Lack of fusion is the defect inspectors fear most: the weld metal touched the joint face or the previous pass but never actually fused with it. The result is a planar discontinuity with crack-like behaviour — invisible from the surface, hard to find even with NDT, and structurally unforgiving.
It’s also the defect that most clearly separates “the weld looks fine” from “the weld is fine”: a cold-lapped bead can have a perfect surface profile over a joint that’s mechanically a butt of two unbonded surfaces.
What is lack of fusion?
ISO 6520-1 group 401 covers three locations:
| Type | Code | Where it hides |
|---|---|---|
| Lack of side-wall fusion | 4011 | Between weld metal and groove face |
| Lack of inter-run fusion | 4012 | Between passes in multi-pass welds |
| Lack of root fusion | 4013 | Between weld metal and root face |
The physics: fusion requires the parent surface itself to reach melting temperature while the filler arrives. If molten filler flows onto a surface that’s too cold — because the arc was aimed at the pool, the heat input was too low, or the metal rolled ahead of the arc — it freezes against it like solder on an unfluxed joint: in contact, not bonded.
Acceptance: zero, at every level
Under ISO 5817, lack of fusion (401) is not permitted at quality level B, C or D. The same applies to incomplete root fusion (4013).
There is no dimensional discussion as with porosity or undercut: planar defects with sharp edges propagate under load, so the standard treats any detected lack of fusion as a repair. Verify your governing code, but expect no relief — fitness-for-service assessments (e.g. fracture-mechanics-based) are engineering exceptions, not shop-floor sentencing.
Edition note: the 2014 edition of ISO 5817 gave Level D a limited allowance for short, non-surface-breaking lack of fusion; the 2023 edition removed it. Contracts invoking the older edition by date still sentence against the older table — confirm which edition applies before re-sentencing legacy work.
Context on how the levels work: ISO 5817 acceptance criteria guide.
What causes lack of fusion
Heat input too low
- Current below WPS range, or voltage mismatched to transfer mode
- Travel speed too high for the deposition rate
- Short-circuit MIG/MAG on thick sections — the classic “cold lap” recipe: low heat transfer plus high deposition
Arc placement and technique
- Arc riding the pool: the arc impinges on molten metal instead of the joint face, so the face never melts — typical with excessive stick-out or push angle
- Molten metal rolling ahead of the arc on downhill work or with excessive pool size
- Weave too wide: edges of the weave deposit filler on unfused side-walls
Joint preparation
- Bevel angle too narrow for the electrode/torch to reach the side-walls
- Root gap too tight; misalignment shadowing one face
- Heavy mill scale, rust, or primer on fusion faces
Robotic cells add a systematic variant: a torch TCP error or fixture shift of 1–2 mm moves the arc off the joint centreline and produces repeating side-wall lack of fusion at the same coordinates on every part — exactly the pattern that per-seam thermal monitoring is built to catch on part one instead of at the customer.
Detection: the hard one
| Method | Effectiveness |
|---|---|
| Visual testing | Surface-breaking cases only — most LoF is subsurface |
| Radiography (ISO 17636) | Unreliable for tight LoF: a planar gap parallel to the beam can be invisible |
| Ultrasonic / PAUT | Preferred — planar reflectors return strong echoes; see PAUT weld inspection guide |
| Real-time thermal monitoring | Detects the cause live: a fusion face that never reached melting temperature has a measurable thermal signature during the pass |
The RT blind spot deserves emphasis: a weld “cleared by radiography” can still contain lack of side-wall fusion. Method selection per ISO 17635 should be driven by the expected defect orientation — this is covered in welding inspection methods compared.
Prevention
- Keep the arc on the leading edge of the pool, aimed at the joint face — not buried in molten metal.
- Respect minimum heat input: lack of fusion is the defect that “fast and cold” buys you (see heat input and cooling rate).
- Open the prep if side-walls can’t be reached: a slightly wider bevel is cheaper than UT repairs.
- Clean fusion faces to bare metal.
- Verify torch alignment on robotic lines after every collision, tip change, or fixture swap — and monitor fusion-zone temperature continuously so a cold joint is flagged during the weld, not after the customer’s UT.
Lack of fusion is one of the planar defects mapped in the welding defects guide with ISO 5817 acceptance criteria.
Frequently Asked Questions
What is lack of fusion in welding?
Lack of fusion (ISO 6520-1 code 401) is an area where the weld metal failed to fuse with the parent metal or with a previously deposited pass. The metal touched but never melted together, leaving a planar, crack-like discontinuity. Sub-types: lack of side-wall fusion (4011), lack of inter-run fusion (4012), and lack of root fusion (4013).
Is lack of fusion acceptable under ISO 5817?
No. Lack of fusion (401) is not permitted at any ISO 5817 quality level — B, C or D. Like cracks, it is a planar defect whose sharp edges act as crack initiators, so the standard gives it zero tolerance regardless of how permissive the rest of the level is. Any detected lack of fusion is a repair, not a sentencing discussion.
What causes lack of fusion?
Insufficient heat input is the root cause: current too low, travel speed too high, or arc impingement on the pool instead of the joint face. Contributing factors include too-narrow joint preparation, wrong torch angle that lets molten metal roll ahead of the arc onto cold plate, oversized weld pools in weave technique, and dirty or oxidized fusion faces.
How do you detect lack of fusion?
It is the hardest common defect to find: planar, subsurface, and often unfavourably oriented for radiography. Ultrasonic testing (including phased array) is the preferred method because planar reflectors return strong signals. RT can miss tight lack of fusion entirely. In production, real-time thermal monitoring detects the cold-joint signature — insufficient fusion-face temperature — while the pass is being deposited.
