Undercut is a groove melted into the parent plate along the weld toe and left unfilled. It’s easy to spot, easy to measure, and easy to underestimate: a half-millimetre notch at the toe sits at the exact location where geometric stress concentration peaks — which makes undercut one of the most consequential surface defects for fatigue-loaded structures.
In ISO 6520-1 terms, continuous undercut is 5011, intermittent undercut is 5012, and root undercut is 5013.
It’s also one of the few imperfections where ISO 5817 gives clean, memorable numeric limits — making it the textbook example of how quality levels B, C and D actually differ.
What is undercut in welding?
ISO 6520-1 distinguishes:
| Type | Code | Where |
|---|---|---|
| Continuous undercut | 5011 | Uninterrupted groove along the toe |
| Intermittent undercut | 5012 | Short, separated grooves |
| Root undercut (shrinkage groove) | 5013 | Groove at the root of a butt weld |
The mechanism: the arc melts a channel into the parent metal at the pool edge, and surface tension plus travel speed prevent the pool from flowing back to fill it. The result is a notch with weld metal on one side and plate on the other — the worst possible orientation, transverse to the principal stress in most joints.
ISO 5817 acceptance limits
Undercut depth h scales with material thickness t:
| Type | Code | Level B | Level C | Level D |
|---|---|---|---|---|
| Continuous undercut | 5011 | h ≤ 0.05t, max 0.5 mm | h ≤ 0.1t, max 1 mm | h ≤ 0.2t, max 2 mm |
| Intermittent undercut | 5012 | h ≤ 0.1t, max 1 mm | h ≤ 0.2t, max 2 mm | h ≤ 0.2t, max 2 mm |
| Root undercut | 5013 | Not permitted | h ≤ 0.05t, max 0.5 mm | h ≤ 0.1t, max 1 mm |
Smooth transitions are generally required — a sharp-bottomed groove is sentenced more severely than a smooth one of the same depth. Values are indicative references: verify the current edition of ISO 5817 and any overriding spec. Fatigue codes can be stricter than Level B for critical details.
On a 10 mm plate that means: 0.5 mm tolerated at Level B, 1 mm at C, 2 mm at D — the same weld passes or fails depending on the level your project specifies. How levels are assigned: ISO 5817 quality levels explained · Level B vs C vs D side by side.
What causes undercut
Electrical and travel parameters
- Current too high — the arc digs more parent metal than the pool can refill
- Travel speed too fast — pool solidifies before backfilling the melted edge
- Arc length / voltage too high — wider, shallower arc melting the toes
Technique
- Wrong work or travel angle — the arc heats one toe asymmetrically; classic in horizontal (PB/2F) fillets where gravity drains the pool off the vertical leg
- Excessive weaving — pool edges freeze at the extremes of the weave
- Wrong electrode diameter for the joint and position
Process-specific
- In MAG spray transfer at high current, undercut risk grows with travel speed — exactly the corner of the parameter envelope production wants to run in for throughput. This makes undercut a drift defect: a cell that produced acceptable toes at qualification slides into undercut as contact tips wear, stick-out shifts, or speed is nudged up. A compact thermographic welding camera like HeatCam IR-S can provide the thermal view in embedded and robotic installations. Continuous parameter and thermal monitoring catches that drift before the seam fails VT.
Detection and measurement
Undercut is a visual-testing item (ISO 17637): raking light and a depth gauge (or profile gauge / laser profilometry for automated lines). Two practical rules:
- Measure depth at the deepest point, not an average — sentencing uses the maximum.
- Record position and length: intermittent vs continuous changes the applicable limit.
Full VT procedure context: ISO 17637 visual weld inspection. For automated geometry checks, see weld bead geometry measurement — and the Therness monitoring product lineup that brings those checks inline.
Prevention
- Respect WPS current and speed — undercut is the canonical symptom of running hot and fast.
- Fix the angle: keep the electrode/torch bisecting fillet joints; slight drag angle; aim the arc so both toes wet evenly.
- Control arc length — shorter arc, narrower dig.
- Limit weave width or switch to stringers in position work.
- Watch the toes, not the bead: a bright, shiny toe channel behind the pool is undercut forming in real time — operators (and welding camera vision systems) can see it before the weld is finished. A weld monitoring system watching every pass live flags that toe channel as it opens, so current or travel speed gets corrected before the defect ever reaches final VT.
For a sense of how these surface defects develop under the arc, watch what happens when a weld goes wrong.
Monitoring undercut in real time — the edge geometry route
Undercut—a groove melted into the base metal at the weld toe—is a surface defect that visual inspection catches directly, but only after the weld has cooled. By then, the joint is already nonconforming to the acceptance limit (whatever your quality level specifies), and rework costs time, heating, and material.
The undercut that gets inspected started forming during the welding process. If the arc angle is too steep at the toe, if the arc energy is too high for the wire diameter or travel speed, or if the torch angle puts the arc directly at the edge of the base metal instead of into the joint, the weld pool runs up the side of the base metal and erodes it. This erosion is also observable during the weld: the pool geometry and the thermal distribution at the toe tell you whether conditions favor undercut formation.
Undercut risk is not the same across positions and joint types. Fillet welds in the horizontal and vertical-up positions carry more risk than flat-position butt welds, because gravity pulls the pool away from the upper toe just as the arc is eroding it, and a welder correcting for one toe can undercut the other. High wire feed speed paired with a current setting above the qualified range for that wire diameter produces the same erosion mechanism on mechanised and robotic welds: the arc delivers more energy to the toe than the travel speed and joint geometry can absorb into the bead, and the excess energy melts base metal instead of filler.
