How to Prevent Undercut in Welding: Causes, Fixes, Code Limits

A fabricator’s inspector set aside a fillet weld the welder was proud of: the bead was uniform, the color was clean, and the whole thing looked sound. But along one toe, a hairline groove had eaten into the steel, and under the acceptance limits in AWS D1.1 it was a reject. That groove is called undercut, and it is the most fixable weld defect most welders will ever meet.

I’ll rewrite this section to hit the 600–900 word target while keeping the heading, facts, and all links intact.

What Welding Undercut Actually Is (and Why Inspectors Care)

Picture a fillet weld that looks textbook: even bead, clean color. The welder reaches for the grinder when the inspector tilts the part under a light and runs a finger along the toe. Base metal has been eaten away next to the weld. Measured against a straightedge, it is a reject under AWS D1.1.

That groove is called undercut, the most fixable defect most welders will hit. It also sneaks past most often, because the part still feels solid.

The Precise Definition

ESAB’s welding defects guide defines undercut as a groove melted into the base metal at the weld toe. You have removed material, so the joint is smaller than the drawing, and cut a notch where the weld transfers load into the base metal. That notch is a stress raiser that eats fatigue life, so inspectors treat undercut as a real defect.

On a fillet weld, the weld metal sits in the corner where two plates meet. The leg is the weld face along the plate; the toe is where it meets the base metal. Undercut usually eats the toe of the larger plate.

The Vocabulary You Need

Undercut comes in a few forms, and which one you see tells you a lot about the cause:

• Toe undercut. The classic: a groove melted along the weld toe on the base metal surface. This is the one that gets caught.
• Root undercut. A groove at the root of a butt weld, at or above the joint bottom, driven by heat in the joint, not arc angle.
• Continuous undercut. A groove along the full bead, usually a steady parameter problem.
• Intermittent undercut. Scattered pits and notches along an acceptable bead: unstable arc, inconsistent shielding, or a moving joint.
• Interpass undercut. Appearing between passes in a multipass weld, where the welder grinds or cleans slag and repeats the habit.

ESAB lists the common weld discontinuities as porosity, lack of fusion, lack of penetration, inclusions, cracks, undercut, spatter, cold roll, arc strikes and distortion. It shares a shelf with far more dangerous defects, which is why it is treated as a lesser problem. It is not one. The Michigan Tech LTAP welding discontinuities training material treats undercut as routinely seen and notes many such defects are correctable with procedure changes.

How Undercut Differs From Its Neighbours

Three other conditions get confused with undercut, and telling them apart saves grinding.

Underfill is a depression in the weld face itself: the weld metal sits below the plate surface while the base metal stays intact. Undercut is the opposite: base metal missing, yet the weld face looks full and proud. Check the base metal, not the bead.

A crack is a fracture with sharp edges, in the weld or the heat-affected zone. Undercut is a melted groove with a rounded bottom and a shadow line. A crack is always a reject; undercut can be within limits.

Lack of fusion is where weld metal failed to melt and bond with the base metal. It looks crack-like but is buried, while undercut is open to the surface.

Why Inspectors Catch It at the Worst Moment

That open surface is why undercut gets caught late. It is a surface-breaking discontinuity, so an inspector finds it with good lighting and a straightedge, while buried defects like lack of fusion or internal porosity need ultrasonic or radiographic testing to expose. Visual inspection is the primary method, and the AWS D1.1 structural welding code has acceptance criteria for conditions like this.

Where It Shows Up Most

The usual trouble spots:

• Fillet welds where plates of different thickness meet; the toe of the larger plate is the classic spot.
• The toe of a butt weld, where reinforcement meets the plate surface.
• Multipass joints, where every pass repeats the mistake after grinding between runs.
• Vertical-up work, where gravity drags the puddle to the lower toe.
• Wide, high-heat fillets, where the bead outruns the filler.

A Two-Minute Self-Check

You can find undercut on your own work. Run this before every weld cools or gets dressed:

1. Look along the toe with raking light. A groove shows as a dark line that changes as you move your head. Fastest method; catches most toe undercut.
2. Run a straightedge along the leg so it bridges the toe, then look for a gap or light under the edge. Rocking means material is missing.
3. Use a weld gauge on the bad ones to measure how deep the groove is, then compare it with the limit in your code.
4. Check the worst point, not the average. Acceptance limits apply to the deepest point of the undercut.
5. Do it before grinding. Once the toe is blended smooth, you have destroyed the evidence.

Once you know a groove in the base metal at the toe is undercut, the next question is why it happens and what it costs. The rest of this guide starts there, with the advantages of stick welding and why the answer changes with your process.

Why Undercut Matters: Strength, Fatigue Life, Corrosion and Cost

It’s easy for many welders to dismiss a thin groove at the toe of a fillet weld as nothing more than a cosmetic flaw. The truth, however, is that this small notch is doing far more damage than it appears to. It’s removing material at the exact spot where the weld is meant to transfer load, and that alone makes it worth taking seriously.

The Load Path and Stress Concentration

Every fillet weld has a job: transfer stress from one piece of base metal to the other. That stress flows through the weld throat and into the base metal at the weld toes. When undercut forms there, it removes real section thickness from that load path. Even more critically, the sharp transition of that toe groove acts as a geometric stress concentration at that corner.

