Welding in Shipbuilding

Welding fuses the parent metals: the joint reaches melting point and the two pieces become one continuous piece of metal, usually with filler of similar composition. Brazing and soldering do not melt the parent metal at all — a lower-melting filler is drawn into the joint by capillary action and bonds to the surfaces. The dividing line between brazing and soldering is 450 °C: above it is brazing, below it is soldering. The practical consequence is that a welded joint can be as strong as the parent metal, while a brazed or soldered joint is limited by the filler.

Weight and watertightness, mostly. A riveted joint needs overlapping plates and a row of holes, so it is heavier and every hole is a stress raiser; a welded butt joint is flush and continuous. Welding is also faster, cheaper, better suited to prefabrication in blocks, and gives a smoother hull. The cost of that continuity is that a crack has an uninterrupted path — a riveted seam arrests a crack at the joint, a welded one does not, which is exactly what the early all-welded ship failures demonstrated and why brittle crack arrest is now a design requirement on large container ships.

The HAZ is the parent metal close to the weld that never melted but was hot enough for its microstructure to change. It is not one thing: nearest the fusion line the grains coarsen, further out the steel is normalised and refined, further still it is only tempered. Coarse-grained HAZ is the problem region — it is typically the lowest-toughness part of the joint and where brittle fracture and hydrogen cracks start. Its width and severity are controlled by heat input and cooling rate, which is why heat input is a qualified variable rather than a matter of preference.

Heat input governs the cooling rate, and cooling rate governs the microstructure. High heat input cools slowly, giving a wide HAZ with coarse grain and lower toughness. Low heat input cools quickly, giving a hard, brittle martensitic HAZ that is vulnerable to hydrogen cracking. The formula is Q = η·V·I/v, where η is the thermal efficiency of the process, and the answer is expressed in kJ/mm or kJ/cm. In practice the parameter a welder can most easily change without noticing is travel speed — and it sits in the denominator.

Preheat slows the cooling rate after welding, which does three things: it avoids a hard martensitic HAZ, it gives dissolved hydrogen time to diffuse out, and it reduces residual stress. The temperature required depends on the carbon equivalent of the steel, the combined thickness at the joint, the hydrogen level of the consumable, and the restraint. Thicker sections need more because they draw heat away in three dimensions. On board and in the yard the practical rule is that preheat is specified on the WPS and checked before the arc is struck, not judged by hand.

Also called cold cracking or delayed cracking, because it can appear hours or days after welding. It needs four things at once, and removing any one prevents it: hydrogen in the weld metal, a susceptible microstructure (hard martensitic HAZ), tensile stress (residual or applied), and a temperature below about 150 °C. The controls map onto the four: low-hydrogen consumables properly baked and stored, preheat and controlled heat input, reduced restraint and good sequence, and post-weld hydrogen release where required. Because it is delayed, final NDT on higher strength steel is deliberately held back — UR W28 ties that delay to the yield strength of the steel.

The leg is the length along the fusion face — what you measure with a gauge on the plate. The throat is the shortest distance from the root to the face of the weld, and it is the dimension that carries the load. For an equal-leg fillet the design throat is the leg divided by √2, about 0.7 of the leg. Drawings and rules may specify either, so the first question on any fillet dimension is which one is meant: an 8 mm throat is a much larger weld than an 8 mm leg.

Dilution is the proportion of the weld metal that came from the melted parent metal rather than from the filler. It matters whenever the two have different compositions — welding a clad plate, joining dissimilar steels, or overlaying a corrosion-resistant layer, where too much dilution pulls the deposit out of specification. Dilution rises with heat input and with processes that penetrate deeply, such as submerged arc, and falls with stringer beads and lower current.

Four, and each has a natural home. SAW welds panel butts and long seams flat in the panel line, where its very high deposition rate pays. FCAW does most positional work on blocks and at erection joints, because it welds out of position with high deposition and a slag that supports the pool. GMAW handles thinner material and robotic fillet welding. SMAW survives for repairs, tack welds, awkward access and anywhere a cable and a rod beat a wire feeder. GTAW is reserved for root runs on pipework and for stainless and aluminium, where quality matters more than speed. One-sided welding with ceramic or flux backing is now standard on panel lines because it removes the need to turn the panel over.

