Undercut in Welding: Definition, Causes, Limits & Repair

A fabricator once passed a visually perfect-looking handrail, only for the QA inspector to reject three of six welds with a fillet gauge and a thumbnail run along the toe. The welds looked beautiful from a distance, but a narrow unfilled groove in the base metal had been hiding at the weld toe the whole time.

What Is Undercut in Welding? A Clear Definition

If you ask three welders what undercut is, you will usually get three versions of the same answer: “a groove at the edge of the weld,” “the thing the inspector flags,” or “that line along the toe.” All of them are circling the same precise definition, and it matters that you know the exact wording, because that wording is what shows up in a written inspection report.

The official definition. AWS A3.0, the standard welding terminology document, defines undercut as a groove melted into the base metal adjacent to the weld toe or weld root and left unfilled by weld metal. Read that slowly and note the two important phrases: melted into the base metal and left unfilled. Undercut is not extra weld metal, and it is not a crack. It is a channel the arc cut into the parent plate that the filler never came back to fill in.

Weld Toe vs Weld Root: Where Undercut Lives

Undercut appears in two locations, and inspectors treat them differently.

• Weld toe undercut sits on the surface, along the edge where the face of the weld meets the base metal. This is the version you see and feel, and it is the one tied most directly to fatigue failures.
• Root undercut is cut into the base metal at the root of the joint, on the back side of a single-sided groove weld. It hides from visual inspection unless the back face is accessible, which is why root side cutting is such a sensitive topic in pipe work.

The distinction matters because the toe is a free surface where stress concentrates, while the root reduces the load-bearing cross section of the joint. Both are unacceptable when they exceed limits, but they are located, measured and sometimes evaluated by different methods.

Undercut vs Normal Weld Profile

A good fillet weld does not stand as a proud triangle perched on the plate. It blends into the base metal through a small curved transition. Beginners often mistake that natural blend for undercut, especially on an aluminum job where the surface is bright and the light plays tricks. A properly fused toe is smooth and continuous, with the weld metal wetting out into the plate. Undercut is different: it is a recess, a shadow line, a small notch you can catch a fingernail in.

That is the practical thumbnail test. Run your thumbnail along the weld toe. If it glides smoothly, the profile is fine. If it catches, drops into a groove, or you can see a dark line that stays dark even after wire brushing, you are feeling undercut. Good visual inspection techniques and good technique habits go hand in hand; welders who learn to feel their own work rarely pass defective welds forward. The same habit applies across processes, whether you are running stick or learning rod selection for out-of-position work.

Discontinuity or Defect? Why the Wording Matters

In code language, undercut is classified as a surface discontinuity, one detected primarily by visual inspection (VT) before any other method is used. A discontinuity is simply any interruption in the normal structure or geometry of the weld or base metal. Porosity, undercut, overlap and minor porosity are all discontinuities by default.

A discontinuity becomes a defect only when it exceeds the acceptance criteria of the governing code or contract specification. In other words, there is no such thing as “undercut” as an automatic failure. There is only undercut within limits or undercut beyond limits, and the number that separates the two comes from the code written on the drawing, such as the acceptance table applied to that joint. This is why a weld can look rejected to one person and pass to another: they are judging appearance, not measuring against criteria.

The scope is broad. Undercut is not a process-specific flaw. It occurs in groove welds and fillet welds, in plate and pipe, and in every arc process that uses heat to melt the base metal: shielded metal arc, gas metal arc, flux-cored arc, submerged arc, gas tungsten arc, plasma arc and their variants. If the arc can cut a groove, an operator can leave one unfilled. The chapters that follow separate it from its look-alikes, map where it comes from, and put numbers on how much is allowed.

Undercut vs Other Weld Discontinuities

Walk a busy shop floor with an experienced inspector and you’ll hear the same opening line over and over: “There’s a groove right there.” What comes next decides whether you’re looking at undercut, underfill, overlap, or something far more serious. Undercut shares visual territory with several other weld imperfections, and misidentification is genuinely common. Worse, the correct fix for one is often useless, or actively harmful, for another.

The cleanest mental model is this: undercut is the one common discontinuity where the base metal itself has been melted away and never refilled. Everything else is either a weld-metal shape problem (too little metal, too much metal, wrong profile) or an internal integrity problem (porosity, lack of fusion, cracks). TWI’s technical review of geometric shape imperfections organizes them the same way, and most weld codes follow that taxonomy.

Underfill vs Undercut: Two Grooves, Two Meanings

Underfill is the most frequent mix-up. In a groove weld, underfill means the weld face or root surface sits below the adjacent base metal surface because not enough weld metal was deposited. The joint edges are intact; the weld simply came up short. Undercut is the inverse geometry problem: the arc gouged a notch into the base metal at the toe or sidewall, and nothing filled it back in. A single weld can carry both, which is precisely why codes list and measure them separately.

Overlap: Undercut’s Mirror Image

Overlap is weld metal that rolled over the base metal without fusing to it, so the toe laps onto the plate instead of tying in smoothly. It looks similar at a glance because both flaws leave a sharp step at the weld toe, but the parameters behind them point in opposite directions. Undercut usually signals too much current, too much voltage, or an arc that wandered. Overlap usually signals too little heat or a travel speed that let the puddle pile up.

Profile Flaws and Internal Defects

Excessive concavity and convexity are profile problems measured on the weld face itself. Concavity in a fillet weld thins the effective throat; convexity builds a bulging face with an abrupt toe angle. Neither cuts a notch into base metal, but both stack additional stress concentration right where undercut already lives.