Weave technique changes the picture as well. A weave that dwells too long at the outer edge of the pattern — common when a welder widens a pass to fill a wide root gap — erodes the toe on that side even when the average heat input for the pass is within range, because the instantaneous heat input at the dwell point is what drives the erosion, not the pass average.
Thermographic monitoring and edge-tracking vision systems catch this in real time by watching the geometry of the heat input at the toe of the joint. A camera system that tracks bead edge position relative to the joint boundaries can measure the actual width and geometry of the bead and flag when edge profile deviates from the qualified procedure. Abnormal edge geometry—a pulled-in edge, asymmetry, or excessive toe angle—signals that undercut conditions are forming, and the operator or robot halts to adjust.
This is why frame-by-frame thermal and geometric monitoring adds value that a single per-pass average parameter record does not: it captures the instantaneous condition at each point along the toe, not just the pass-level summary. A shop that monitors bead edge geometry continuously, position by position and pass by pass, can show an auditor exactly where along a multi-metre seam the edge profile stayed within the qualified procedure and where it did not, rather than certifying the pass as a whole on parameters that were correct on average but not at every point.
This is the monitoring value for undercut: you do not wait for visual inspection to tell you the bead edge is bad. You adjust it during the weld, on the first seam that starts to show the wrong edge profile. The corrective action is quick (arc angle, travel speed, contact tip angle) and takes minutes, not hours of rework.
The result goes into your traceability record the same way porosity prevention does: a continuously recorded evidence stream showing whether edge-formation conditions remained within the qualified procedure throughout the pass. This record supports your ISO 5817 acceptance decision and your ISO 3834 quality system audit trail.
Monitoring setups: the welding camera hub describes end-to-end; HeatCam IR-C for fixed mounting; comparison of monitoring cameras to pick your system. For the full architecture of how real-time monitoring feeds traceability, see weld bead geometry measurement: thermal and laser profilometry.
”Under Cut” vs. Undercut: One Defect, Different Spellings
Undercut turns up in the field as “undercut,” “under cut,” and “undercutting” almost interchangeably — welders speak the term the way they learned it on the shop floor, and none of the variants point to a different defect. Whether it is written as two words, one word, or turned into the “-ing” form, the underlying imperfection is the same groove melted into the parent metal at the toe or root and left unfilled. ISO 6520-1 codes it as reference 5011 (continuous), 5012 (intermittent) or 5013 (root), and the applicable ISO 5817 limit for that code does not change with the spelling.
The spacing and the -ing form are not entirely neutral, though. “Undercutting” as a verb can describe the action — a welder or a robot undercutting a joint while it welds — while “undercut” as a noun is the defect that results once the arc has passed and the pool has solidified without backfilling the groove. In practice the distinction rarely matters for sentencing: an inspector measuring depth h against the quality level does not care whether the finding is logged as undercutting or undercut, and neither does ISO 6520-1’s classification, which names the result, not the verb that produced it.
Where the spelling does matter is in a weld log or an inspection report meant to survive an audit. A traceability record that mixes “undercut,” “under-cut” and “undercutting” across different reports for the same joint reads as three findings to someone reconciling the file six months later, even though it is one. Standardising on the ISO 6520-1 term and reference number — “undercut, 5011” rather than a locally invented phrase — keeps the finding searchable and keeps a repeat occurrence on the same line visible as a trend rather than as three unrelated line items.
For the one-paragraph version of the definition and mechanism, see the undercut glossary entry; for the classification reference and the causes/detection matrix mapped to code 5011 specifically, see the continuous undercut defect page.
Undercut is one of the defects in our complete reference: welding defects guide with ISO 5817 acceptance criteria.
Stop undercut before it reaches final VT
See how real-time monitoring flags a toe channel as it opens, so current and travel speed get corrected on the weld where it happens.
Book a demo See how HeatCore AI catches driftTherness solution: Welding monitoring system — Every weld inspected in the cycle, with defect flags and the ISO 3834 / ISO 17635 evidence kept per weld.
Frequently Asked Questions
What is undercut in welding?
Undercut (ISO 6520-1 code 5011 for continuous undercut or 5012 for intermittent undercut) is a groove melted into the parent metal along the weld toe that is left unfilled by weld metal. Root undercut is classified as 5013. It reduces the load-bearing cross-section locally and creates a sharp notch exactly where weld geometry already concentrates stress, making it one of the most fatigue-relevant surface imperfections.
How much undercut is acceptable under ISO 5817?
For continuous undercut, ISO 5817 limits depth h to 0.05t (max 0.5 mm) at Level B, 0.1t (max 1 mm) at Level C, and 0.2t (max 2 mm) at Level D, where t is material thickness. Intermittent undercut gets slightly more allowance, root undercut less. Smooth transitions are generally required. Always verify the current edition of the standard and your governing code.
What causes undercut in welding?
The classic causes are welding current too high for the position, travel speed too fast for the weld pool to backfill the melted groove, wrong torch or electrode angle, excessive arc length or voltage, and excessive weaving. In fillet welds in the horizontal position, gravity pulling the pool away from the vertical leg is a frequent contributor.
Can undercut be repaired?
Yes. Light undercut within acceptance limits can be left or blended by grinding to a smooth transition. Undercut beyond limits is excavated if needed and filled with an additional stringer pass under a qualified procedure, then re-inspected visually — and the repair recorded against the joint ID for traceability.
Is 'under cut' the same defect as undercut in welding?
Yes. 'Under cut,' 'undercut' and 'undercutting' all describe the same imperfection: a groove melted into the parent metal at the weld toe or root and left unfilled by weld metal. ISO 6520-1 classifies it under reference 5011 (continuous), 5012 (intermittent) or 5013 (root) depending on where it occurs, and the ISO 5817 limit for each of those codes applies regardless of which spelling or verb form was used to describe it.
