Knowledge / Shipyards and production

Welding distortion in shipbuilding: process, measurement and quality assurance

Understand welding distortion through process qualification, measurement conditions, inspection and traceable correction.

On this page

A welded panel can meet a dimensional target while still containing an unacceptable weld, and a sound weld can belong to an assembly that no longer fits its neighbours. Shipbuilding quality therefore needs separate evidence for the welding process, the joint itself and the geometry of the finished structure. Distortion control connects these questions, but it does not make them interchangeable. This article explains the connection without providing a production welding procedure.

Start with the physical mechanism

Welding heats a small region much more strongly than the surrounding structure. Expansion and subsequent contraction interact with restraint from colder material and fixtures. Permanent strain and residual stress may remain after the assembly returns to a nearly uniform temperature. TWI's explanation of distortion mechanisms distinguishes shrinkage, angular change, bending, buckling and twisting. The observed shape depends on material, geometry, fit-up and process conditions; a single shrinkage allowance cannot represent every joint.

It helps to describe the error before trying to correct it. Is a panel too short, a stiffener rotated, an edge lifted, or the entire block twisted? Record where and when the deviation was measured and whether the assembly was clamped or free. A photograph of a wavy plate does not establish the through-thickness stress field, fatigue strength or cause. Diagnosis needs the production history as well as the final appearance.

Keep three qualification questions distinct

A welding procedure specifies the controlled way a joint is to be produced. A procedure qualification demonstrates performance within the scope of the applicable qualification rules. A welder or operator qualification addresses a person's or operator's permitted scope. Production inspection assesses actual work. None of these records automatically replaces the others. An experienced welder does not make an unsuitable procedure valid, and a valid procedure does not prove every production weld acceptable.

Revision status matters. On the review date, 6 October 2026, IACS's W28 listing identifies Revision 2 as current and says Revision 3 comes into force on 1 January 2027. A project must establish its applicable class rules, implementation conditions and approved procedure set. Publication of a future revision is not evidence that it already governs every contract. This article does not reproduce qualification tables or claim access to paid standards.

Plan geometry and welding together

The useful question is not simply “How can movement be stopped?” Ask which movement is expected, which dimensions matter at the next assembly stage, and what stress or cracking risks a restraint method could introduce. Fixtures also need a release sequence and an extraction path. A part that fits perfectly while held by strongbacks may move when those strongbacks are removed.

TWI's guidance on presetting and restraint explains the trade-off between limiting movement and introducing residual stresses, and the value of representative trials. Such measures should be developed by the responsible welding and structural specialists. Adding weld metal, attaching improvised strongbacks or applying uncontrolled heating is not a harmless workshop adjustment. Each can alter the material or load path and may require an approved corrective procedure.

Heat input is informative but incomplete

For a simplified arc process, electrical energy per unit length can be written as E = VI/v, where voltage V, current I and travel speed v use consistent units. The energy transferred into the joint also depends on process efficiency. The equation does not alone predict peak temperatures, penetration, toughness or distortion. Joint geometry, pass sequence, interpass conditions and heat losses matter, and pulsed processes require appropriate treatment of electrical measurements.

Consider an invented comparison with steady 24 V and 200 A. At 5 mm/s, electrical energy per length is 960 J/mm, or 0.96 kJ/mm. At 4 mm/s it becomes 1200 J/mm, or 1.20 kJ/mm: a 25% increase. No thermal efficiency has been assumed. The calculation does not authorise either setting or predict a 25% distortion increase. It demonstrates why “same current” does not mean “same thermal history”.

An original measurement example

Assume two educational stiffened panels have the same nominal geometry and are measured at the same temperature on the same support arrangement. After welding, panel A shows an edge lift of 8 mm while still restrained and 13 mm after release. Panel B shows 10 mm while restrained and 11 mm after release. The numbers are invented and are not acceptance limits or measured shipyard results.

Comparing only the restrained measurements would favour A. Comparing the released condition would favour B for this single geometric measure. Neither comparison proves that B has a better weld, acceptable residual stresses or adequate service strength. The study also lacks repeat trials and a measurement-uncertainty budget. Its immediate lesson is to define the state being compared. Otherwise, two teams may report different “final” dimensions while measuring different mechanical conditions.

A defensible trial record would identify material batches, joint preparation, procedure revision, consumable, sequence, restraint arrangement, relevant temperatures, measurement datums and release timing. If a variable changes between trials, record it rather than attributing the whole result to one chosen cause. Small trial programmes are especially vulnerable to confusing correlation with a repeatable process improvement.

Inspection should follow the failure question

Visual inspection, dimensional survey and nondestructive examination answer different questions. A geometric survey can locate misalignment but cannot generally establish internal weld quality. A volumetric examination can reveal certain internal indications without proving the complete assembly is dimensionally correct. Inspection methods, extent, timing, personnel qualifications and acceptance criteria must follow the approved inspection plan and applicable rules.

The IACS Recommendation 47 listing identifies a shipbuilding and repair quality reference, Revision 10 Corrigendum 1 of October 2025. The listing establishes the document's identity, not a universal tolerance for every vessel. Do not lift an isolated tolerance from a secondary chart without checking the relevant structure, material, rule adoption and project specification. A numerical tolerance has meaning only within its defined application.

Correction needs a decision trail

When a deviation is found, preserve its location, extent and condition before correction. Identify whether the response is acceptance within existing criteria, engineering assessment, repair or replacement. A corrective action should state the authorised method, the responsible party and the evidence required afterwards. Grinding a visible feature away does not demonstrate that the underlying joint is sound; measuring after forced fit does not establish acceptable unloaded geometry.

Work safety is a separate gate. OSHA's hot-work rule illustrates the need to control the work area and adjacent fire hazards under United States law. Actual work elsewhere follows its own legal and yard requirements. A technically approved weld repair does not by itself establish that the space is safe for hot work, that isolations are correct, or that nearby coating and cleaning activities are compatible.

Common mistakes and the useful output

Avoid treating low visible distortion as proof of low residual stress, increasing weld size “for safety” without design approval, comparing measurements taken in different support states, or using a generic heating recipe across materials. Another frequent error is to improve one local dimension while moving the mismatch into the next block interface. The complete assembly sequence should remain visible in the decision.

The useful output is a traceable connection between the approved process, production records, inspected joint, measured geometry and any authorised correction. This guide intentionally omits production settings, heat-straightening instructions and universal acceptance values. Those require vessel-specific engineering and competent execution. Its central principle is that geometry, weld integrity and work safety are related requirements, each needing its own evidence.

Sources