Porosity and lack of fusion belong to an entirely different category: internal discontinuities that a thumbnail test cannot confirm. Porosity is trapped gas; lack of fusion is an unbonded joint interface. These normally require radiographic or ultrasonic examination to size and locate. Some processes are simply more prone to them. Porosity in aluminum, for example, is frequently driven by inadequate surface cleaning or hydrous oxide, which is why correct prep guidance such as our aluminum cleaner for welding page matters as much as machine settings. Similarly, aluminum MIG welding carries a well-known lack-of-fusion risk because aluminum hides heat input so effectively.

Cracks sit at the top of the severity ladder. A crack is a tensile rupture, not a shape imperfection, and it propagates under load. The distinction carries straight into repair work: treating a hairline crack as cosmetic undercut is how a part fails in service.

Side-by-Side: How Inspectors Tell Them Apart

Discontinuity	What it actually is	Where to look	Typical detection
Undercut	Notch melted into base metal at toe or sidewall, left unfilled	Weld toe, sidewall	Visual, undercut gauge, fillet gauge
Underfill	Weld face or root below adjacent base metal surface	Groove weld face/root	Visual, profile gauge
Overlap	Weld metal lapping over unfused base metal	Weld toe	Visual
Concavity / convexity	Hollow or bulging weld face	Weld face	Visual, fillet gauge
Porosity	Trapped gas pores in weld metal	Weld interior or surface	RT or UT; surface-breaking pores visible
Lack of fusion	Unbonded weld-to-joint interface	Root, sidewall, or between passes	UT, RT (limited sensitivity)
Crack	Tensile rupture, linear and sharp	Weld metal, HAZ, or base metal	Visual if surface; RT/UT if buried

The practical field test sequence is simple: dry the weld, wipe off the spatter, light it obliquely, then run a thumbnail along the toe. A definite step that stops the nail is undercut. A smooth edge with a low weld surface is underfill. A rolled-over lip is overlap. If nothing catches but the part still matters structurally, internal examination is the next conversation.

Why the Distinction Changes the Repair Route

Each discontinuity demands a different remedy, and this is where misidentification gets expensive:

• Undercut is corrected by blending grinding or by depositing a small stringer bead at reduced current to restore the profile.
• Underfill needs weld metal added, not grinding — grinding here removes base metal you can’t spare.
• Overlap is blended back to a smooth, tied-in profile with grinding; no weld metal needs to be added.
• Porosity and lack of fusion generally demand gouging or back-gouging down to sound metal, then rewelding with a qualified procedure — or outright rejection when the discontinuity exceeds the acceptance class.
• Cracks must be removed completely to sound metal before anything else happens, followed by a root-cause look at procedure, parameters, and restraint. Grind a crack flush and call it fixed — the logic some people apply to cosmetic wheel damage, which is exactly why crack handling is a central part of any alloy wheel welding procedure — and you have a repair that fails the next inspection, or worse, fails in service.

The habit worth building is this: name the discontinuity before you reach for a grinder. Undercut has one repair logic, underfill another, and cracks have none to share with either. Once you can sort shape problems from internal problems, the correct fix — and the correct code limit to check against — usually identifies itself.

Types of Undercut: Toe, Root, and Arc-Blow Undercut

Undercut is not one defect with one look. Once you can define undercut in welding, the next skill is naming the type in front of you, because the type decides how you find it, how you measure it, and which repair is allowed. TWI’s review of geometric weld imperfections sorts these variations by location and origin, and inspectors are trained to do the same. Working shops meet four: toe (sidewall) undercut, root undercut, multi-pass undercut that traps slag, and undercut driven by magnetic arc blow.

Sidewall (Toe) Undercut

Toe undercut is the classic form: a groove melted into the base metal along the fusion line, where the weld face meets the plate. On a groove weld it sits on one or both sidewalls; on a fillet weld it shows at one or both toes. The arc gouges the parent metal and the filler fails to flow back in, leaving a sharp line a thumbnail catches on even when the bead looks sound from a distance. Excessive edge heat is usually to blame: high current or voltage, a long arc, fast travel, or a work angle aimed at the plate rather than the joint corner. The Fabricator’s coverage of acceptable undercut makes the practical point: the groove matters because it is a notch, not because it is ugly.

Root Undercut in Single-Sided Groove Welds

Root undercut occurs at the root of a joint welded from one side only, such as a single-sided butt or V-groove weld. The root pass runs too hot or too fast and melts away part of the root face instead of tying into it. The trouble is access: on pipe and closed structures that groove sits on the bore surface or back side, where no inspector’s eye can reach. The Iowa DOT engineering research letter on weld inspection explains why internal root discontinuities push projects toward volumetric examination.

Back-gouging matters here too. When a welder uses carbon arc gouging to open the root for a sound second-side pass, a careless profile leaves a groove that behaves exactly like undercut if the refill pass does not wet those edges.

Multi-Pass Undercut That Traps Slag

In a multi-run weld, an undercut groove at the toe of an intermediate pass is not just a blemish; it is a pocket. The next pass may bridge across it instead of filling it, sealing flux and slag inside, and that trapped slag becomes a second discontinuity that reads as an indication on a radiograph. Flux-based processes are most vulnerable, and welders running flux core wires on aluminum or steel know how stubborn a bridged-over slag pocket can be. The fix is technique, not grinding: each pass should wash into the previous toe, and root and hot passes must be slag-free before the next layer goes on.

