Underfill in Welding: Causes, Acceptance Limits, Inspection, and How to Fix It

A structural welder on a bridge project laid a beautiful-looking cap pass all day, only for the inspector’s gauge to sink into a shallow groove running the full length of the seam — and 40 feet of weld got rejected as underfill. It looked fine to the eye; the numbers told a different story.

Introduction: The Defect That Looks Fine Until Someone Measures It

A structural welder on a bridge project spent the better part of a shift laying cap passes on a long butt joint. The beads came out uniform, the ripple pattern was even, and the finished seam looked like something you would photograph for a portfolio. Then the inspector walked over, pressed a weld gauge into the face of the weld, and the ramp sank into a shallow trough running the full length of the seam. Forty feet of weld was tagged as underfill and rejected. The welder’s first reaction was the familiar one: “It looks fine.” But looking fine and measuring right are two different things, and on that job the numbers ruled.

That story is not really about a bad welder. It is about a defect that hides in plain sight. Underfill is a surface, dimensional imperfection: the finished weld face sits below the plane of the adjacent base metal because not enough filler metal was deposited into the joint. It is subtle enough that an experienced eye can pass right over it, which is exactly why the TWI’s technical guidance on geometric shape imperfections treats profile problems as a separate class from cracks or inclusions. Because it lives on the surface, it is caught primarily by visual inspection (VT) using a weld gauge, not by radiography or ultrasonics. And unlike a crack or a zone of lack of fusion, underfill is not usually a zero-tolerance condition: most codes give it numeric acceptance limits you can measure against.

What You Actually Want to Know About Underfill

If you searched for “underfill in welding,” you are probably trying to settle one of five practical questions:

• What is it? How do you define underfill, and how do you tell it apart from undercut, an undersized fillet, or root concavity?
• Why does it happen? Which parameters, joint-prep mistakes, and habits cause it?
• Is it acceptable? What do ISO 5817 and AWS D1.1 actually allow, and which quality level applies to your job?
• How do I find it? What does an inspector look for, which gauge do they use, and when is visual inspection not enough?
• How do I fix and prevent it? Can you add a wash pass, do you have to gouge it out, and what changes at the machine to stop it recurring?

This article answers all five, in that order, with the acceptance numbers, inspection workflow, repair options, and prevention checklist you can take back to the shop floor.

Who This Guide Is For

The guide is written for four overlapping audiences. Welders who keep getting profile comments and want to understand the numbers behind the feedback. QC inspectors who need a clean reference for gauging and reporting. Fabricators and shop supervisors who are weighing a rework decision against a delivery date. And students and apprentices who are learning to read codes for the first time. If you are setting up equipment or brushing up on fundamentals beforehand, our MIG welder setup guide covers the parameter groundwork that so often decides profile quality, and our welding PPE guide covers the protection you need when you are close to the arc doing fit-up and inspection work.

The One Idea to Hold Onto

Everything in this article rests on a single principle: underfill is a dimensional defect, not an opinion. It has a depth, that depth can be measured in millimetres or fractions of an inch, and the governing code sets a limit for it. Guessing from the truck cab or eyeballing it under a flashlight is not inspection. That shift in mindset, from “does it look acceptable?” to “does it measure within tolerance?” is what separates a welder who passes first time from one who reworks a seam every time an inspector arrives.

Where specific limits, tolerances, and edition details appear below, they are tied to primary or well-established sources: ISO, AWS, TWI, ESAB’s training material, and university-level welding education publications. Read on, and you will finish with a clear picture of what underfill costs a joint, exactly how much of it each code tolerates, how to catch it before the inspector does, and how to make sure your next pass comes out flush the first time.

What Exactly Is Underfill in Welding? Definition and Distinctions

Underfill can be defined in a single sentence: the finished weld face sits below the surface of the adjacent base metal because not enough filler metal was deposited into the joint. The groove was cut, the arc ran, the bead looks like a bead, but the metal that should fill up to the parent surface simply is not there. Nothing has been burned out of the base plate. The shortfall is in the filler.

That is why underfill is a dimensional defect rather than a cosmetic one. ISO catalogs it as imperfection 511, \”incompletely filled groove\”, in the ISO standard covering weld imperfections and quality levels, and the acceptance codes built on that classification give it hard numeric limits in millimetres or fractions of an inch. You cannot settle an argument about underfill by looking harder. You settle it with a straightedge and a gauge.

Root Underfill and Cap Underfill

The same defect shows up in two places:

• Cap underfill is the everyday kind. The visible top surface of the finished weld lies below the plane of the surrounding metal, so the bead sits in a shallow trough along the seam. This is the version a gauge sinks into during a final visual check.
• Root underfill happens at the bottom of the joint, where the first pass failed to build the root up to profile. On open-root pipe and pressure work you often cannot see it until the weld is back-gouled, cut open, or scanned, which is why pipe welding procedures specify root pass technique and reinforcement so precisely.