SAW works under a blanket of granular flux, which is held in place by gravity. Turn the joint out of the flat or near-flat position and the flux falls off, taking the shielding, the slag cover and the arc stability with it. In exchange for that restriction it offers deposition rates several times those of manual processes, deep penetration, no arc flash and no visible arc — which is why the panel line is built around it and the blocks are turned to suit rather than the other way around.

Flux cored arc welding uses a tubular wire filled with flux instead of a solid wire. It combines the continuous feed and high deposition rate of a wire process with the slag of a covered electrode, and that slag is what supports the molten pool in vertical and overhead positions. Rutile cored wires are easy to run positionally; basic cored wires give better toughness and lower hydrogen where the steel demands it. Self-shielded variants avoid a gas bottle and tolerate draughts, which matters on an open berth.

Thickness, access, process and cost. Below about 6 mm a square edge with a root gap will do. Above that a single-V opens the joint enough for the arc to reach the root. Beyond roughly 20 mm a double-V halves the weld metal and lets the joint be balanced about the neutral axis to control angular distortion — but it needs access from both sides. A U or double-U preparation costs machining but removes still more weld metal on very thick sections. Every degree of included angle is consumable, arc time and distortion, so the preparation is chosen as the smallest opening the process can reliably fill from the root.

Weld metal shrinks as it cools. In a fillet or a single-V butt the weld sits on one side of the neutral axis of the joint, so the shrinkage pulls that side in and the plate rotates. Controls, in order of preference: balance the weld about the neutral axis with a double-V or welding alternately either side; pre-set the plates so the shrinkage pulls them into line; use the smallest weld that satisfies the rules; sequence with back-step or skip welding; and use restraint or strongbacks — which trade distortion for residual stress, so they are a last resort rather than a first one. Straightening afterwards by line heating is common, expensive, and something a good sequence avoids.

Both spread heat rather than concentrating it. In back-step welding the seam is divided into short lengths, each run in the direction opposite to the overall progression, so shrinkage in one length is partly resisted by the already-cooled metal beside it. Skip welding deposits short runs at intervals along the joint and then fills the gaps. Both reduce cumulative longitudinal shrinkage and angular distortion on long seams, at the cost of more starts and stops — and every start and stop is a potential defect, which is the trade-off.

• Undercut — a groove melted into the parent metal at the toe. Current too high, travel too fast, wrong angle. A fatigue crack starter, so it is limited tightly at every acceptance level.
• Porosity — gas trapped as the pool freezes. Damp or contaminated surfaces, damp consumables, lost gas shielding in a draught, or too long an arc.
• Slag inclusion — slag from a previous run not cleaned out. Poor interrun cleaning, wrong bead sequence, too wide a weave.
• Lack of fusion / incomplete penetration — the arc never melted the side wall or reached the root. Too little heat, wrong angle, too narrow a preparation or too small a root gap. The most dangerous of the group because it is planar and sharp.
• Cracks — hot cracks form on solidification in a deep narrow bead or with sulphur contamination; cold cracks are hydrogen-driven and delayed.
• Overlap and excess reinforcement — cosmetically reassuring and structurally bad, because both create a sharp toe angle where fatigue cracks start.

A step-like tear that opens beneath the surface of a plate, parallel to the rolled surface, when a weld applies high tensile stress through the thickness of that plate — typically at a T or corner joint under restraint. The cause is elongated non-metallic inclusions in the rolling plane, so the plate has much lower ductility through its thickness than along it. Prevention: specify a through-thickness tested plate (Z quality, covered by IACS UR W14), redesign the joint so the load runs in the rolling plane, reduce restraint, butter the surface with a lower-strength weld metal, or reduce the weld volume.

Because fatigue crack growth in steel is almost independent of yield strength. Higher strength buys thinner scantlings for the same static capacity, which means higher working stresses, a more flexible structure and a higher stress range at every detail — while the fatigue resistance of the welded joint has not improved at all. So the same detail that was comfortable in mild steel may not be, and hatch corners, bracket toes and weld terminations need better geometry and finish rather than less.

A pWPS is the preliminary welding procedure specification — the proposed recipe, written before anything is proved. A test weld is made to that recipe, tested, and the results recorded in a PQR, the procedure qualification record: what was actually done and what the tests gave. If the results pass, the recipe is issued as an approved WPS, which is what the welder works to. The essential variables — process, steel grade, thickness range, position, consumable, heat input range, preheat — define how far the qualification stretches. Move outside any of them and the WPS no longer covers the joint.