Arc-Blow and Magnetism-Induced Undercut

When a part is magnetized, or the field around a DC arc is unbalanced, the arc does not travel where the welder points. This arc blow drags it sideways and concentrates heat on one edge, so the signature is an uneven groove along one toe while the opposite side looks fine, often worse near starts, stops, or the end of a run. Demagnetizing the workpiece, using AC where the process allows, lower current with a drag technique, and a tight arc all reduce it; heavy sections handled with magnetic lifting gear are the usual suspects.

How Position Changes the Picture

Gravity is a co-defendant in positional undercut. In horizontal fillets the pool sags, starving the upper toe, especially if the work angle favors the top plate. Vertical-up runs invite undercut at weave edges when current is high and travel is fast, while vertical-down is notorious because speed outpaces the pool’s ability to wet the sides. Overhead demands short arcs, low current, and small stringers, since excess heat undercuts the toes. Flat position is the most forgiving but still melts sidewalls at high amperage. A practical all-position welding rod guide is useful reading, since electrode choice and positional technique are inseparable.

Type	Where it appears	Usual trigger	How it is found
Toe / sidewall	Fillet toes, groove sidewalls	High heat, long arc, fast travel	Oblique light, thumbnail, gauges
Root	Root face, pipe bore	Hot or fast root pass	Back-side visual, RT or UT
Multi-pass	Intermediate pass toes	Bead bridges the groove	Radiography, UT, gouge-out
Arc blow	One side, near ends and stops	Magnetized part, DC field	Visual asymmetry, gaussmeter

Before any repair decision or code acceptance call, identify the type. Toe undercut is a surface geometry problem you can see and gauge; root undercut is a hidden geometry problem that may need NDE; arc-blow undercut is a symptom of magnetism that will recur on the next bead unless the cause is fixed. Naming the type correctly is the first step of every competent repair plan.

Why Undercut Happens: Root Causes by Process

Undercut is rarely one careless moment. It is an arc-management problem that builds over a few inches of travel: the arc melts more base metal than the filler can refill, and nothing is left to restore that lost volume. Nearly every cause in welding practice traces back to that imbalance. Once you see it that way, troubleshooting gets much simpler.

The Core Imbalance: Heat In, Metal Back

Every weld puddle sits between two competing forces. Arc energy melts the base metal and opens a groove; deposited filler fills that groove back in. Undercut appears whenever the first outpaces the second. TWI identifies the classic combination as a wide-spreading arc from high voltage, paired with insufficient filler input from low current or overly fast travel. The arc carves a shallow trough along the toe, and the electrode or wire simply does not deliver enough metal to cover it before the puddle freezes.

Seven variables drive this imbalance in practice:

• Excessive current — the most frequently cited cause, since it deepens penetration and widens the puddle faster than deposition keeps up.
• High arc voltage or a long arc — spreads the arc cone sideways into the toe instead of concentrating it ahead of the puddle.
• Travel speed too fast — leaves no time for filler to flow into the groove before it freezes.
• Wrong electrode or torch angle — directs heat and force away from the corner that needs filling.
• Poor weaving technique — sweeping the joint with no dwell at each toe.
• Oversized filler or a steep deposition angle — pushes molten metal out of the corner instead of into it.
• Base metal already too hot — interpass drift or long uninterrupted passes keep the puddle fluid and draining.

SMAW: Arc Length, Electrode Size and Technique

Stick welding gives the welder the most direct control and the most ways to go wrong. Excessive amperage for the electrode diameter is the usual culprit, closely followed by a long arc held as the rod burns down. Weave width matters too: a wide side-to-side motion with no pause at the toes lets the arc carve each edge and move on before filler arrives. On horizontal fillets, pointing the arc too far into the vertical plate melts that plate’s edge, and gravity pulls the molten metal away from the toe, leaving a notch behind. Angle is fixed before amperage, which is why Advantages of Stick Welding is worth reading while you build rod control.

GMAW and FCAW: Voltage, Stickout and Speed

Short-circuit and spray transfer fail differently. Spray transfer combines high voltage with high current to produce a wide, fluid puddle that runs ahead of the wire, and fast-lap or out-of-position work compounds it through travel speed. Self-shielded flux core adds a variable of its own: incorrect stickout changes effective current, while a push gun angle concentrates heat at the leading toe. Aluminum spreads heat through the plate so readily that the problem grows, which is why the parameter ranges in Aluminum MIG Welding: The Complete Practical Guide deserve close attention.

SAW and GTAW: Opposite Extremes

Submerged arc rarely undercuts from operator technique, since nobody is hand-carrying the arc. Instead, excessive current for the wire size, travel speed that outruns the flux’s filling ability, and wrong electrode angle or multiple-wire spacing push the arc into the side plates. The result usually shows up as a smooth, consistent sidewall groove running over a long length.

GTAW undercuts for different reasons: a torch held too upright with too much heat, an arc aimed into the joint corner, or a weave that never dwells at the toes. Because GTAW deposits so little metal per pass, any excess base-metal melting shows up immediately as a notch.

Troubleshooting Logic: Work the Variables in Order

Experienced inspectors and welders narrow causes the same way, in a fixed order:

1. Check current first. Reduce by about 10% and retest; this resolves more undercut than any other single change.
2. Check arc length and voltage. Shorten the arc, and on constant-voltage processes confirm the setting matches the required bead width.
3. Check travel speed. If the puddle looks narrow and sluggish against the groove, slow down.
4. Check angle and technique. Point the arc at the joint corner and confirm a real dwell at each toe.
5. Check heat history. Verify interpass temperature and what earlier passes have already put into the joint.

