Structural Beams: Undersized Fillets and a Delayed HAZ Crack at 4 °C
A fabricator is producing twelve 8 m roof beams from 254 × 146 × 37 UB in S355J2, welded to IS EN 1090-2 execution class EXC2. Stiffeners and end plates are attached with 8 mm leg-length fillet welds using 1.2 mm solid wire MIG with an 80% argon / 20% CO2 mix. It is February and the workshop is at 4 °C; the plates were stacked outside overnight and were brought in the same morning, and condensation is visible on the flange faces. Twenty-four hours after welding, magnetic particle inspection finds a 40 mm long indication in the parent metal heat-affected zone of one beam, about 8 mm from a weld toe. A weld gauge check across the batch shows fillet leg lengths of 6.5–7.0 mm against the 8 mm specified. The client is threatening liquidated damages of €500 per day and wants the beams on site this week.
1. The site foreman argues that a 6.5 mm leg is close enough to 8 mm. Assess that argument quantitatively and state what you would do with the affected welds.
Read the worked answer
Convert leg to throat. The design strength of a fillet weld is carried by the throat, not the leg.
- Specified: a = 8 × cos 45° = 5.66 mm.
- As welded: a = 6.5 × 0.707 = 4.60 mm.
- Shortfall: 1.06 mm, about 19% loss of throat, which is roughly a 19% loss of the weld's design resistance.
Is it within tolerance? Under the ISO 5817 quality levels normally invoked for EXC2, undersized fillet throats are permitted only to a limited degree (of the order of 1 mm for throat sizes above 3 mm at level C, and not at all at level B). A 1.06 mm throat shortfall sits at or beyond that limit, so it cannot be signed off by the fabricator. Critically, it is not the fabricator's call to accept a reduction in a designed weld: the weld size comes from the designer's connection check, and any reduction must be referred back to the designer in writing.
Action. Default position: build the welds up to the specified 8 mm leg with a qualified WPS, having first cleaned the toes, and control heat input so the repair does not introduce cracking. Where a beam is heavily stressed or the client's engineer will not accept a build-up, the alternative is to obtain a documented design check showing the reduced throat is adequate. Do not ship on the assumption that "close enough" is acceptable - the paper trail is what the client's inspector will ask for.
2. Diagnose the 40 mm indication: what type of crack is it likely to be, why did it appear 24 hours later, and what in the workshop conditions contributed?
Read the worked answer
Diagnosis. An indication in the parent metal HAZ, parallel to and a few millimetres back from the fusion line, appearing 24 hours after welding rather than immediately, is the classic signature of hydrogen-assisted cold cracking (delayed cracking). It is not solidification cracking, which would sit along the centreline and appear while welding; it is not lamellar tearing, which would run parallel to the flange surface in the through-thickness direction.
Why delayed. Diffusible hydrogen trapped in the weld and HAZ needs time to migrate to the region of high triaxial stress near the toe. Once enough hydrogen accumulates where the hardened HAZ microstructure and residual stress coincide, a crack initiates and grows. That is why the indication only shows up on inspection the next day - a same-day inspection would have passed it.
Contributing conditions, in order of significance.
- No preheat at 4 °C ambient. S355J2 with a carbon equivalent around 0.42 on a 10.9 mm flange cools fast; the HAZ transforms to hard martensite and the hydrogen has nowhere to escape.
- Moisture. Plates stacked outdoors, brought in and welded with condensation on the faces, plus any damp wire spool or gas line, are a direct hydrogen source.
- Hydrogen from the process and consumables - a CO2-bearing MIG mix is less forgiving than a basic flux system, and a damp wire or contaminated nozzle adds hydrogen.
- Restraint. Stiffener and end-plate details create high triaxial restraint exactly where the crack formed.
Verification. Run a hardness survey (HV10) across the HAZ; elevated readings support the diagnosis and are also a useful check for the client.
3. Set out the technical controls you would put in place before any further welding on this batch, and the repair and inspection regime for the cracked beam.
Read the worked answer
Controls before restarting.
- Preheat. Determine the required preheat using the IS EN 1011-2 method for the actual CEV, thickness and heat input; for this combination expect around 100–150 °C, applied by electric blankets or thermal lances with temperature crayons or a contact thermometer to verify through the thickness.
- Condition the material. Bring plates into the warm shop at least 24 hours before welding and confirm the surface is at least 3 °C above dew point before an arc is struck. Weld refusal at low ambient is a legitimate position, not an excuse.
- Consumables. Store wire in a dry, heated cabinet, discard spools showing corrosion, purge gas lines of condensate, and keep the gas hose out of puddles.
- Procedure. Use a qualified WPS with a controlled heat input range and an interpass limit; do not let welders "turn it up to go faster".
- Qualification. Confirm all welders are qualified to IS EN ISO 9606-1 for the process, position and thickness range.
Repair of the cracked beam.
- Determine crack extent by MPI and, if needed, ultrasonic testing - hydrogen cracks often run further than the surface indication suggests.
- Grind or air-arc gouge out the full length plus 50 mm each end, back to sound metal, and re-check by MPI before welding.
- Preheat to the procedure value, weld the repair to a qualified repair WPS, and control heat input.
- Inspect the repair after a 48 hour delay, not immediately, so any further delayed cracking is caught.
- Record the repair, and re-inspect the other eleven beams after the same delay, since they were welded under identical conditions.