As noted by ESAB’s welding defects guide, undercut reduces the effective cross-section of the joint and introduces a stress raiser that harms fatigue life. Most importantly, ESAB explicitly states that although undercut is sometimes viewed as cosmetic, it can be structurally significant in dynamically loaded joints. This is especially true for structures that see constant movement, vibration, or changing loads, such as those found in marine fabrication.

Fatigue: The Notch That Becomes a Crack Starter

For a joint under one, steady load, a small amount of undercut may not be an immediate failure point. For cyclic loads, it’s a completely different story. Each time the joint is loaded and unloaded, the stresses at that notch multiply far beyond what the surrounding metal sees. Simply put, cyclic loads find the notch, and the notch becomes a crack starter.

From there, a fatigue crack can grow little by little with each cycle until it reaches a critical size. The biggest factor for this is orientation. An undercut transverse to the primary tensile stress is by far the worst case, whilst an undercut parallel to the principal stress is far less serious. That small detail can be the difference between a minor repair and a catastrophic failure.

Corrosion: A Hidden Crevice

The mechanical risk isn’t the only one to worry about. A groove at the weld toe creates a tight, hard to clean crevice. Moisture, dirt, and road salts love to settle inside of it and stay trapped there, unable to fully dry out. For coated or painted structures, this becomes a coating breakdown point. Instead of corrosion starting out on an open surface, it can start directly at that undercut. Once rust begins in that notch, not only does it eat away at the already weakened section, but it also adds more stress to the area, fueling the fatigue risk even further.

The True Cost of Repair

All of these risks come with a price tag. Repairing undercut isn’t just laying down an extra pass of weld. It means grinding, gouging, re-welding, re-inspection, possible NDT, and the schedule and material cost of the rework itself. On top of that, if the job involves stainless steel, grinding that material adds its own fume exposure to the bill. For many shops, the labor hours lost for hold points and rework will far outweigh the seconds that could’ve been saved by preventing it in the first place.

At the end of the day, undercut isn’t something to shrug off. It’s a loss of section, a stress riser, a corrosion trap, and a direct hit to both safety and profit. With that in mind, every effort put towards preventing it is an investment that pays for itself.

The Root Causes of Undercut, Explained

Before reaching for a grinder, it’s important to understand one thing: undercut isn’t a random flaw. It’s a symptom. More often than not, the arc is melting away more base metal than your filler metal can replace. To fix it for good, you need to diagnose the cause, not just cover up the groove.

The Core Causes

Two industry sources align almost perfectly when it comes to the most common culprits. According to Michigan Tech LTAP, the main causes are welding parameters set too high, inadequate filler metal addition, incorrect work angle, and travel speed that is too fast. ESAB expands on this, listing excess current or voltage, long arc length, a steep torch or electrode angle, and high travel speed. All of these boil down to the same physical problem: too much heat, not enough replacement.

• Excess current or voltage: Cranking up your amperage or voltage dumps an immense amount of heat into the joint. The arc begins to dig into the base metal edges, and if you aren’t feeding enough filler at the same time, there is simply no material to fill that gouge.
• Long arc length: Stretching your arc is one of the easiest mistakes to make, and one of the most destructive. A longer arc raises voltage, widens the heat spread, and becomes far more erratic. This causes the arc to wander away from the puddle and melt the thin toe of the joint instead of focusing its heat where your filler can reach.
• Steep torch or electrode angle: Your angle dictates where the arc force goes. If your work angle is too steep, that force is driven directly into one toe rather than across the fillet. The result is that one edge melts drastically faster than the filler can flow into it, carving undercut whilst the rest of the bead may still look deceptively full.
• Travel speed too fast: The puddle needs time to wet out to the toes and flow into the corner. Move too quickly and it freezes before it can do so. With nothing left for it to bond to, the melted base metal edge is left behind as a groove.
• Inadequate filler metal addition: Heat and angle can be spot on, but without enough filler, the ratio will always be off. Whether it’s a lack of wire feed in GMAW or inconsistent filler rod cadence in SMAW, if you aren’t replacing what is melted away, undercut will always be a possibility. This is especially true when working with finicky alloys such as those covered in our guide to aluminum MIG welding.

Root Undercut in Butt Welds

Not every undercut forms at a fillet toe. Root undercut forms at the bottom of a butt joint, and it’s typically driven by a different set of factors. Excess heat concentrated directly at the root is the biggest offender, however this can be made much worse by a bad root face, an oversized root gap, or an arc that wanders off of the root line. Unlike a toe groove, root undercut is usually hidden until it’s viewed from the back side or caught by inspection, making it all the more important to control your heat at the bottom of the joint.

Shielding, Cleanliness, and Fit-Up

Even with perfect machine settings, the environment around your joint can throw all of that balance off. Dirt, rust, paint, oil, and moisture will all create an unstable, erratic arc, and an unstable arc is what walks heat directly into the base metal edges. This can be amplified by shielding issues, such as incorrect gas type or flow, hose leaks, or wind blowing across an open job.

Fit-up is just as crucial. Mismatched or offset joint edges, an excessive root gap, and joint movement during welding all make it incredibly difficult for filler to reach the edges cleanly. If your plates aren’t clamped tight or are pulling apart as you weld, the puddle has nowhere to wet out to, forcing the arc to gouge the edge in an attempt to compensate. For work that requires an immaculate surface, utilizing a proper aluminum cleaner for welding prior to striking an arc can go a long way in keeping that arc stable to begin with.