Only then should anyone change filler size, shielding gas, or the whole procedure. Reaching for a new wire diameter before verifying amperage is one of the most common and most expensive misdiagnoses on a production floor: it sends the WPS through a revision while the real cause sits in the machine’s settings.

Why Undercut Matters: Stress Concentration and Fatigue

Plenty of welders look at a shallow groove at the weld toe and shrug. It is not a hole, the joint still holds, and it “looks fine.” That reaction is understandable, but it is also where fatigue failures begin. Undercut matters not because of how much metal it removes, but because of the shape it leaves behind: a sharp, notched re-entrant corner sitting exactly where the load path already crowds into the weld.

The Notch Effect: Why Geometry, Not Metal Loss, Is the Real Problem

Stress does not spread evenly through a welded joint. The load lines squeeze together as they pass through the weld throat and bend around the toe, so the local stress at the toe is higher than the average (nominal) stress you would calculate from the applied load. Engineers describe that ratio as the stress concentration factor. A smooth, gently blended weld toe spreads the load over a generous radius. An undercut replaces that radius with a sharp notch, and the sharper the notch root, the more intensely the stress concentrates in a tiny zone of base metal.

This is the notch effect, and it is why geometry dominates the fatigue question. TWI’s technical review of geometric shape imperfections is blunt about it: sharp discontinuities such as undercut are among the most damaging imperfections a welded joint can carry, precisely because they concentrate stress at a point where the metal is already affected by the thermal cycle of welding.

The fatigue failure sequence then runs in three stages: local plastic straining at the notch, crack initiation, and stable crack growth cycle after cycle. Undercut effectively pre-packages the first stage. Instead of waiting years for a smooth toe to develop micro-cracks, the joint starts life with a sharp feature that behaves like an initiated crack. In components that survive millions of load cycles, the initiation phase normally consumes most of the service life. Remove that phase and you remove most of the life.

What the Research Actually Shows

The numbers from published fatigue studies are sobering:

• Research published in the International Journal of Fatigue (2025) reports that fatigue cracks most commonly initiate from local imperfections such as undercut, confirming what inspectors see in the field: the toe is the starting line.
• An earlier fatigue study found that a 1 mm deep undercut can reduce fatigue strength by roughly 30% in fillet-welded joints (published in the International Journal of Fatigue). One millimetre is a defect you can miss with a casual glance, yet it costs nearly a third of the joint’s endurance limit.
• Work on butt welds found that the fatigue strength reduction from a toe undercut is at least twice that of a flush-ground joint without undercut (University of Texas at Austin fatigue research). The comparison is instructive: same weld, same throat, same metal. The only difference is the toe geometry.

Static vs Cyclic Service: Why Codes Split the Rules

Undercut removes very little cross-section, so on a structure loaded once, or loaded gently and held, its effect is minor. The metal can redistribute stress, and the joint behaves close to its nominal capacity. That is why codes allow more undercut on statically loaded members.

Cyclic service is a different world. Bridges, crane booms, pressure-cycling vessels, chassis, and mobile equipment all subject welds to repeated stress ranges, and every cycle works on the notch. This is the engineering logic behind the split acceptance limits you will find in the codes: AWS D1.1’s Table 8.1, for example, sets a noticeably tighter depth allowance for cyclic-loaded welds than for static ones, and the length of the undercut run matters as well as its depth, since a long continuous notch behaves worse than a short isolated nick. The exact numbers are laid out in the acceptance-limits section that follows; the point here is why they exist at all.

Service condition	How undercut behaves	Inspection posture
Static / single overload	Minor net-section loss, stress redistributes	More tolerant; depth and length limits are relaxed
Cyclic / fatigue loaded	Serves as a crack initiation site every cycle	Depth tightly limited, length accumulates, rejection threshold is low

Real-world cyclic examples are everywhere. A boat hull welds through wave-induced flexing for decades, and an alloy wheel repair joint sees a road-load cycle every few seconds at speed. Neither part will fail from one pull; both can fail from a notch that was never properly evaluated.

The practical takeaway for anyone reading a visual inspection report: treat undercut as a fatigue problem first and a cosmetic problem never. A weld that is beautiful everywhere except the toe is a weld that has already started aging.

Acceptable Undercut Limits: AWS D1.1, D1.5, ISO 5817 and More

“Is this undercut acceptable?” has no universal answer. Codes judge undercut by numbers, and those numbers move with three variables: base metal thickness, whether the joint carries cyclic load, and which document the contract actually names. A groove that passes on a static handrail fails on a bridge girder, and both are judged differently again on a pressure vessel or pipeline.

AWS D1.1: The Limits Most Structural Shops Live By

In the United States, structural steel fabrication runs on AWS D1.1, Structural Welding Code — Steel, and undercut acceptance sits in Table 8.1, the visual criteria an inspector applies with a gauge. The Fabricator’s plain-English breakdown of acceptable undercut pairs well with the table itself. The short version:

• Static welds in base metal thinner than 1 in [25 mm]: undercut limited to 1/32 in [1 mm].
• Static welds in base metal 1 in [25 mm] or thicker: 1/16 in [2 mm], because a thicker section carries the notch with less penalty.
• Cyclic welds transverse to the primary tensile stress: 0.01 in [0.25 mm]. Not a typo, roughly the thickness of a sheet of paper, shallow enough that most welders find it only with a thumbnail.
• Cyclic welds running parallel (longitudinal) to that stress: 1/32 in [1 mm], since the notch sits aligned with the stress flow instead of across it.
• Root undercut gets its own line in the same table; in cyclic service it is also held to 0.01 in [0.25 mm].

