction is defined relative to the weld, not necessarily to the overall part's longest dimension. A weld transverse to a beam can therefore shorten the beam through contraction across the weld.

TWI describes longitudinal and transverse contraction together with angular movement caused by nonuniform through-thickness contraction. It also relates bowing to weld shrinkage acting away from a section's neutral axis. These mechanisms explain why one scalar “shrinkage allowance” cannot describe every dimension of a welded assembly. TWI, distortion mechanisms

The FHWA Bridge Welding Reference Manual, FHWA-HIF-19-088, describes transverse shrinkage moving free joint edges together, and angular distortion pulling flexible members toward the weld. Its steel-bridge examples include both groove and fillet welds; they are explanatory examples rather than guaranteed movements for other geometries or materials. FHWA, September 2019 with September 2020 errata, printed pages 119–120

Why free contraction is not the finished movement

A simple thermal-expansion relation describes a uniform, unconstrained body over a range where its expansion coefficient can be treated as constant. A weldment instead contains changing temperature gradients, a locally molten region, elastic/plastic restraint and potentially metallurgical transformations. The heated region's size and the surrounding structure's stiffness change during the sequence.

Explanatory synthesis: permanent movement depends on compatibility between the shrinking region and the rest of the structure. Some contraction appears as displacement, some is resisted through stress, and the balance changes when restraint is removed or more welds are added. Cooling to ambient temperature does not undo plastic deformation accumulated while the material was hot.

Material, joint preparation, weld volume, heat distribution, pass arrangement and structural stiffness all affect this interaction. An empirical allowance from a familiar assembly is evidence about that route; it is not a material constant applicable to every weld of the same length.

Original thermal-contraction comparison

Assume a hypothetical free bar with uniform temperature, length 200 mm at the higher temperature, constant thermal-expansion coefficient 12 × 10⁻⁶ K⁻¹, and a 500 K temperature decrease. Using the small-strain relation:

Contraction magnitude = α × L × |ΔT|

= 12 × 10⁻⁶ × 200 × 500 = 1.20 mm

The corresponding thermal strain magnitude is 0.006, or 0.6%. This original example uses an assumed coefficient; it is not a specified alloy property or a weld shrinkage allowance. It excludes solidification, temperature dependence, phase changes, nonuniform heating and restraint. Applying 1.20 mm to a real 200 mm weld would add assumptions the model does not contain.

As a separate geometric sensitivity example, an assumed rigid flange rotates 0.5° about its root. At a point 100 mm from the rotation axis, the normal displacement is 100 × sin(0.5°) = 0.873 mm. This converts an assumed angle to movement; it does not predict that welding produces the angle. It illustrates how a small local rotation can matter at a distant feature even when linear shrinkage at the joint is modest.

Restraint changes the response

Tacks, clamps, temporary stiffeners and the surrounding assembly resist movement to different degrees. TWI's fabrication discussion explains that tacks help maintain joint gap and resist transverse shrinkage, while inadequately controlled long seams can close as welding proceeds. It also describes back-to-back assembly and possible movement when paired components are separated. TWI, Distortion control: Prevention by fabrication techniques

Restraint is consequently not synonymous with stress elimination. A restrained dimension and a released dimension represent different mechanical states. Joint gap also influences the volume deposited, so a changing gap can alter the subsequent contraction as well as the immediate fit-up.

An assembly's stiffness is not constant throughout fabrication. The first weld can make the structure more rigid for the second. The FHWA manual uses this distinction to explain why nominally balanced double-sided groove welds may require different preparation depths to offset angular distortion. It also shows that double-sided fillet welds can still distort the horizontal member. Symmetry can reduce a particular movement without eliminating every distortion mode. FHWA, printed page 120, figure 79 discussion

Sequence and weld size are design-dependent

SSAB's trailer-design guidance for Strenx hot-rolled high-strength structural steel discusses limiting weld cross-sectional area, symmetric welding, root-gap consistency and sequence as distortion-control measures. Its page explicitly restricts the recommendations to original SSAB products. These principles explain supplier-specific design choices; they do not authorize reducing a drawing's required throat, penetration or weld length. SSAB, Welding distortion: Trailer Design Guideline

Lower deposited volume can reduce one source of contraction, but an undersized or unfused weld is not an acceptable distortion remedy. Likewise, lower heat input is bounded by fusion, metallurgy, hydrogen-control and procedure requirements. Intermittent welds and alternate joining methods change the joint's structural and sealing behavior; they are design alternatives rather than interchangeable shop corrections.

Presetting or prebending introduces an initial shape intended to offset later movement. Its success depends on the repeatability of material, preparation, restraint and sequence. A measured compensation for one assembly does not automatically transfer to a different thickness, fixture or pass arrangement.

Dimensional evidence and prediction limits

Dimensional comparisons depend on the temperature, support and restraint state. A hot weldment, a cooled clamped assembly and a released component can have different dimensions without the measurements contradicting one another. Machining or later heat treatment can further redistribute stress and change shape.

A numerical welding simulation can represent thermal history, deformation and restraint, but its prediction remains conditional on the heat-source model, material behavior, boundary conditions and sequence. A simple contraction coefficient or energy-per-length figure cannot replace those relationships.

Explanatory synthesis: dimensional conformance, weld integrity and residual-stress assessment are separate forms of evidence. Correct final geometry does not prove adequate fusion; sound weld examination does not prove that the assembly meets its dimensional tolerances. Straightening changes the mechanical and possibly thermal history and is not merely an erasure of the original shrinkage.

Source versions

Reviewed 2026-11-02 PST. The downloaded FHWA PDF cover states September 2019 (with errata September 2020), publication FHWA-HIF-19-088; figure copyright dates are not treated as the manual's edition. No separate technical issue date was verified for the cited TWI or SSAB live guidance. The worked calculations and explicitly labeled synthesis are original, bounded explanations, not fabrication allowances or operating instructions.

Related reference

Manufacturing process references