Both types eat the same quantity: the effective throat, the shortest load path through the weld. A cap 2 mm shallow and a root 2 mm short both leave the joint carrying less metal than the engineer designed for.

Underfill vs. the Imperfections It Gets Confused With

Tradespeople mix these up, because several defects look alike at a glance. The differences are worth memorizing:

Undercut deserves a name check because it is underfill’s usual partner at the weld toe: Xiris notes that undercut is a groove melted into the base metal at the toe even when the weld face is properly filled. We break that one down separately in our guide to welding undercut, causes, and acceptance limits. Likewise, a weld specified at 3/8 in. and measured at 1/4 in. is an undersized weld rather than an underfill, because the comparison is against the drawing callout instead of the base metal surface, as PrimeWeld’s weld quality testing guide points out.

Root concavity and sagging sit in their own columns entirely. TWI’s review of geometric shape imperfections covers how each is generated, and each carries its own designation and limit. Writing \”underfill\” on the report when the real issue is a teardrop root depression sends the repair crew after the wrong problem.

The 10-Second Shop-Floor Test

Lay a straightedge or the blade of a steel rule across the finished bead so it spans both edges of the joint. If daylight passes underneath, or the rule rocks, the face is below the parent metal and you are looking at underfill. If the rule sits flat but the groove at the toe catches your thumbnail, that is undercut. If the bead sits proud, measure its height against the code limit. The surface reveals geometry only; it cannot tell you whether the weld is sound inside.

Filler supply feeds this defect directly, since an electrode or wire too small for the joint deposits metal more slowly than the travel speed uses it up. If you are unsure how filler sizes pair with current settings, our guide to welding electrodes and how to choose them explains the relationship. On thin material the margin is narrower still: one fast pass on sheet TIG work can drop a flush cap into rejection, as our TIG welding sheet metal guide describes.

Get the definition right and everything downstream gets easier: the cause you hunt, the gauge you pick, the code clause you cite, and the repair you write up.

Why Underfill Happens: Root Causes Across Processes

Underfill almost never comes from one dramatic mistake. It comes from a small imbalance that repeats itself inch after inch: the arc is moving, melting and tying in, but the filler metal is not arriving fast enough to fill the crater it leaves behind. TWI lists low welding current, excessive travel speed, insufficient filler deposition and an excessively wide groove as the classic triggers for incompletely filled welds (TWI, geometric shape imperfections), and in practice almost every shop-floor cause you will encounter is a variation of one of those four.

Travel Speed Outrunning Deposition

This is the number one cause, and it is intuitive once you picture it. The torch lays down a fixed volume of metal per second. If you walk faster, that same volume spreads over a longer distance, so the bead gets thinner and sits lower in the joint. A welder chasing a schedule will often speed up first and only later notice the cap sinking below the plate surface. The fix is not to slow down blindly, but to restore the ratio: either reduce travel speed or raise the filler delivery rate to match it.

Current, Voltage and Heat Input

Too little amperage means the filler melts faster than the base metal can accept it, or the wire simply stalls in the joint instead of washing out to the sidewalls. Set the machine too low and you get a cold, narrow, sunken bead no matter how careful your hand is. Setting current correctly is process-specific; a 3/32 E7018 amperage chart is a good example of how much the right rod diameter and polarity matter.

Voltage behaves differently. Excessively high voltage produces a flatter bead and a concave weld profile (The Fabricator), because the wide, soft arc washes the molten pool outward instead of building it up. The weld still looks smooth and well fused, but the face arcs gently downward at the toes. That concavity is textbook underfill.

Joint Preparation: Root Opening and Groove Width

Every extra millimetre of gap is metal someone has to deposit. A root opening opened up beyond the WPS, a bevel ground too wide during fit-up, or repeated grinding between passes can make the joint physically wider than the filler schedule was written for. Over-gapping on a heavy butt joint is one of the fastest ways to produce underfill even with perfect technique, because the volume of the groove has quietly grown while the deposition rate stayed the same.

Multipass Technique and Bead Sizing

In multipass work, underfill starts when stringer beads are laid too thin and too far apart, leaving valleys that the final cap cannot realistically bridge. A welder who stacks narrow, canted beads instead of properly sized, evenly spaced ones ends up with a rippled surface whose low points fall below the base metal line. Work angle matters just as much: tilting the electrode too far toward the centre of the joint builds a high crown in the middle and leaves the edges starved.