Depth is only half the test. A weld can measure under 1/32 in at every point and still fail if the toe is chewed along the run, so inspectors sum undercut across a defined length of weld, with 12 in [300 mm] the usual gauge length. Project specifications then cap that accumulated total, with 2 in [50 mm] in any 12 in a common threshold. Treat fragmented undercut as one continuous problem, not dozens of small nicks.

The 2025 D1.5 Bridge Welding Code: A Measured Shift

Bridge welding falls under AWS D1.5, always stricter than D1.1 because bridges live on cyclic traffic load, where undercut is treated as a fatigue problem rather than a cosmetic one. The 2025 edition rewrote several provisions, as explained in the AWS Welding Digest analysis of the new undercut tolerances and treatments:

• The 0.01 in [0.25 mm] limit stays in place where fatigue demand is highest.
• A 1/16 in allowance appears in less demanding locations, bundled with finishing requirements.
• A new 1/8 in [3 mm] tolerance at the corners of stiffeners and connection plates, zones that were automatic rejects under the old text.

This was not relaxation for its own sake. Fatigue research on welded joints, including work published in the International Journal of Fatigue, shows that the damage undercut causes depends on notch orientation, location, and whether the toe was blended smooth afterward. The wider tolerances arrive paired with grinding and treatment requirements, and fracture-critical applications remain tightly controlled.

ISO 5817 and the Quality-Level System

European practice, and much of the world outside the United States, sorts weld quality into three levels under ISO 5817. Level B applies to highly stressed and dynamically loaded joints, Level C to general fabrication, and Level D to non-critical work. Undercut limits follow that ranking:

Code and Condition	Maximum Undercut Depth
AWS D1.1, static, base metal < 1 in [25 mm]	1/32 in [1 mm]
AWS D1.1, static, base metal ≥ 1 in [25 mm]	1/16 in [2 mm]
AWS D1.1, cyclic, weld transverse to primary tensile stress	0.01 in [0.25 mm]
AWS D1.1, cyclic, weld longitudinal to primary stress	1/32 in [1 mm]
AWS D1.5 (2025), stiffener and connection plate corners	1/8 in [3 mm], new tolerance
ISO 5817, quality level B	0.05t, max 0.5 mm
ISO 5817, quality level C	0.5 mm
ISO 5817, quality level D	1 mm

Material matters just as much as geometry. ISO 5817 covers steel, while aluminum falls under ISO 10042 and, on the US side, AWS D1.2, and those numbers are not interchangeable with structural steel limits. Whether you are running MIG (see our aluminum MIG welding guide) or flux-cored wire (our aluminum flux core welding guide), never quote D1.1 acceptance values on an aluminum job.

Other documents carry their own tables. API 1104 for pipelines controls undercut with both depth and length criteria, while ASME B31.3 process piping and ASME BPVC Section VIII pressure-vessel work defer to the acceptance criteria written into the fabrication specification. When a project tightens limits for high-strength low-alloy steels or fatigue-critical machine parts, filler selection and procedure qualification, covered in our guide to alloy steel welding electrodes, become part of holding that line.

Reading the Tables Without Getting Burned

1. Identify the loading category first: static or cyclic, and which way the primary stress runs relative to the weld.
2. Use the thickness of the thinner joined part, not the thicker one.
3. Measure at the deepest point on a clean surface. Mill scale and spatter lie to a gauge.
4. Accumulate length before accepting, rather than judging each nick in isolation.
5. Confirm the edition named in the contract. The current D1.1 edition and its amendment are not the 2015 text, and D1.5 2025 is not D1.5 2020.
6. Remember that the project specification outranks the code minimum.

Undercut acceptance is a set of numbers, not an opinion. Learn which table governs your joint, measure it honestly, and half the argument at the inspection table disappears.

How Undercut Is Inspected and Measured

Finding undercut starts with your eyes, not a gauge. Because undercut is a surface-breaking discontinuity, visual inspection remains the primary acceptance check under AWS D1.1 Clause 8.9, and most codes agree. But “visual” doesn’t mean a casual glance under shop lighting. Done properly, it is a trained, methodical process, and the difference between catching a rejectable groove and missing it is usually technique.

Visual Detection: Light Angle and the Thumbnail Test

The two most practical detection methods are also the oldest. First is oblique lighting: instead of shining a light straight down on the weld, angle it across the weld toe at a low raking angle so shadows reveal depth. A 1/32 in (0.8 mm) undercut that looks like nothing under flat overhead light suddenly reads as a dark, continuous line when the beam comes in from the side. AWS guidance recommends exactly this approach, and it costs nothing but a flashlight held at the right angle.

Second is the thumbnail test, sometimes called the fingernail test. Draw your thumbnail along the weld toe. If it drops into a groove and rides there, you have undercut; if it slides smoothly over a rounded transition, the profile is acceptable. It sounds primitive, but experienced inspectors use it as a fast screen before committing to gauges, and it works well on fillet welds where undercut hides on the toe line. AWS Certified Welding Inspector resources describe it as standard field practice.

Measuring Depth: Undercut Gauges and Depth Micrometers

Once you suspect undercut, you measure it. The instrument of choice is an undercut gauge (also called a weld undercut comparator or depth gauge). The principle is simple: the gauge body sits flat and is zeroed on the base metal surface adjacent to the weld, then the fine pointer is lowered to the bottom of the groove. The depth reading tells you whether the undercut is 0.005 in or 0.030 in — which in most codes is the difference between accept and reject.