Preventing Undercut with Parameters: Current, Arc Length and Travel Speed

Parameters are the part of the problem you can actually measure, so they are the first place to look. ESAB’s prevention guidance for undercut is compact: reduce current and arc length, aim filler into the toes, slow the travel speed, and use a controlled weave. Angle is handled elsewhere because it is a craft skill rather than a machine setting. The other four levers live on the machine and in your hands, and many technique problems are really setting problems.

1. Arc Length and Voltage

A long arc is the quietest way to create undercut, because nothing looks wrong until the toe is already eaten. Extend the electrode or torch away from the work and arc voltage rises, widening and stiffening the pool so the arc wanders off the joint and strikes the edge of the base metal.

The working rule is old and simple: keep arc length roughly proportional to electrode diameter. On GTAW, arc length sets how precisely you can focus the pool. On GMAW, a long stick-out lets the shielding gas spread out and the arc wander off the joint.

2. Current

Too much current melts base metal faster than filler can replace it. Undercut is just the difference between what the puddle ate and what the filler put back. Lower amperage reduces melting at the toe and shrinks the pool, making it easier to control. The catch is that current drives penetration and fusion, so drop it too far and you create a different defect.

Think in ratios: heat input per unit length is roughly (arc voltage × amperage × 60) divided by travel speed in inches per minute. Trim the amperage a notch, or slow the travel, and heat input falls with it.

3. Travel Speed

If the puddle freezes before it flows into the toes, undercut is the result even on a perfectly aimed arc. Slow the travel speed and the molten metal gets more time to reach the corners of the joint before it solidifies. Michigan Tech’s guidance on weld defects is direct about this: slow the travel speed and make sure enough filler metal is added to replace the base metal that has been melted away. Slow down without adding filler and you just build a bigger version of the same problem.

4. Filler Addition Rate

Undercut is a balance between melting and filling, so watch the filler rate as closely as the machine settings. In GMAW and FCAW, match the wire feed speed to the amperage you are running. If the wire feeds slower than the current demands, the puddle consumes base metal that nothing replaces.

With SMAW the equivalent is manual rod cadence, where stick welding quietly earns its reputation for control. If you are new to running a low-amperage stringer, the advantages of stick welding explain why that control is easier with a consumable than with a bare wire. Dip the rod into the leading edge of the puddle and let the heat of the puddle melt the rod rather than melting it with the arc. Rod diameter has to match the amperage, and the all-position welding rod guide covers sizes and applications. A rod too large for your current cannot be laid into the toe cleanly.

5. Weave

Where the joint is wider than your puddle, a weave is the only way to fill it, and that is where toe undercut starts. Use a narrow, controlled weave and add a slight pause at each toe. That pause is the whole trick: it holds the arc over the corner long enough for filler to flow into the edge. A wide, fast weave does the opposite: arc time spent in the middle of the joint and both toes left cold.

The Balancing Act: Do Not Trade Undercut for Lack of Fusion

Every lever here trades one defect for another. ESAB links lack of fusion and lack of penetration to low current or heat input combined with excessive travel speed, an incorrect torch or electrode angle, and a long arc length. Cut the heat too far and you may stop melting the base metal altogether, which is worse than a shallow groove: that defect stays buried and invisible until NDT finds it.

Parameter adjustments must stay inside the ranges set by your qualified WPS. AWS D1.1/D1.1M ties production welding to qualified procedures, and those ranges come from the supporting PQR. If your qualified procedure produces undercut, that is a signal to fix the procedure and requalify it, not to experiment on the work.

Torch Angle, Travel Angle and the Art of Filling the Toes

With the procedure qualified and the machine set, angles are the fastest lever left, and the one thing no machine adjusts for you: where the arc is aimed, how it is dragged, and how you steer molten metal into the toes.

Worth remembering: ESAB lists a steep torch or electrode angle among four direct causes of undercut, alongside excess current or voltage, long arc length and high travel speed.

Work Angle: Where the Torch Points at the Joint

The work angle is the angle between the torch or electrode and the surface of the work, the plan view of your aim.

On a fillet weld the classic reference is roughly 45 degrees, where arc force, filler feed and puddle flow cooperate, splitting the joint evenly so the arc shares heat between both toes.

Tilt steeper, aiming into the near toe, and the arc concentrates its heat on one side. That toe melts faster than your filler can replace the metal, and a groove opens where the arc was pointing, the most common way good parameters turn into an undercut bead.

Too flat is just as bad: hold the torch nearly parallel to the work and the arc spreads out, the puddle goes wide and shallow, and root penetration suffers. You trade undercut for lack of penetration or lack of fusion, both onMichigan Tech’s LTAP list of classic weld defects.

Travel Angle: Push, Pull and Drag

The travel angle is the angle in the vertical plane between the torch and the direction the joint runs. This is where push versus pull lives.

A modest push of roughly 10 to 15 degrees is the usual reference. Holding the torch slightly behind the direction of travel pushes the puddle ahead of the arc, improving penetration and driving metal into the toes.