Robot and Automated Programming Errors

Manual errors show up on one weld. Programming errors show up on four hundred. In robotic welding, a mistake in wire feed rate or travel speed causes consistent underfill across an entire production run (Xiris welding defects guide). Someone edits the teach pendant, a wire feed slider drifts, a spool change is compensated incorrectly, or a program written for 1.2 mm wire runs unchanged with 1.0 mm wire. The arc parameters look plausible on the controller screen, so the defect is only caught at final inspection.

Which Processes Are Most Prone

Underfill is especially common in high-speed automated GMAW and SAW, because those processes run fast and deposit heavily, so any mismatch between parameters, travel speed and joint design converts immediately into a low cap (Fractory defect overview). Orbital TIG on pipe is the other hot spot: small filler wire, tight root gaps and automated travel speeds mean a slightly low current or a cold, undersized wire leaves a shallow root or cap that no amount of arc optics can hide. Plasma arc welding, with its narrow joint and precise but unforgiving penetration profile, behaves the same way when the filler hand-off is mistimed.

The common thread is calibration. When travel speed, current, voltage, wire size, wire feed speed and joint geometry are treated as one system rather than independent dials, underfill stays off the plate. Start with correct setup, as covered in how to set up a MIG welder, and the defect becomes far easier to avoid.

How Underfill Weakens a Joint: Strength, Fatigue, and Failure Risk

Underfill rarely fails a joint the moment the arc goes out — and that is what makes it dangerous. A sunken cap still looks like a weld and still gets bolted into place. What changed is the amount of metal doing the work: reduced weld volume hits load distribution directly, leaving the joint most vulnerable under dynamic loads, shock loading, and concentrated stress.

The Load Path Runs Through the Thinnest Point

Every welded joint carries force through its smallest cross-section. In a butt weld, that is the throat of the weld metal left after the root and face are accounted for. In a fillet weld, the design throat is roughly 0.707 times the leg length, so a fillet called out at a 6 mm leg is only designed around 4.2 mm of throat. Underfill eats into that number directly.

The arithmetic is unforgiving: stress equals force divided by area. Remove 1 mm of throat from a 6 mm fillet and roughly a quarter of the load-carrying area is gone, so the remaining metal sees proportionally higher stress. A small shortfall on a frame lug under steady load may sit inside the safety factor; the same shortfall on a joint under repeated or impact loads pushes local stress past where the design intended it to sit.

The Concave Face Works Like a Notch

The shape of the defect matters as much as its depth. A properly finished weld face blends smoothly from base metal to weld toe. An underfilled face dips, creating a geometric discontinuity — a stress raiser. As Xiris’s welding defects guide explains, the concave profile creates a notch effect that can serve as a fatigue crack initiation site under cyclic loading.

Fatigue cracks almost never start in the middle of a sound weld. They start at geometric stress raisers: a sharp toe, a crater, a grind mark, or a concave surface where load lines crowd together. Once a crack initiates at an underfilled toe, each load cycle nudges it deeper while the shrinking load-bearing area raises stress again. It is a slow, self-accelerating process that visual inspection misses until the crack is long enough to see. The same notch logic is why undercut carries its own acceptance limits — undercut cuts into the base metal rather than leaving the weld low.

Where It Matters Most: Fillets and Dynamic Structures

Two situations turn a millimetre of underfill from an annoyance into a risk:

• Fillet welds loaded through the throat. Because the fillet’s design strength is tied directly to throat size, any depth shortfall is a straight reduction in capacity. Partial-load situations and eccentric loading make it worse.
• Structures under cyclic or impact loading. Bridges, crane booms, mining equipment, exhaust and piping supports, and rail infrastructure all live on repeated stress cycles. Thermite-joined rail, for instance, sees millions of axle passes, which is why railroad thermite welding is held to such tight profile and grinding tolerances.

Pipe work is another area where a low root or cap profile is a service-integrity issue, not just a strength issue; procedures covered in pipe welding procedures build profile control into the pass schedule for exactly this reason.

Misalignment Makes Underfill Worse

Underfill rarely arrives alone: poor fit-up that produces a low cap often brings high-low misalignment, and the two compound each other. TWI notes that geometric imperfections such as misalignment can compound into lack of fusion and in-service problems. Misalignment forces the joint into secondary bending under load, concentrating stress on the already-thinned, underfilled side, so a defect tolerable on an aligned joint becomes where the crack starts.

Reading Fatigue-Oriented Quality Levels

This is why codes tie acceptance levels to fatigue performance rather than treating underfill as a cosmetic check. ISO 5817 links its quality levels to detail categories used in fatigue design:

Harmless or Hazardous? A Simple Test

Ask three questions before accepting a shallow weld: Is this throat carrying the load? Is the structure subject to cycles, impact, or vibration? Is misalignment or another imperfection stacked on top? If any answer is yes, treat underfill as a strength defect and measure it with a gauge rather than trusting the eye. What looks like a minor cosmetic dip is a missing slice of the joint’s capacity — and on a fatigue-loaded structure, a ready-made starting line for the crack.