The critical rule: depth is measured from the parent metal surface, not from the face of the weld. Reading from the weld cap is a classic beginner error, because deposit height has nothing to do with how deeply the base metal was melted away.

Alternatives include depth micrometers on a bridge stand, small machinist’s rules read at a raking angle, and comparison templates. Fillet weld gauges (the classic Reynolds or “smart” sets) are excellent for leg size and throat, and their notched edges help flag profile problems, but they are poor tools for quantifying undercut depth — use them for sizing, then switch to a proper depth gauge.

When UT or RT Enters the Picture

Undercut is surface-breaking, so radiography shows it poorly compared to internal porosity or lack of fusion, and ultrasonic testing isn’t designed to size surface notches. Where UT genuinely helps is in specialized setups: phased-array UT with TOFD or angle-beam scans can detect and size toe cracks that started at undercut, which matters when fatigue-critical joints are assessed after service. Magnetic particle and dye penetrant testing are the practical complement — they tell you whether undercut has already nucleated a crack, not how deep the undercut itself is. For background on how geometric imperfections interact with cracking, TWI’s review of geometric shape imperfections is a solid reference.

The Inspector’s Workflow and Common Measurement Errors

A repeatable field workflow looks like this:

1. Clean the weld and surrounding base metal thoroughly.
2. Perform an unaided visual scan along the entire weld length at both toes.
3. Re-scan with oblique lighting, checking for continuous shadow lines.
4. Run the thumbnail test along each toe as a tactile screen.
5. Gauge every suspect location for depth and note the length of each continuous run.
6. Record against the applicable acceptance table, including the accumulated-length rule.
7. Escalate to PT/MT only if a crack-like indication is suspected.

Measurement errors are depressingly common. Mill scale on the base metal can offset your zero point by a few thousandths, producing false depth readings — file or grind a small clean patch beside the groove before zeroing. Weld spatter beads near the toe can hide undercut entirely or catch the gauge body and tilt it. Reading at an angle exaggerates depth on a V-shaped groove. And inspectors sometimes measure the deepest point of a 6 in groove and record it as representative of the whole weld, when codes judge both depth and the length of each continuous run.

Finally, the acceptance limits you measure against change with the application. A handrail under static-load criteria is judged differently than a bridge detail under cyclic traffic loading. The The Fabricator’s guide to acceptable welding undercut and the 2025 D1.5 bridge code changes both stress one point: confirm which code and which table governs before the first measurement, not after the argument starts. If you’re training inspectors and welders alike, structured programs such as an aluminium welding course typically build these measurement skills into their quality-control modules.

Repairing Undercut: Grinding, Rewelding, and the 2025 Code Shift

Finding undercut is only half the problem. The next question is how to fix it without making the joint worse. The route depends on depth, loading type, and the governing code. A bad repair often costs more than the original weld.

The Two Traditional Repair Routes

Undercut repairs generally fall into two categories:

• Blend grinding. A soft abrasive wheel or rotary tool rounds the sharp notch at the weld toe into a smooth transition. It removes the stress raiser, but also base metal, so depth and width have limits.
• Weld-metal build-up. Added weld metal fills the groove and restores the profile. This keeps the cross section intact but puts a new heat cycle into an already heat-affected zone.

Under AWS D1.1, undercut requiring repair is cleaned per the weld-cleaning clause, then repaired by depositing additional weld metal. In practice, that means grinding the groove bright and sound, removing oxide, oil, and spatter, and running a controlled bead tying into the base metal and existing weld face. A small-diameter electrode or reduced-current stringer bead keeps heat input low, so the repair does not recreate the undercut it is fixing.

Gouging, Preparation, and Re-Inspection

When undercut is deep or sits beside another discontinuity, blending is not enough. The area may need arc-air gouging or grinding back to sound metal before rewelding. The preparation must leave a clean, accessible groove with no trapped slag. Undercut bridged over by a capping pass often hides slag inlay underneath, and that must be removed, not covered again.

Every repaired weld must be re-inspected by the same method originally used and against the same acceptance criteria. A joint that passed visual plus ultrasonic the first time passes visual plus ultrasonic again, not visual alone, and the limits do not relax because the weld has already been scrutinized once.

The 2025 D1.5 Shift: Grinding Gets Preferred

The biggest change for fabricators is in the 2025 AWS D1.5 Bridge Welding Code. For undercut within allowable depth, it now prefers grinding to a smooth surface over cosmetic weld repair. This is a philosophy shift: instead of adding a small bead that re-heats the toe and risks a fresh notch, grinding removes the notch geometry itself.

The catch is a firm limit: grinding must not reduce the net cross section below 98% of nominal area. Light blending that holds you above 98% is the fast, code-preferred fix. If the undercut is wide or deep enough that blending would drop the section below that threshold, grinding alone is unacceptable and weld metal must be added back.

Situation	Preferred Treatment
Undercut within allowable depth, light blending	Grind to a smooth surface (D1.5 2025 preference)
Grinding would drop net section below 98%	Deposit additional weld metal; grinding alone not permitted
Deep undercut, or trapped slag/other defects	Excavate to sound metal, re-weld, re-inspect
Cosmetic toe profile on non-critical detail	Optional toe grinding as a fatigue improvement

D1.5 and D1.1 do not always agree, so confirm which code the project specification invokes before anyone touches a grinder. The fabricator summarizes acceptable welding undercut and how the limits differ between static and cyclic service.