Swing the torch way forward, past 45 degrees of push, and you lose toe coverage and root penetration, because the puddle is shoved away from the corners: an undercut toe with a poorly fused root. ESAB notes that poor manipulation, which fails to wash metal into the toes, turns good parameters into a visibly undercut bead, and bad angles are a big part of it.

The SMAW Drag Angle

Stick welding adds a wrinkle: with a covered electrode the classic technique is a drag angle, tilting the electrode 10 to 15 degrees toward the direction of travel so the trailing edge points back along the weld and arc force pushes the puddle forward.

Reverse it and the arc digs. The electrode gouges the puddle, carving a crater that fills behind you as a hole rather than solid metal. The same tilt also stops the arc wandering off the work and undercutting the toe out of position.

Newcomers to all-position stick welding learn the drag angle first: if the electrode leans back into the puddle instead of ahead of it, fix the angle.

Toe Washing: The Habit That Saves the Bead

Even with perfect angles you still have to steer the arc into the corners deliberately. This is toe washing, and it is the most useful manipulation habit a welder can build.

1. Start at the center of the joint and confirm your heat and speed.
2. Move the arc toward one toe, deliberately, not a fast sweep.
3. Dwell briefly at the toe so the filler flows into the corner before the puddle freezes. This pause is where undercut is prevented or created.
4. Return to the center and repeat toward the other toe.
5. Close the arc slightly at each toe so the heat stays where you put it.

The dwell is what people skip. A weld that undercut every time can often be fixed on the next bead by adding nothing but a one or two count pause at each toe. For a wider look at the process, the case for SMAW is worth a read.

Vertical-Up: The Hardest Case

Vertical-up tests all of this at once, and it is the position most associated with undercut for a physical reason: gravity pulls molten metal toward the lower toe, so the puddle runs downhill and metal runs out of that corner instead of into it.

• Keep the puddle small so it freezes fast enough for you to control where metal goes.
• Use stringer beads or a very slight weave. A wide weave enlarges the puddle and gives gravity more room.
• Pause at each toe so filler flows into the corner before you move on.
• Watch the heat, not just the appearance. Undercut in vertical-up usually shows mid-bead, where the welder was rushing.

With latitude in your procedure, a slower travel speed and more filler at the toes beats any parameter change. And a vertical-up test is the hardest joint to pass visually, which is why AWS qualification and acceptance criteria are quoted separately for it.

Angles do not change how much heat goes into the joint, but they decide where that heat lands and whether the filler keeps up. Hold the work angle near 45 degrees on fillets, keep push or drag at 10 to 15 degrees, and wash the toes on purpose.

Consumables, Shielding Gas, Joint Preparation and Shop Conditions

Welders obsess over the arc, current and travel speed, and rightly so. But undercut is often created by what surrounds the arc: a stick-out that is too long, a leaking hose, a dirty edge that steals heat, or a plate too cold for the puddle to flow.

Consumables: size, stick-out and diameter

The most common consumable mistake is an electrode or wire too large for the plate. A fat electrode melts a wide puddle that spreads instead of hugging the toe. Follow the rule of thumb: roughly a 1/8 in electrode for plate around 3 mm thick, and 5/32 in for 5 to 6 mm plate, stepping up for thicker plate.

Stick-out matters just as much. In GMAW, an over-long stick-out preheats the wire and steals heat from the base metal, so the puddle arrives cooler and less fluid at the toe. Hold stick-out consistent from start to end, and shorten it as you climb to the thicker side of a dissimilar joint.

Consumable condition matters too. Damp electrodes, contaminated wire and a spatter-clogged gun liner make the arc erratic, and an erratic arc wanders into the toe. The same logic applies to wire choice for stainless or aluminum, as described in our guide to alloy steel welding electrode selection.

Shielding gas: type, flow and leaks

Wrong gas makes the arc restless, and a restless arc digs holes. According to ESAB’s welding defects guide, incorrect gas type or flow, gas line leaks and a long arc length drive porosity and erratic arc behaviour, and each destabilises the arc the way undercut does: it wanders off centre, hesitates near the toe and melts a groove the filler never reaches. ESAB also recommends Ar/CO₂ blends and pulsed or spray transfer where the material allows, and notes that pure CO₂ in MIG/MAG throws more spatter.

The process choice matters for fume, too. An AIHA summary of NIOSH research reports that the best GMAW processes, such as pulsed spray, produce roughly 1/40 the fume emission rate of SMAW, and some GMAW choices cut hexavalent chromium emissions to as little as 1/250 of SMAW. Before switching anything, read our notes on gas and cylinder supply considerations.

Joint preparation: clean steel, controlled geometry, no movement

Undercut loves dirty edges. Scale, rust, paint, grease and embedded water steal heat from the arc and force you to compensate with more current or slower travel, which is how you melt the toe. Clean every joint face back to bright metal first.

Control three things:

• Root face: too thick and the root melts before the filler reaches it, driving root undercut; too thin and you get lack of penetration.
• Root gap: an oversized gap forces the arc to bridge a wider span at the bottom, concentrating heat at the root.
• Bevel angle: a wide bevel creates a deep V the puddle fills from the bottom, slowing travel and overheating the toes.

Then clamp the work. Joint movement is a quiet cause of undercut: as plates shift under the arc, the toe moves mid-bead and the puddle is pulled off the corner. Strongbacks and clamps keep the joint where you left it. For aluminum, prep matters even more, as covered in how to prep aluminum for welding.