Reading the Codes: ISO 5817, AWS D1.1 and Other Acceptance Limits

Two welders can look at the same cap pass and disagree all afternoon. The code cannot. Acceptance limits exist precisely because “looks fine” is an opinion, and an inspector needs a number. When it comes to underfill, the two documents you will meet most often in the shop are ISO 5817 and the AWS D1.1 structural welding code — and they approach the defect from slightly different directions.

ISO 5817 and Imperfection 511

The international standard ISO 5817:2023 lists weld imperfections in Table 1 and grades them into three quality levels: B (highest, most restrictive), C (medium) and D (most tolerant). Underfill appears there as item 511, “incompletely filled groove,” and the acceptance value is expressed as h, the depth the weld face sits below the adjacent base metal surface, against t, the thinner of the joined parts.

Those figures apply to thicknesses above 3 mm. Below that, ISO 5817 allows only very short imperfections on thin sheet, because a shallow dip in 1.5 mm plate is proportionally far more serious than the same dip in 20 mm plate. Notice how the limit scales with thickness: on a 10 mm butt joint at Level C, you are down to 1 mm anyway (0.1 × 10 = 1 mm), so the cap governs. On a 40 mm joint, 0.1t would allow 4 mm, but the 1 mm maximum holds it back.

What AWS D1.1 Says

The American Welding Society’s AWS D1.1 Structural Welding Code handles profile imperfections through groove weld profile rules and fillet size tolerances rather than one neat table. The provisions fabricators trip over most often:

• Fillet undersize allowance. A fillet weld is permitted to underrun its nominal specified size by up to 1/16 in. (1.6 mm) without correction, provided the actual size stays within the code’s minimum requirements. A 1/4 in. fillet called on the drawing can therefore legally land at 3/16 in. on the measuring.
• Undercut. The familiar limit is 1/32 in. (0.8 mm) depth, but D1.1 attaches conditional allowances based on the length of the undercut and the loading the joint sees — a short, isolated undercut in a statically loaded member is treated differently from a continuous line in a cyclically loaded one. For the full reasoning behind that distinction, see our piece on welding defects undercut: causes, acceptance limits, and how to fix it.
• Convexity and concavity. Excess reinforcement and root concavity are capped in much the same spirit: small is tolerable, deep is not.

One bookkeeping detail worth knowing: AWS D1.1/D1.1M:2025 is the 25th edition, released in April 2025, and it supersedes earlier editions on projects that call for “the current edition.” Specifications written years ago may silently point at an old revision, so confirm which edition your WPS and inspection plan reference. For bridge work, D1.1 is not the governing document at all — AWS D1.5 Bridge Welding Code applies, and it is deliberately stricter on profile, toughness and inspection because of the fatigue cycles bridges accumulate over decades. Pipework follows yet another path; for pressure and utility piping, the relevant code and purging rules are covered in our gas line welding code and purging guide, while structural pipe fabrication specs are outlined in pipe welding procedures.

Geometric imperfections like these are catalogued with useful cause-and-effect detail in TWI’s review of geometric shape imperfections, which is a good companion read when you are trying to tell profile problems apart.

How the Drawing Picks Your Level

In practice, the chain works like this: the designer selects a quality level or cites a standard, the fabricator writes a WPS that meets it, the inspector verifies against the same document. If a drawing says “ISO 5817 Level C,” every gauge reading is compared to h ≤ 0.1t with a 1 mm cap. If it says “AWS D1.1, statically loaded,” fillet underrun and the 1/32 in. undercut rule take over. The single most common cause of an argument on the fabrication floor is not a bad weld — it is two people reading two different documents. Confirm the governing standard, the edition, and the quality level before fit-up, and most underfill disputes disappear before they start.

How Underfill Is Detected: Visual Inspection, Gauges, and Monitoring

Underfill is a dimensional defect, so the eye alone cannot judge it: a cap can look smooth and even and still sit below the surface it should reach. Detection works at three moments, before the arc, during the weld and after cooldown. Miss one and you either chase the defect backwards or ship it.

Stage 1: Pre-Weld Fit-Up Checks

Most underfill found at final inspection started at fit-up. Before welding, check root opening, land thickness, bevel angle, alignment and cleanliness against the WPS. A gap opened up by aggressive grinding gives the welder far more volume than the procedure assumed, and the same travel speed into that wider joint brings the cap up short every time.

Use fixtures and spacers instead of eyeballing the gap, confirm tacked parts have not pulled out of alignment, and keep the groove free of mill scale and grind dust. Consistent, correctly spaced tack welding holds the gap steady through thermal cycling. Fit-up is the cheapest inspection there is: it prevents the defect rather than finding it.

