A slag inclusion is non-metallic flux residue trapped inside a weld, usually between passes or against a groove sidewall. Codes judge it by size, not presence. ASME Section VIII full radiography, for example, rejects an elongated slag indication longer than 1/4 in. in a weld up to 3/4 in. thick, while AWS D1.1 classes UT indications by echo rating.
Slag itself is normal. Shielded metal arc (SMAW), flux-cored (FCAW) and submerged arc (SAW) welding all produce a flux layer that protects the molten pool. The defect starts when that layer isn't fully removed and the next pass buries it.
This guide covers why slag gets left behind, how to prevent it, what a slag inclusion looks like on an ultrasonic A-scan, how to locate one with an angle-beam probe, and how the main US codes accept or reject it. It is part of our weld defect types guide, which compares all the common weld flaws.
Contents: What is a slag inclusion? · Causes · Prevention · UT signal · Worked example · Acceptance criteria · Limits · FAQ
What is a slag inclusion in welding?
A slag inclusion is a solid non-metallic particle or layer, left from the welding flux, that is trapped in the weld metal or at the fusion line. ISO 6520-1:2007 places it in group 3, solid inclusions, separate from cavities such as porosity (group 2) and from lack of fusion (group 4).
TWI describes how slag appears on a radiograph:
"Slag is normally seen as elongated lines either continuous or discontinuous along the length of the weld." — TWI, Job Knowledge 43: Slag inclusions
TWI also notes that slag is less severe than a crack because it lacks sharp edges, which is why codes allow small amounts.
| Form | Where it sits | Usual cause |
|---|---|---|
| Linear (interpass) | Along a pass boundary, often running the length of the weld | Slag not removed between runs; poor bead overlap |
| Sidewall | Against the groove face, at the toe of an earlier pass | Undercut or a convex bead that leaves a slot slag can't escape from |
| Isolated | A single pocket | Local cleaning miss, a stop-start |
| Clustered | Several small particles in one area | Turbulent pool, fragments of broken slag |
| Root area | Near the root pass | Narrow root opening, poor access for cleaning |
Slag inclusions are volumetric: they have width and height as well as length. That puts them in the same family as porosity and separates them from planar flaws such as cracks and lack of fusion. The distinction decides both the best NDT method and how strictly the code treats the flaw.
Key takeaway: a slag inclusion is flux residue that a later pass buried; it is volumetric and usually elongated along the weld.
What welding practices leave slag behind?
Most slag inclusions come from incomplete cleaning between passes, made worse by bead shapes that create places for slag to hide. Technique, parameters and joint design decide how easy that cleaning is.
Incomplete cleaning between passes
Interpass cleaning is the most direct control. Chipping removes the bulk of the slag but can leave fine residue at the bead toes and in corners. Wire brushing, a needle scaler or light grinding may be needed. TWI advises grinding for difficult butt joints and notes that basic fluxes, with their convex bead and adherent slag, carry a higher risk.
Poor electrode angle and manipulation
An unsuitable electrode angle can push slag ahead of the weld pool, where the arc rolls over it instead of floating it out behind. Excessive weaving makes an uneven bead that is harder to clean. The correct angle depends on the process, consumable, position and joint, so the qualified procedure is the reference.
Travel speed and heat input
A travel speed that is too fast gives a small pool that freezes before slag can float to the surface. A heat input that is too low for the joint reduces fluidity and penetration, so slag at the sidewall isn't melted out. Raising the current isn't a universal cure; current, voltage and travel speed must stay inside the range the welding procedure specification (WPS) allows.
Bead shape, overlap and undercut
TWI names too little overlap between adjacent beads, and undercut at the toe of earlier runs, as the main geometric causes. A convex bead leaves a narrow slot at each toe. The next pass bridges over it and seals the slag in.
Joint preparation and surface condition
A narrow groove, a tight root opening or a poorly prepared bevel limits electrode access and cleaning. Rust, mill scale, oil, paint and moisture interfere with the arc and the flux chemistry. Consumables should be stored and handled as the manufacturer and the WPS require, since damp flux behaves differently in the pool.
How do you prevent slag inclusion in welding?
Prevent a slag inclusion by cleaning every pass completely, keeping the bead flat and well overlapped, and running parameters inside the WPS. The cheapest chance to fix it is before the next pass covers it.
| Step | What to do | Why it works |
|---|---|---|
| 1. Prepare the joint | Groove angle and root opening per the WPS; remove rust, scale, oil and paint | Gives access for welding and cleaning |
| 2. Set parameters | Current, voltage and travel speed inside the WPS range | Keeps the pool fluid long enough for slag to float out |
| 3. Control the bead | Flat or slightly convex profile, adequate overlap, limited weave | Leaves no toe slots for slag |
| 4. Clean each pass | Chip, then wire brush; grind toes and crevices where needed | Removes residue before it is buried |
| 5. Look before the next pass | Visual check of toes, corners and the sidewall | Catches a miss while it is still on the surface |
| 6. Grind out problems | Grind back convex beads, undercut or visible slag lines | Removes traps a later pass can't fuse out |
Visual testing between passes catches surface slag, undercut and poor profile. It can't see into a completed weld, which is where ultrasonic testing (UT) or radiographic testing (RT) comes in.