Shop conditions: cold plate, wind and damp

Base metal temperature changes how the puddle behaves. Cold plate, wind chill and dampness pull heat away faster, so the puddle stiffens and stops flowing into the toes. Preheat is not just a code requirement, it is a practical undercut control: it thins the puddle and keeps it fluid enough to wet the toe.

Wind is the other enemy. Shielding gas gets blown off the puddle, the arc goes unstable, and spatter and porosity follow. A simple frame and a welding blanket beat an afternoon of welding a bad bead in the open.

Safety: grinding and repair add fume exposure

Fixing undercut means grinding, gouging and re-welding, and each step puts the welder back near the fume plume. On stainless and chrome-bearing alloys that means hexavalent chromium exposure.

OSHA 29 CFR 1910.1026 sets a permissible exposure limit of 5 micrograms per cubic meter for chromium (VI) as an 8-hour TWA, with an action level of 2.5 micrograms per cubic meter, and requires engineering and work practice controls first, with respirators supplementing them. 29 CFR 1910.252(c) requires local exhaust or general ventilation to keep toxic fumes, gases and dusts below the maximum allowable concentration. A NIOSH evaluation of welding operations reported total fume of 2 to 60 mg/m³ without local exhaust ventilation and 3 to 13 mg/m³ with it.

None of this replaces good technique, but it makes good technique possible: clean metal, correct consumable size, stable stick-out, sound shielding gas, a prepared and clamped joint and a warm plate take most of the guesswork out of the toe.

Process by Process: Where Undercut Sneaks In (SMAW, GMAW, FCAW, TIG, SAW)

The cure for undercut is not universal. The same 0.25 mm groove a skilled TIG welder clears on a stainless tank can be a daily fight on a 10 mm fillet run with a short-circuit GMAW gun. Each process melts, fills and freezes metal differently, so a trigger that ruins one joint may be irrelevant to the next.

Process	Typical undercut trigger	Best preventive action
SMAW (stick)	Manual filler addition at the toes; repeat chance in multi-pass cleanup	Toe-wash deliberately with the rod; slag removal and grinding between passes
GMAW (MIG/MAG)	Short-circuit transfer, fast-freezing pool, travel speed leaving toes cold	Slow travel, match wire feed to the puddle, use spray or pulsed transfer for heat control
FCAW (flux core)	High inductance and long stick-out with a large hot pool	Tune inductance to the wire, keep stick-out consistent, prefer gas-shielded flux core if fume is a concern
TIG (GTAW)	Arc wandering at the root of a butt weld rather than at the toe	Place filler at the root, keep a slow deliberate arc angle, let the arc dwell in the corner
SAW (submerged arc)	Toe undercut uncommon with the large fluid pool; process control dominates	Hold the parameter window steady, manage flux delivery and drying, keep travel speed low

SMAW: the filler rod is the weak link

Stick welding has no wire feeder to protect the toes, so the weld is only as good as the welder’s rod cadence. If the puddle reaches the corner before the rod does, base metal melts faster than filler can replace it and you get a groove. Walk the arc toward each toe, add filler there first, then come back to the center of the bead. Multi-pass work multiplies the risk, because grinding between passes creates repeated chances at the same mistake. SMAW is also a cost problem in the fume literature: AIHA’s summary of NIOSH research notes that multiple passes are often required for SMAW welds, further increasing the cost of the job. For shops that value the simplicity of stick, the trade-offs are covered in why SMAW still wins.

GMAW: transfer mode decides everything

MIG and MAG come in four common transfer modes, and each behaves differently at the toes. Short-circuit transfer gives a small pool that freezes almost instantly, so at high travel speed on a fillet the puddle never reaches the toes. Globular sits between the two. Axial spray and pulsed spray deliver better heat control, less spatter and a longer, more fluid pool, making them the go-to for fillet work where toe coverage matters. Spatter is also a diagnostic clue: ESAB’s defect guide attributes heavy GMAW spatter to excess current or voltage, long arc length, pure CO2 shielding, unstable short-circuit transfer and low inductance. All five are parameter-window problems, and a machine with the wrong inductance and shielding gas can undercut all day long. Get the gas and equipment right before blaming the welder; the welding equipment guide covers the hardware side.

FCAW: a bigger pool, different problems

Flux core wire gives good fillet capability and a larger, more forgiving puddle than solid wire, so toe undercut is generally less of an issue. Two things still catch people out: inductance settings that make the wire sluggish or harsh, and excessive stick-out, which starves the base metal of heat at the toe. Self-shielded flux core also adds a significant fume load, one reason gas-shielded flux core is preferred where the code allows.

TIG: thinnest pool, root problem

GTAW produces very little fume and an extremely thin, controllable puddle, but demands the most skill. The characteristic TIG undercut is not a toe groove at all: it appears at the root of a butt weld where the arc wanders at the bottom of the joint and eats the root reinforcement. The cure is root-level filler placement plus a slow, deliberate arc angle, building the root before you travel. Demanding TIG specifications are covered in aerospace TIG welding.

SAW: uncommon at the toe

Submerged arc welding runs large, fluid pools at high heat input, so toe undercut is comparatively rare. Process control dominates instead: consistent travel speed and voltage, correct flux delivery and drying, and attention to arc position changes around the joint.