Stage 2: In-Process Monitoring

While the weld is being laid, the operator is the first line of defence. A cap that keeps sinking below the joint edges, a concave ripple pattern, or a bead that looks thin for the joint width are live warnings. Adjusting travel speed or deposition at that moment costs seconds; finding it after cooldown costs hours.

In automated cells, underfill is one of the easiest defects to spot in real time by monitoring the pool and bead profile with a weld camera; manually, confirm the machine delivers what the procedure calls for. Wire feed speed, amperage, voltage and travel speed interact, as The Fabricator explains, so verify them against the WPS before the first pass.

Stage 3: Post-Weld Visual Inspection (VT)

Visual inspection is the primary method for underfill, but timing matters. AWS D1.1 allows visual inspection once welds have cooled to ambient temperature, and some steels need 48 hours of cooling before inspection to avoid misleading results from thermal contraction. Measure at the specified temperature, not while the joint is hot.

The standard VT kit is simple:

• A combination weld gauge or fillet gauge for leg, throat and face height
• A bridge cam gauge for groove depth and profile on plate and pipe
• A machinist’s rule or depth gauge for small measurements
• A magnifier and comparator templates checked against the drawing callout
• Low-angle lighting for profile relief, plus appropriate welding PPE near live arcs

What Visual Inspection Cannot See

VT is bounded: it covers the final cap and the accessible root, with no view between passes. Visual inspection sees surface discontinuities only. Anything sealed inside the weld body, such as porosity, lack of fusion or slag inclusions, is invisible to eye and gauge.

Volumetric and surface NDT go further:

• RT (radiography) and UT (ultrasonic testing) reveal internal conditions along the weld length
• PT (penetrant testing) and MT (magnetic particle testing) find surface-breaking cracks and lack of fusion at the toes

Underfill is a surface, dimensional condition, so VT with a gauge is the right tool. Critical joints get both: VT for profile, NDT for soundness. Pipework shows the combined approach covered in pipe welding procedures, while groove profiles are assessed against the geometric imperfection categories TWI describes. The acceptance level comes from the standard on the job, typically ISO 5817, and each imperfection is logged against its clause.

Modern Aids: Cameras and Real-Time Profile Monitoring

High-dynamic-range weld cameras resolve the bead and joint edges through arc glare, giving inspectors a clear view without standing in the plume. In robotic and orbital systems, laser profile sensors trace the finished bead, flag a shallow cap automatically and log the measurement with the part ID. The trend is clear: measure during welding, not just after.

Documentation and Reporting

A measurement that is not recorded is nearly worthless. Every VT report should state the joint ID, measured depth, the gauge used, temperature and time of inspection, the code and level applied, a photograph where possible, and the disposition: accept, repair or reject. Records must survive re-inspection, audit and the life of the structure, because the same weld may be judged again years later.

Underfill does not care how the weld looked to the person who ran it. Measure it, log it, let the numbers decide.

Fixing Underfill: Repair, Rework, and When to Reject

Once the gauge has confirmed underfill, the conversation shifts from measurement to judgment. What happens next depends on three variables: how deep the depression is, what the drawing or code actually allows, and what the joint does for a living. A shallow dip on a static handrail and the same dip on a bridge girder look identical and behave nothing alike. Fixing underfill welds is a decision before it is a task.

The Three-Way Call: Accept, Repair, or Reject

Sort every underfilled joint before anyone reaches for a grinder.

• Accept: The measured depth sits inside the applicable acceptance level, the member is statically loaded and non-critical, and the inspector signs off. Document it, do not polish it away.
• Repair: The depth exceeds the limit, the joint carries fatigue loading, or the specification demands full throat. It needs metal added back.
• Reject and cut out: The underfill travels with cracking or lack of fusion, or the joint is on a safety-critical connection where no dimensional shortfall is tolerated at all.

Repair Options, From Lightest to Heaviest

Blend grinding. For marginal underfill, a light grind dresses the toes and evens the transition so the profile carries no notch. It adds no metal, so it only works where the surface is already within tolerance and the issue is profile quality, not thickness.

A cover or wash pass. The everyday fix. The welder cleans the area, confirms settings against the welding procedure specification, and lays a thin bead over the affected length to bring the face flush with, or slightly proud of, the base metal. This is real repair, not cosmetics, because it restores the cross-sectional area and kills the concave notch. Doing it to a documented procedure rather than by feel is exactly the discipline described in our guide to pipe welding procedures, positions and inspection.