Repairing Undercut on Aluminum

Aluminum deserves its own warning: it shows no color change to signal overheating, so a repair bead can burn through thin plate or soften the joint if amperage is not controlled. Clean the groove with the dedicated Aluminum Cleaner for Welding: How to Prep Aluminum Right process and remove every trace of oxide before tying in a bead. On thin-wall work such as Aluminum Boat Welding: The Complete Guide for Strong Hull Joints, the safer route is usually a low-amperage TIG tie-in rather than a grinder-heavy fix. Cracked wheel rims, where distortion is unforgiving, need the procedures in Alloy Wheel Welding: A Comprehensive Guide to Safe and Effective Aluminum Wheel Repair first.

A Practical Repair Sequence

1. Measure the undercut depth and length against the governing code limit.
2. Choose grinding or weld-metal build-up using the 98% net-section test.
3. Clean the area per the weld-cleaning clause: scale, spatter, and slag out.
4. Blend to a smooth contour, or deposit a low-current stringer bead.
5. Dress the repair so no new notch is introduced.
6. Re-inspect by the original method and criteria, then document the result.

Done this way, repair stops being guesswork: the grinder or the electrode is chosen by measurement and code, not by whichever tool sits nearest the bench.

Common Undercut Mistakes, Prevention, and Practical Shop Tips

Undercut rarely comes from a welder who does not know what it is. It comes from small shortcuts repeated a few hundred times a shift until a groove at the weld toe becomes normal. The fix is usually a settings change and a hand-position change, not a new machine.

The Mistakes That Keep Showing Up on the Repair Pile

• Capping hot and fast. The root and fill passes are careful, but on the last bead the machine is still set for root-pass heat and the travel speed is pushed up. Because earlier passes have already heated the joint, filler wets easily while the arc still cuts the toes, and the undercut appears only in the final bead. The cap should usually run cooler and slower than the passes beneath it.
• Chasing speed with a long arc. Stretching the arc raises voltage, which widens and deepens the crater at the toe. Trade guidance on acceptable undercut keeps returning to the same basics: arc length about one electrode diameter, moderate voltage, and travel slow enough that filler wets the groove. See why SMAW still wins in so many applications for more on arc-length discipline with stick.
• Ignoring the WPS range on purpose. Running at the top of the amperage band because it is technically in range bakes undercut into a procedure. That upper end usually exists for out-of-position or heavy-tack work, not flat production.
• Weaving without dwell. A fast side-to-side zigzag deposits metal in the center of the joint and starves the toes. The weave needs a momentary pause at each side so filler can flow into the corner.
• Wrong angles and oversized filler. Dragging an oversized rod on a horizontal fillet forces heat input up and control down. Consumable choice matters as much as technique; see our notes on selecting alloy steel electrodes for how filler choice changes puddle behavior.

Prevention Starts on Paper: WPS, pQR, and Qualification

Technique fixes stick only when the paperwork behind them is right. A qualified welding procedure specification backed by a procedure qualification record (pQR) keeps amperage, voltage, travel speed, wire size, gas, and position inside ranges proven to work. Welder qualification tests close the loop: many shops add a fillet-gauge check on the test plate so undercut becomes an explicit pass/fail criterion, and continuity records keep that qualification current.

The cheapest control is a first-article check: gauge the first weld of every shift against the code limit before the rest of the lot runs. Catching a hot setting on weld one saves a dozen repairs on weld twelve.

Technique Fixes by Position

• Flat (1G/1F): the most forgiving position. Stay in the lower half of the amperage range, use a moderate drag angle, and travel slow enough that the puddle reads wide and shallow.
• Horizontal (2F/2G): the top toe is the danger zone because gravity pulls the pool away from it. Drop heat slightly, run stringers rather than wide weaves, and keep the electrode pointed at the upper edge. Our all-position welding rod guide covers the right consumables here.
• Vertical-up (3G/3U): cut current about 10 to 15 percent below flat settings, use short stepped weaves with a real pause at each side, and resist a long continuous weave.
• Overhead (4G/4F): use the smallest effective filler diameter, the lowest current that still fuses, and a tight arc. Overhead undercut usually means too much heat and a hand moving too fast.

Shops working to aerospace tolerances run this as culture, as our guide to aerospace TIG welding standards and training shows.

When Post-Weld Improvement Makes Sense

Two accepted improvements sit alongside repair: toe grinding, where the weld-to-base-metal transition is blended smooth to remove the notch, and toe peening or high-frequency mechanical impact treatment, which introduces compressive residual stress at the toe. Both are legitimate fatigue-life enhancements, and TWI’s review of geometric imperfections makes clear that shape, not just depth, drives the stress-raising effect. Neither is a license to leave nonconforming undercut in place: grinding still must respect code depth and net-section limits, and peening must be specified in the procedure. The 2025 D1.5 bridge welding code favors controlled grinding removal over simply adding weld metal.

Undercut prevention is not complicated: slower travel, a shorter arc, a corrected angle, and a procedure that actually gets followed. The welds that pass the thumbnail test are almost always the ones where somebody resisted finishing five minutes early.

Conclusion: Treat Undercut as a Fatigue Problem, Not a Cosmetic One

Undercut is one of those weld discontinuities that rewards a second look. It hides at the toe or root, it flatters the weld under casual lighting, and it convinces plenty of experienced welders that a joint is finished when it is not. By now you have the full picture: undercut is a defined, measurable, code-controlled groove melted into the base metal beside the weld, and its real danger has almost nothing to do with how the weld photographs. It has to do with fatigue.