Pick the row that matches your machine and change one thing at a time: a fix for the toe groove on one process usually creates a different defect on another.

Code Limits: What Undercut Is Actually Allowed to Be

Codes don’t prevent undercut from happening. They sort it after the fact into acceptable or not, and anything over the limit must be corrected before a job can ship. That line in the sand changes based off of the code you’re working to, the load on the connection, and the contract that you’re bound to.

AWS D1.1/D1.1M: The Structural Steel Standard

For most structural steel fabrication, acceptance for undercut falls under Table 8.1 of AWS D1.1/D1.1M. At the time of writing, the listed edition is AWS D1.1/D1.1M:2025-AMD1, however the edition named within your contract documents is always the one that governs.

The limits are directly tied to stress orientation and base metal thickness:

Situation	Maximum undercut
Undercut transverse to the primary (tensile) stress in the part	0.010 in (0.25 mm)
Other cases, where the base metal thickness is under 1 in	1/32 in (0.8 mm)
Base metal 1 in thick or thicker, for any length of weld	1/16 in (about 2 mm)

The 0.010 in (0.25 mm) allowance is by far the tightest for a reason. When a groove runs transverse to the primary tensile stress, it sits directly in the load path. The 1/32 in (0.8 mm) limit covers most welds that run parallel to that stress on material under 1 in. For base metal 1 in or thicker, the allowance opens up to 1/16 in (about 2 mm) for any length of weld. This is less leniency and more practicality, as holding a clean toe on heavy plate can be much harder to achieve.

Static vs. Cyclic Loading

Not every connection sees the same demands. A statically loaded structure, such as a storage rack or mezzanine, can be held to its applicable limit, however cyclically loaded connections are the ones that call for the tightest scrutiny. Bridges, cranes, and fatigue sensitive machinery bases all fall into this category.

This is due to fatigue. Under millions of load cycles, a notch doesn’t need to be deep to initiate a crack. For this reason, any undercut transverse to the primary tensile stress on a cyclic connection must still fall within the 0.010 in (0.25 mm) limit. Most importantly, the loading case is a design input, not an inspection detail. It needs to be established before inspection begins, otherwise a weld could be incorrectly accepted for its intended service.

ASME B31.3 and International Codes

If you’re welding process piping, AWS D1.1 is the wrong book entirely. ASME B31.3 carries its own visual examination acceptance criteria for piping welds. These limits are tailored to internal pressure, thermal cycling, vibration, and potential corrosion from the inside of the pipe, all of which can be far more punishing than many structural applications. Never carry numbers from one code over to the other, as they are not interchangeable.

For European and many international jobs, ISO 5817 is used insteadof a single number. Instead, it defines quality levels for welded joints, from B (highest) down through C and D, with different allowable defect limits at each level.

The contract, welding procedure specification, or drawing will state which level applies, and that choice is what determines just how much undercut is allowed. As ESAB notes in its welding defects guide, international standards including ISO 5817, AWS D1.1, and ASME Section IX all define their own terminology and acceptance levels. With that being said, critical industries such as nuclear, aerospace, and high-pressure applications frequently go beyond these minimums and require defects to be eliminated entirely.

At the end of the day, “acceptable” isn’t a fixed property of a weld. It’s a function of the code, the edition, the quality level, the service, and the contract.

The Caveats That Actually Matter

• Codes define acceptance, they do not prevent defects. Nothing in any of these documents will stop a welder from running too hot. The code tells you what to do with the result.
• Your contract specification always wins. When the project documents specify a tighter limit than the base code, use the tighter limit. Owner requirements routinely supersede them.
• A shop standard is not a code standard. Many shops publish internal weld criteria. Treat them as a floor for training and consistency, never as evidence that a weld meets AWS, ASME, or ISO requirements.
• Check the edition. The edition cited within the contract documents is the governing one, not necessarily the most recent revision.
• Repair provisions still apply. Acceptable-looking welds still have to meet minimum weld size and throat requirements, and any undercut that exceeds the limit must be excavated, repaired, re-inspected, and documented.

Understanding these acceptance criteria is only half of the job. The first half is preventing the groove in the first place, which is where the parameters, torch angle, and filler discipline come in.

How to Inspect, Measure and Repair Undercut

Detection: Finding it Where it Matters

Because undercut breaks the surface, visual inspection is always your first and most reliable tool. ESAB’s welding defects guide lists visual inspection as the method that reveals surface defects such as undercut, porosity, spatter and cracks, with UT, RT, MT and PT reserved for subsurface indications. If a job asks for NDT specifically on undercut, you are probably chasing a separate indication.

Inspect the toes, not just the face. Angle the light across the weld so it casts a shadow into any groove running along the toe, and pay extra attention to fillet welds, butt weld toes and welds made out of position. On critical work such as alloy wheel welding, toe inspection should never be rushed. A shallow notch is easy to miss under flat overhead light.

Measuring Against the Code

Depth decides acceptability, not how noticeable the groove looks. Lay a straightedge across the adjacent base metal and measure to the deepest point with a weld gauge or a depth gauge, then compare that reading against the acceptance criteria in the code governing your work. The edition named in the contract always wins, so check the current version of AWS standards for structural steel or the applicable piping and fabrication code. Codes do not exist to stop undercut; they define when it must be repaired.