Gouging and re-welding. Where the underfill runs deep or a root region is starved, the correct move is to gouge or grind the area back into a clean uniform groove, take it to bright base metal, then deposit new weld metal to fill it. Weld to the approved procedure and re-inspect. Capping over contaminated or cracked material is not a repair; it is a burial.

Why the Rework Bill Is Real

Industry reporting puts it plainly: underfill welds frequently require rework wherever precision and safety are paramount, because a reduced throat section means reduced load capacity. Welding vision specialists at Xiris make the same case in their defect guide, and the geometric reasoning behind it is covered in TWI’s review of geometric shape imperfections, types and causes. In fatigue service, missing metal is a notch collecting cycles.

What Rework Actually Costs

The weld is rarely the expensive part. The bill breaks down like this:

Multiply that across forty feet of seam and the economics flip. Which is why the parameter discipline in our flux core wire welding techniques matters far more than any heroic cleanup effort later.

Where No Underfill Is Allowed

Some joints get zero tolerance. Bridge fabrication, pressure-retaining components and fatigue-critical connections generally require full throat and a flush profile, because an inspector cannot assume a hidden shortfall is harmless. Under those specs, any measurable underfill gets repaired, however small the reading. At the other end, a short depression on an unstressed bracket may pass at the drawing’s most relaxed level and stay in service untouched.

One closing note for whoever does the grinding: the repair bay is where eyes and lungs take the hit, and dust, sparks and hot metal all demand the right kit. Our guide to welding PPE and protective gear covers it. And keep the mirror image of this whole problem in mind: excess reinforcement is the opposite failure mode governed by the same weld profile tables. Too little is a defect. Too much is a defect too.

Prevention Playbook: Parameters, Procedure, and People

Underfill is almost never a mystery. Ask a welder who laid a shallow cap and they can usually tell you within seconds what went wrong: they moved too fast, the wire was too thin, the amp setting was low, or the gap was wider than the procedure allowed. The good news is that every one of those causes is controllable. Preventing underfill comes down to three things working together: parameters that match the joint, a procedure that is actually followed, and people who know what to look for before the inspector does.

Balance Deposition Rate Against Travel Speed

The single most common cause of underfill is simple: the welder is covering ground faster than filler is being deposited. Travel speed directly changes heat input, so you cannot fix a shallow bead by simply slowing down if the current is also too low — you will just make a slower, still-thin weld. The fix is to balance the two. As a rule of thumb drawn from common defect-prevention guidance, proper electrode size, correct current setting, and avoiding excessive travel speed are the first three checks for any underfill-prone joint.

Practically, that means:

• Verify amperage and voltage against the WPS, not against habit. A 3/32 E7018 that runs beautifully at one setting will underfill if the machine has drifted down. Our 3/32 E7018 amperage guide shows the actual ranges and why polarity matters.
• Increase deposition before increasing speed: step up to the next wire or electrode diameter when the joint is wide, rather than stretching a thin bead across it.
• Re-check settings after every machine change, extension lead change, or power supply dip. Long cables and poor connections rob output.

Control Fit-Up and Joint Preparation

Wide root openings and over-ground joints are silent underfill producers. Every extra millimetre of gap is volume that has to be filled, and if the bead schedule was written for a tighter gap, the welder runs out of passes before the face is flush. Use fit-up fixtures and tack welds to hold the root opening within the range the WPS and joint design call for, and inspect fit-up before the arc is struck. Grinding that widens a bevel beyond drawing tolerance should be treated as a fit-up nonconformance, not a minor cleanup.

Surface condition matters too. As ESAB’s welding defects guide notes, clean surfaces, correct shielding, and proper preheat and interpass control prevent the related family of defects — porosity, lack of fusion, and hydrogen cracking — that often show up alongside underfill when procedures are skipped.

Parameter Checklist Before You Start

Tune the Program, Then Train the Operator

On robotic GMAW lines, underfill is usually a programming artifact: wire feed speed and travel speed were validated on flat coupons and never adjusted for the actual joint. Run first-article parts, measure the profile with a gauge, and tune the program before releasing the lot. Orbitals TIG and SAW lines need the same discipline — seam tracking drift, incorrect filler addition rates, and uncorrected travel-speed ramps all produce shallow caps.

People close the loop. The reliable workflow is: qualify the WPS, train the operator on work angle and travel speed, perform a first-article inspection, then keep in-process gauge checks in the routine. Operators should also be current on the basics that support quality, from machine setup discipline to correct electrode selection and everyday protective gear that keeps them comfortable enough to work steadily rather than rushing.

Common Mistakes, Myths, and Underfill vs Undercut Confusion

After spending hours perfecting a weld, the last thing any fabricator wants to hear is that it has been rejected. However, when it comes to dimensional imperfections, a large portion of these rejections stem from avoidable mistakes. By recognizing these errors and debunking a few myths, welders can prevent unnecessary rework while still upholding the quality that their jobs demand.