Under IACS UR W28 the test assembly is first examined non-destructively over its full length — visual, surface crack detection by dye penetrant or magnetic particle, and volumetric examination by radiography or ultrasonics. It is then cut up for destructive tests: two transverse tensile specimens, four bend specimens (two root and two face, or four side bends on thicker material), Charpy V-notch impact specimens at defined notch positions across the weld and HAZ at the specified temperature, one macro examination, and hardness testing for steels of 355 N/mm² yield and above. A longitudinal tensile test is added when the consumable is not separately approved.

Procedure qualification proves the recipe; welder qualification proves the person can follow it. IACS UR W32 sets the qualification scheme for welders of hull structural steels — a test weld in a defined position and material group, examined and tested, giving an approval with its own range of validity for position, thickness, joint type and process. Approvals lapse if the welder stops using the process, so continuity of employment records matter as much as the original certificate. In a yard audit, the two questions are always the same: is there an approved WPS for this joint, and is the person welding it qualified for that WPS?

• Visual — profile, undercut, overlap, spatter, obvious surface cracks. Cheapest and most under-used.
• Magnetic particle (MT) — surface and slightly subsurface flaws in ferromagnetic material. Fast, sensitive to tight cracks.
• Dye penetrant (PT) — surface-breaking flaws only, but works on aluminium and stainless where MT cannot.
• Ultrasonic (UT) — internal flaws, with depth information. Excellent for planar defects such as lack of fusion; needs a skilled operator and good surface access.
• Radiography (RT) — internal volumetric defects such as porosity and slag, with a permanent record; poor at tight planar defects lying parallel to the beam, and it needs radiation control.

IACS UR W33 covers NDT of ship hull steel welds and UR W34 the advanced techniques — phased array and time-of-flight diffraction.

On thick sections and wherever the expected defect is planar. UT detects lack of fusion and cracks — flaws that are dangerous precisely because they are sharp — far more reliably than RT, which can miss a tight crack lying parallel to the beam. UT also needs access from one side only, gives depth information, produces no radiation hazard so work can continue nearby, and is not limited by thickness in the way film radiography is. RT keeps its place where a permanent image is wanted and where the likely defects are volumetric.

In practice, IACS Recommendation No. 47, the Shipbuilding and Repair Quality Standard, which is the reference most yards and owners write into the building contract. It gives standard and limit values for fabrication and welding — alignment, gaps, undercut depth, weld profile, distortion, straightness — for both new construction and repair. It is a recommendation rather than a unified requirement, so its force comes from the contract; the current edition is Rev.10 with a corrigendum issued in October 2025.

A running brittle crack in a welded hull can travel the length of a ship, because welding makes the structure continuous. The risk grew as container ships adopted very thick, high-strength deck plating, where a crack initiating at a hatch corner or a weld could propagate catastrophically. The response was to require steels with a demonstrated crack arrest capability and to design arrest features into the structure, so that a crack that does start is stopped within a plate rather than crossing the ship. IACS UR W31 covers YP47 steels and brittle crack arrest steels.

Heat input above roughly 50 kJ/cm — and above about 35 kJ/cm for quenched and tempered steels of 420 N/mm² yield and above. It matters because very high heat input coarsens the HAZ and can cost toughness in exactly the material chosen for its toughness. IACS completely revised UR W28 to Rev.3 in September 2025, aligning the treatment of high heat input welding across W11, W16 and W28 with ISO 15614-1 and AWS D1.1, adding explicit definitions in kJ/cm, widening the qualification requirements, and tying the timing of final NDT to the yield strength of the steel. The revision applies to new welding procedure qualification tests from 1 January 2027.

Establish what the indication actually is and where — method, size, depth, orientation — and compare it against the acceptance criteria that apply to this joint, because an indication is not automatically a defect. If it is rejectable, stop and ask why before repairing: one isolated pore in a long seam is a different problem from lack of fusion appearing repeatedly on the same shift. Check whether the joint was welded to an approved WPS by a qualified welder, and whether anything drifted — consumable storage, gas, preheat, travel speed. Then repair to an approved repair procedure, re-examine by the same method plus surface crack detection, and record it. Finally close the loop: extend the examination to adjacent joints from the same welder or shift if a systematic cause is suspected.