Repair: Grind, Excavate, and Refill Correctly

Not every acceptable undercut needs repair, but once a groove exceeds the governing limit it must be corrected properly. Michigan Tech LTAP’s welding discontinuities guidance says to grind the affected area slightly to create a uniform groove and then add weld, while ESAB lists depositing corrective toe runs to refill the groove, blending it smoothly and re-qualifying the parameters so it does not recur. Between the two you have a workable method: standardize the groove, fill it properly, and change the settings that produced it.

The Rules a Repair Still Has to Satisfy

Repairing a groove is not a free pass. Under AWS D1.1/D1.1M clause 8.25, a defective weld must be excavated out to sound metal before it is reweld, and the repaired weld is then subject to the same acceptance criteria as the original. That last part is where a lot of repairs fail on paper.

A weld can look perfect and still be a reject if it is under-size. AWS D1.1 Table 7.7 sets minimum fillet weld sizes, so a repair that gouges out too much metal can leave the joint under-size even after the groove is full. Measure the throat and leg after grinding, not before, and confirm the remaining section is still strong enough to accept the repair load.

Excavation can be done by hand with a grinder for small areas, or with carbon arc gouging for deeper and wider removals. Gouging is fast and leaves a clean, well-defined cavity the following pass can key into, though it is mostly a carbon and low-alloy steel tool rather than a stainless one.

How to Avoid the Classic Repair Mistake

The most common failed repair is grinding a groove and dropping a fat, sloppy pass over the top. It fills the notch visually, leaves a lump where a fillet used to be, and often leaves the real problem untouched underneath or just beside it. A sound repair follows a fixed sequence:

1. Grind to a uniform groove. Shape the defect into a clean V or U running the full length of the undercut, with smooth transitions into sound metal. No pits, no feathered edges the new puddle cannot wet.
2. Clean it out. Remove all grinding dust, slag and oxides. Contamination at the toe is a leading cause of the original undercut, and it will fail the repair the same way.
3. Weld a controlled toe run. Short stringers with a slight pause at each toe beat a fast wide weave. SMAW gives the fine control needed for a short, precise toe run without the heat input of a heavy flux core pass.
4. Blend and document. Blend the repair into the surrounding weld and base metal, re-inspect it under the code’s repair provisions and record what was done.

Then change the parameters that caused the defect. Re-welding with the same current, angle and travel speed that produced the undercut means you have not repaired anything, you have just paid for the same mistake twice.

Troubleshooting by Symptom, Common Mistakes, and a Prevention Program That Holds

Most undercut defects announce themselves. By the time a groove appears at the toe, the settings, fit-up and torch position behind it have already told a story. Read that story before reaching for the grinder. Michigan Tech’s welding defects guidance makes the same point: correcting a defect usually comes down to adjustments in technique and confirming the procedure is actually followed on the floor.

Match the Symptom to the Cause

Use the pattern of the defect, not just its size. Undercut is consistent enough that the location and rhythm of the groove usually narrow the cause fast.

What you see	What is usually happening	First move
Undercut only along the top toe of a fillet	Work or travel angle is off, dumping extra heat onto that side of the joint	Return the work angle toward roughly 45 degrees, trim current slightly, and dwell at the top toe
Undercut only at the bottom toe on vertical-up work	Gravity pulls the puddle downhill and filler arrives too late to refill the corner	Keep the puddle small with stringer beads, add filler earlier, pause at the bottom toe
Undercut comes and goes along an otherwise good bead	Unstable shielding, a wandering arc, or joint movement during welding	Check gas flow and hose leaks, kill drafts, re-clamp the joint so it cannot move
Undercut right next to heavy spatter	Parameter window too hot, or shielding gas wrong for the transfer mode	Back off current and voltage, verify gas and flow rate, confirm the transfer mode
Undercut on the first pass only, clean passes after	Excess heat at the start, dirty or cold steel, no preheat	Clean to bright metal, verify preheat per the WPS, control first-pass heat input
Root undercut in a butt weld with no toe issue	Too much heat at the root, a poor root face, an oversized root gap, or a wandering arc	Control root-pass heat, fix fit-up and root gap, hold a slow, deliberate arc angle

Two details tighten the diagnosis. First, run the pattern against the job: a groove on structural fillets but never on pipe points to technique, not machinery. Second, check what came out of the process. If the arc sputtered and the bead had rough edges, suspect shielding and parameters before blaming the welder’s hands.

The Classic Mistakes

These errors show up again and again in defect reviews, and they travel together:

• Current and voltage set too high. More heat than the filler can replace melts the toe faster than it can be filled.
• An overly long arc. Longer arc means higher voltage, a wider, stiffer pool, and an arc that wanders off the toes.
• Travel speed too fast. The puddle freezes before it can flow into the corners.
• An oversized electrode or wire. Too much filler capacity for the joint and too much heat at once.
• No filler discipline. Filler added late, or in one lump instead of steadily through the toe.
• Dirty, rusty or oily joint surfaces. These steal heat, upset the arc and prevent wetting at the toe.
• Skipping the toe wash. The arc never visits the corner, so the corner never gets filled.