The Definition Worth Remembering

Strip away the code tables and the technical language, and the core definition from TWI’s review of geometric shape imperfections stays simple: undercut is a groove melted into the base metal adjacent to the weld toe or root that is not filled by weld metal. It happens on groove welds and fillet welds, on every arc process from stick to submerged arc, in every position. That single notch is a pre-built crack starter, which is why structural codes treat it as a fatigue problem rather than a beauty problem.

What You Learned About Causes and Prevention

Almost every root cause of undercut traces back to four controllable variables: amperage too high, arc voltage too high from a long arc, travel speed too fast, and wrong electrode or torch angle. Add poor weave technique with no dwell at the toes, an oversized filler for the joint, and base metal already running hot from a previous pass, and you have the complete recipe. The encouraging news is that most undercut is preventable, and prevention is cheaper than repair. Holding consistent parameters, dwelling at the toes during a weave, and staying inside your all-position rod selection and technique range will eliminate the majority of undercut before it ever reaches an inspector.

Process choice matters too. Aluminum work brings its own undercut triggers, from heat input sensitivity to oxide contamination, so technique guides like the complete practical guide to aluminum MIG welding and aluminum flux core welding are worth reading before you blame the arc.

Limits Are Contextual, Never Universal

One of the most important takeaways from this article is that there is no single “acceptable” undercut depth. Acceptance depends on the governing code, the loading condition, the material thickness, and the weld length. AWS D1.1 draws a hard line between static and cyclic service, with cyclic joints held to much tighter tolerances. ISO 5817 grades the same discontinuity differently at quality levels B, C, and D. And the 2025 revision of the D1.5 Bridge Welding Code tightened undercut tolerances and shifted the preferred treatment toward grinding. Always check the project specification before you decide a weld is acceptable, because the spec and the drawing govern, not shop habit.

Inspection and Repair: The Short Version

Inspection stays visual more often than people expect. Oblique lighting, a thumbnail run along the toe, and a fillet gauge or depth micrometer will resolve nearly every case. Reserve ultrasonic or radiographic testing for situations where root undercut on a butt joint is visually inaccessible. Measurement errors are usually the result of mistaking spatter or mill scale for a groove, so clean the weld first.

Repair is a two-path decision. Shallow undercut within code depth limits is usually corrected by blend grinding to a smooth transition, which removes the notch without adding a repair bead. Where depth exceeds what grinding alone can fix, the classic route remains a stringer bead with a small-diameter electrode at reduced current, followed by re-inspection to the same acceptance criteria as the original weld. Increasingly, codes prefer grinding over cosmetic repair welds precisely because a poorly executed repair bead can introduce a fresh notch, residual stress, or even lack of fusion right where fatigue cracks start.

Your Undercut Inspection Checklist

Carry this to your next inspection and work through it in order:

1. Identify the joint type and service condition — static or cyclic, and what the governing code requires.
2. Clean the weld toe — remove spatter and loose mill scale before judging anything.
3. Inspect under oblique light — light raking across the toe reveals grooves that frontal light erases.
4. Test with a thumbnail or gauge — confirm the groove is real, then measure depth.
5. Measure depth and length — check against the code limits, including accumulated-length rules.
6. Decide the remedy — blend grinding within limits, repair welding beyond them, or accept where the code allows.
7. Re-inspect after any repair — every repair re-enters the same inspection gate.

Undercut will never be as dramatic as a crack or as obvious as porosity, which is exactly why it causes so many structural failures under cyclic load. Treat it as a fatigue problem from the first pass, and it stops being a rejection problem later.

Frequently Asked Questions

What does undercut mean in welding?

Undercut is a groove melted into the base metal next to the weld toe or weld root and left unfilled by weld metal, per AWS A3.0. It looks like a narrow channel running along the edge of the bead and reduces local thickness.

Is a small amount of undercut acceptable?

Yes, most codes allow shallow undercut. AWS D1.1 permits 1/32 in [1 mm] on statically loaded material under 1 in thick, with limited accumulated length up to 1/16 in [2 mm], while cyclically loaded tension welds are limited to 0.01 in [0.25 mm].

What causes undercut in welding?

The main causes are excessive current, arc voltage that is too high or an arc that is too long, travel speed that is too fast, incorrect electrode or torch angle, and weaving without dwelling at the toes. Arc blow and magnetism can also contribute.

How do you measure undercut depth?

Zero an undercut gauge on the base metal surface beside the weld and set its pointer at the bottom of the groove, then read the depth. For shallow or internal undercut, inspectors may use a depth micrometer or ultrasonic examination.

Can undercut be repaired by grinding alone?

Only when grinding produces a smooth transition and still leaves acceptable depth and section thickness. The 2025 D1.5 Bridge Welding Code prefers grinding for undercut within allowable depth, but requires the net section to stay at least 98% of nominal and deeper cases rewelded.

Why is undercut dangerous for fatigue?

Undercut acts as a sharp notch at the weld toe, raising local stress and giving fatigue cracks an easy initiation site. Research shows deep undercut can cut fatigue strength substantially, which is why codes impose much tighter limits on cyclically loaded welds.

What is the difference between undercut and underfill?

Underfill is the weld simply sitting below the surrounding base metal surface because the joint was not filled enough. Undercut is metal actually melted away at the toe and not refilled, creating a notch rather than just a low profile.

Which inspection method finds undercut?

Surface undercut is found by visual inspection using oblique lighting and a thumbnail check, then measured with an undercut gauge. Undercut at the root of a groove weld that cannot be seen requires ultrasonic or radiographic examination.