The Most Frequent Mistakes in the Shop

One of the biggest errors is treating underfill as purely cosmetic. While it may not have the jagged appearance of a crack, underfill directly reduces the cross-sectional area of a weld. This loss of material can become a major concern for fatigue-prone structures, making it a structural concern rather than an aesthetic one. For those working with structural joints, resources such as our Pipe Welding Procedures guide can help in understanding how these dimensional tolerances are applied in the field.

Another extremely common mistake is judging acceptance by eye instead of a gauge. No matter how experienced a welder may be, depth perception can be deceiving. The slightest concave face can be well below the drawing callout, yet appear flush under shop lighting. This is why calibrated tools, such as a fillet or bridge cam gauge, will always triumph over a visual guess. Proper inspection goes hand in hand with the fundamentals covered in our How to Set Up a MIG Welder guide, as stable parameters make accurate measurements far easier to achieve.

Perhaps the most costly error is ignoring the drawing’s specified quality level. As noted by ISO 5817, acceptance for imperfections is not universal. Depending on whether level B, C, or D is specified, the allowable depth for underfill can vary drastically. Skipping fit-up checks and chasing speed over deposition are also major culprits. When fit-ups are rushed or travel speed is increased to meet quotas, the filler simply does not have time to fill the joint. For processes that are sensitive to these changes, such as Flux Core Wire Welding, small tweaks to wire feed and travel can be the difference between a conforming pass and a reject.

Lastly, a trap that catches even veteran fabricators is using one parameter fix for the wrong defect. Cranking up amperage may remedy an undersized weld, but if the root cause is excessive travel speed, it can worsen underfill or create other issues. Most importantly, welders must remember that visual inspection cannot reveal internal defects. A cap that passes VT can still conceal lack of fusion or incomplete penetration, which will require RT or UT to be detected. This is a critical fact echoed by Xiris and TWI Global alike. These realities further highlight the importance of safety, a practice that starts with proper Welding PPE on every job.

Myths That Need to Be Retired

A few shop myths only serve to reinforce these mistakes. The most persistent is that underfill “will hold” if it looks fine. The truth is that its stress-raising concavity is what makes it dangerous, especially under cyclic loading. Another myth is that undercut and underfill are interchangeable terms. While they can both be dimensional in nature, their location and root causes differ greatly. This distinction is further solidified by Fractory’s welding defects guide, which evaluates profile imperfections separately under established standards.

Underfill vs Undercut: Clearing the Confusion

Final Takeaway

At the end of the day, the prevention of underfill comes down to discipline, not guesswork. Measure with gauges, respect your specified quality level, adjust for the correct root cause, and never assume a cap pass tells the full story of joint integrity. By avoiding these mistakes and correcting the myths surrounding them, you will save your shop time, money, and the costly rework that comes with rejecting a weld that should have been conforming from the start.

Practical Checklist, Real-World Scenarios, and Conclusion

Underfill rarely fails dramatically. It shows up as a number on a gauge at the worst moment: after the seam is finished and the jig is unloaded. It is one of the few defects you can catch with your own eyes and a cheap gauge while the arc is still hot. Everything above collapses into three habits: measure early, measure honestly, and measure against the right code level.

The Pre-Weld, In-Process, and Post-Weld Checklist

Before the arc:

• Confirm the drawing or WPS specifies the quality level (ISO 5817 B, C, or D) and the nominal fillet or groove size. If nobody wrote it down, ask before you strike an arc.
• Check root opening, land, bevel angle, and cleanliness against the procedure. Over-ground or over-gapped joints are the top cause of a thin cap.
• Match filler diameter to the joint — slim rod in a wide groove is how underfill starts. The electrode selection guide pairs rod size with material thickness.
• Verify amperage, voltage, and wire feed speed against the WPS, not memory, and tack evenly so shrinkage does not open the joint — see the MIG tack weld guide.

During the weld:

• Watch deposition against travel speed. If the puddle races and the bead flattens, slow down before the pass ends.
• Gauge the first 6–12 inches of a long seam, then re-check at intervals — the cheapest place to catch underfill is mid-seam.
• On multipass work, keep stringer beads full-sized rather than thin ripples that leave a permanent low spot.

After the weld:

• Let the joint cool to ambient before judging profile, then measure with a fillet or bridge cam gauge rather than by eye.
• Record the reading, weld ID, operator, and procedure in the weld map or inspection report.
• VT only sees the final cap and accessible root; internal soundness needs RT, UT, PT, or MT.