Several of these sit in the manual process, where technique carries the most weight. Running SMAW in the field, the case for stick welding and its control advantages are worth weighing against the discipline it demands. Choosing consumables that behave in every position is covered in the all position welding rod guide.

A Prevention Program That Holds

Individual welders fix their own grooves. Programs fix shop-wide trends. Six elements separate a shop that repeats the same undercut from one that stops making it.

1. Enforce qualified procedures and qualified welders. The AWS standards and credentialing framework includes Welder Performance Qualification (WPQ) and Welding Procedure Qualification (WPQ2) pathways, because codes assume welds are made by qualified welders using qualified procedures. If the job runs on an unqualified setting, the defect was guaranteed before the arc struck.
2. Run mock-up coupons before production. A test joint on scrap plate, made by the actual welder at the actual settings, exposes a bad angle or a hot parameter set before it reaches a customer.
3. Self-inspect while the bead is still hot. A toe check two hours later is an inspection. A toe check as you release the torch is a correction. Checking the toe as the bead cools, then adjusting at once, prevents most rework.
4. Hold a first-off inspection point. On repeat work, the first weld of a run gets inspected before the rest is made, so a wrong setting costs one weld instead of fifty.
5. Track causes, not counts. “Twelve undercuts this month” tells you nothing. “Twelve undercuts, all at the top toe, all from the same welder on 10 mm plate” tells you exactly what to fix. The AIHA source-reduction work makes the same economic point: multi-pass SMAW work with slag removal and grinding between passes adds cost and rework, so cutting defect-driven rework lowers shop cost.
6. Re-qualify parameters after any change. New wire diameter, a different gas cylinder, a revised fit-up or a changed machine all shift the window. Treat them as new conditions, not tweaks to an old habit.

Inspections only work if someone is qualified to sign off. That is why AWS CWI and Senior CWI certification pathways exist: inspection quality depends on documented, qualified competence rather than opinion.

That leaves the last piece: a repeatable check any welder can run on their own work. The closing checklist pulls everything in this article into a short, shop-floor routine that takes under a minute per joint and catches the defect before an inspector does.

Frequently Asked Questions

What is the acceptable amount of undercut in welding?

It depends on the code and the loading case. Published summaries of AWS D1.1 Table 8.1 give 0.010 in (0.25 mm) maximum where the weld is transverse to the primary tensile stress, 1/32 in (0.8 mm) for other cases on base metal under 1 in thick, and 1/16 in (about 2 mm) for base metal 1 in and thicker. Always confirm the edition named in your contract documents, since the AWS standards page currently lists AWS D1.1/D1.1M:2025-AMD1.

What is the most common cause of undercut in welding?

Welding parameters set too high, combined with a travel speed that is too fast, is the classic combination. Michigan Tech’s LTAP training material lists high parameters, inadequate filler metal addition, incorrect work angle and fast travel speed as the four main causes, while ESAB adds long arc length to the list. In practice, the arc is usually too hot, too long, and moving too quickly for the filler to reach the toes.

Can undercut be repaired without removing the whole weld?

Usually yes. Michigan Tech describes grinding the affected area to a uniform groove and adding weld, while ESAB recommends depositing corrective toe runs to refill the groove and blending it smoothly. The defect still has to be excavated back to sound metal under the code’s repair provisions, the repaired section must meet minimum weld size requirements, and the repair must be re-inspected.

Does undercut affect the strength of a welded joint?

Yes, in ways that depend on where it sits and what the joint is doing. The groove removes effective cross-section and acts as a stress raiser at the weld toe, so ESAB notes it can be structurally significant in dynamically loaded joints even when it looks cosmetic. An undercut transverse to the primary tensile stress is far more serious than one running along the stress direction.

Why do I get undercut when welding vertical-up?

Gravity pulls your puddle toward the lower toe faster than filler metal can reach it, so the toe on the lower side is left starved and the arc starts eating into the base metal edge. Keep the puddle small, use stringer beads or a gentle side-to-side weave, and pause briefly at each toe so the filler flows in before you move on.

Is undercut ever acceptable without repair?

Yes, if it falls inside the acceptance criteria of the governing code and the contract documents, and if the loading case allows it. Codes set maximum depth and length limits rather than a blanket ban, so a small groove on a statically loaded non-critical member may pass, while the same groove on a cyclically loaded connection may be a reject. When in doubt, measure the worst point and check the criteria that apply to your joint.

Can you weld over an undercut instead of grinding it out?

No, not on a structural joint where the code applies. Welding over a groove leaves the defect buried in the joint, which removes it from visual inspection without removing it. The defect has to be excavated to sound metal, repaired, and re-inspected under the procedure the code requires.

Does TIG welding avoid undercut?

It reduces the risk but does not remove it. TIG gives you a small, very controllable pool and AIHA notes it has very low fume emission rates, but the root of a butt weld is still where the arc can wander and cut a groove. The cure is the same one you would use elsewhere: fill the root deliberately with filler in the puddle and keep the arc angle steady.

How do I stop undercut from coming back after I repair it?

A repair without a parameter change usually comes back, because ESAB’s remedy guidance explicitly includes re-qualifying the parameters to prevent recurrence. Change the one lever you identified during troubleshooting, usually arc length, travel speed or work angle, and prove it on a scrap coupon before you run it on the joint. Then make the toe check part of the pass, while the bead is still hot enough to shape.