Three Scenarios Where the Gauge Decided

Structural steel butt joint, 10 mm plate. A fitter opens the root slightly to compensate for fit-up, the welder speeds up to close the shift, and the cap lands 1.4 mm below the plate surface. Against ISO 5817 Level C (h ≤ 0.1t, max 1 mm) the weld fails. Two minutes of in-process gauging halfway along would have caught it; a wash pass at the end would have saved it.

Robotic GMAW production line. A wire-feed drift of a few percent across hundreds of identical parts turns one marginal cap into a systemic reject. Real-time profile monitoring and camera-based checks flag the trend early, as covered in welding defect monitoring literature. The fix is re-teaching the program, not re-inspecting the bin.

Orbital pipe TIG. A hot pass run too fast leaves the cover concave on sanitary or pipe welding procedure work where profile requirements are tight. The repair is a careful cover pass at correct parameters, followed by full re-inspection of the coupon.

A Simple Accept / Repair / Reject Flow

1. Identify: Is the face below the base metal, or is it a groove melted into the toe? If it is a groove at the toe, you are looking at undercut, not underfill — different limits, different fix.
2. Measure: Depth h and plate or leg thickness t.
3. Compare: h against the specified ISO level or AWS tolerance.
4. Decide: Within limit and non-critical → accept and document. Outside limit but accessible → repair (blend, wash pass, or gouge) under an approved procedure, then re-inspect. Safety-critical joint or internally suspect → reject and escalate.

As ESAB’s welding defects guide puts it, combining skilled operators, qualified procedures, and rigorous QA prevents most weld defects — and proper PPE keeps that QA crew safe, as covered in this welding PPE guide.

Ten Key Takeaways

1. Underfill is dimensional, not cosmetic — it is measured, not admired.
2. It is formally designated as imperfection 511 in ISO 6520-1.
3. Too much travel speed, too little heat, or too little filler causes it.
4. Reduced throat means lower load capacity and a fatigue crack notch.
5. ISO 5817 Level B forbids it; C and D allow graded depths.
6. AWS D1.1 tolerates limited fillet underrun; bridge work is stricter.
7. Gauges beat eyes, every single time.
8. In-process checks cost minutes; end-of-job rework costs days.
9. Repair with an approved procedure and always re-inspect.
10. Prevention is people, procedures, and parameters working together.

The bottom line: the welder who got 40 feet rejected did not lack skill — he lacked a gauge at the halfway point. Underfill rewards the welders and inspectors who measure rather than guess, and punishes anyone who trusts a good-looking cap instead of a number. Check the level, check the depth, and check it early.

Frequently Asked Questions

What is underfill in welding?

Underfill is a weld whose face sits below the surface of the adjacent base metal because too little filler metal was deposited. It reduces the effective throat and creates a concave, notch-like profile. It is classified as ISO 6520-1 imperfection 511, ‘incompletely filled groove.’

Is underfill an acceptable weld defect?

It depends on the code and the quality level specified. ISO 5817 forbids it at Level B, allows h ≤ 0.1t (max 1 mm) at Level C, and h ≤ 0.25t (max 2 mm) at Level D for thickness over 3 mm. AWS D1.1 permits small undershoots within its weld profile tables, including a 1/16 in. fillet size tolerance.

What causes underfill in welding?

The main causes are travel speed that is too fast, amperage or heat input too low, insufficient filler deposition, an electrode or wire that is too small, an overly wide root opening from poor fit-up, and beads laid too thin during multipass welding. In robots, wrong wire feed or travel speed programming causes it across whole runs.

How does underfill affect weld strength?

It lowers the weld’s cross-sectional area, so the joint carries less load than designed. The concave profile also acts as a stress raiser, and under cyclic loading it can become the site where a fatigue crack starts. That is why fatigue-critical work specifies stricter quality levels.

How is underfill detected during inspection?

By visual inspection using a weld gauge, such as a fillet or bridge cam gauge, measuring weld face height against the drawing callout. Because it is a surface defect, it does not normally need radiography or ultrasonics — though those methods are still needed for internal flaws that vision cannot see.

What is the difference between underfill and undercut?

Underfill is the entire weld face sitting below the base metal surface from insufficient filler. Undercut is a sharp groove melted into the base metal at the weld toe that filler never refilled, and it occurs even when the weld face itself is fully filled. Both concentrate stress but have different causes and limits.

How do you fix a weld with underfill?

Grind the area to a clean uniform profile and add a cover or wash pass so the face is flush or slightly proud, following an approved repair procedure, then re-inspect with a gauge. Deep or root-area underfill may need gouging out and rewelding.

How do I prevent underfill on the next job?

Balance deposition rate with travel speed, verify amperage and voltage against the WPS, control root opening and joint prep, choose the right electrode or wire size, size each stringer properly in multipass work, and check the bead with a gauge early so you catch the trend before the whole seam is run.