Key takeaway: prevention happens pass by pass; once the weld is capped, finding a slag inclusion needs volumetric NDT.
Which NDT method finds a slag inclusion best: UT or RT?
RT shows slag clearly as dark, irregular lines, while UT locates it in depth and is more sensitive to the planar flaws that slag is often confused with. Many inspection plans use one as the main method and the other to resolve doubtful indications. Our guide to NDT for welding inspection compares all the methods.
| Consideration | Ultrasonic testing (UT) | Radiographic testing (RT) |
|---|---|---|
| Principle | Reflected sound | Differential absorption of X-rays or gamma rays |
| How slag shows | Low to moderate, multi-peak echo | Dark, irregular, elongated image |
| Depth information | Yes, from sound path and angle | No, unless extra exposures are taken |
| Planar flaws (lack of fusion, cracks) | Strong when the beam faces the flaw | Can be missed if not aligned with the beam |
| Radiation safety | Not needed | Exclusion zone and licensing needed |
| Record | A-scan data; PAUT and encoded scans give images | Film or digital image |
| Effect of orientation | Strong | Moderate |
Phased array ultrasonic testing (PAUT) sweeps a range of angles from one probe position and records an image, which makes it easier to see the extent and shape of a slag line than with a single-angle A-scan.
What does a slag inclusion look like on a UT A-scan?
A slag inclusion usually gives a lower-amplitude, multi-faceted echo that is not mirror-like (non-specular), and its peaks shift and change height as the probe is moved or orbited. Lack of fusion gives a sharp, strong echo at one angle, and porosity gives a cluster of echoes that is often stronger than slag.
Evident's waveform interpretation tutorial puts the comparison plainly: "Slag can look very similar to porosity." It adds that the multi-faceted slag echo may be weaker than the porosity echo, and that porosity peak shapes and amplitudes change rapidly as the probe is rotated (Evident, Basic Waveform Interpretation).
The behavior comes from shape. Slag is rough and irregular, so it scatters sound in many directions. Only part of the energy returns to the probe, and the facets that face the beam change as the probe moves. Lack of fusion is a smooth, flat face, so it acts like a mirror: very strong when the beam hits it squarely, very weak when it doesn't.
| Feature | Slag inclusion | Lack of fusion (sidewall) | Porosity |
|---|---|---|---|
| Echo shape | Broad, jagged, several peaks | Single sharp peak, fast rise and fall | Cluster of sharp peaks (single pore: one sharp peak) |
| Typical amplitude | Low to moderate | High when the beam is normal to the fusion face; can be very low when it isn't | Moderate; can exceed slag |
| Orbital scan (probe circled around the indication) | Peaks change and wander; amplitude drops off gradually | Amplitude falls off quickly as the beam turns off the plane | Peaks change rapidly; response similar from most directions |
| Swivel (probe rotated in place) | Moderate change | Sharp drop | Little change for single pores |
| From the other side of the weld | Usually seen from both sides at similar amplitude | Often strong from one side only | Seen from both sides |
| Response with a different angle | Similar amplitude at 45°, 60° and 70° | Strongest at the angle nearest perpendicular to the bevel | Similar at all angles |
| Typical location | Pass boundaries; toe of earlier runs; can be anywhere through the thickness | On the bevel line | Anywhere in the weld metal |
| Length by scanning along the weld | Often elongated, with an uneven echo along its length | Elongated, steady echo | Short, or scattered over an area |
Evident gives the lack-of-fusion signature as a "strong reflection with fast rise and fall time in the second leg" when tested from the side with the fusion gap. Our guide to angle-beam testing in UT explains the wedge angles and probe motions used to build these pictures.
Practical sequence for characterizing a suspected slag indication:
- Peak the echo and record sound path, surface distance and gain.
- Plot depth and position against the weld cross-section (see the worked example below).
- Orbit and swivel the probe and watch whether the echo is one clean peak or several shifting ones.
- Scan from the other side of the weld and, if the procedure allows, with a second angle.
- Scan along the weld to find the length, usually by the 6 dB drop method at each end.
- Compare the location with the bevel line. An indication sitting exactly on the fusion face needs more evidence before it is called slag.
Key takeaway: no single echo proves slag; the combination of amplitude, echo shape, response to probe motion and location against the weld geometry supports the call.
Worked example: locating a slag indication with a 70° probe
The depth and surface distance of an indication tell you whether it sits in the weld body (more likely slag or porosity) or on the fusion face (where lack of fusion must be ruled out). This example uses the standard geometry: leg sound path = t ÷ cos θ, depth = sound path × cos θ in the first leg, depth = 2t − sound path × cos θ in the second leg, and surface distance = sound path × sin θ.
Setup. A 1.25 in. thick single-V butt weld in a statically loaded joint, 30° bevels (60° included angle), 1/8 in. root opening. The weld is scanned with a 70° shear-wave probe. The reference level from the 0.060 in. side-drilled hole is b = 46 dB.
Geometry check first. First-leg sound path = 1.25 ÷ cos 70° = 1.25 ÷ 0.342 = 3.65 in. Any indication with a shorter sound path is in the first leg; a longer one is in the second leg, after the bounce off the back wall.
| Step | Indication 1 | Indication 2 |
|---|---|---|
| Sound path | 2.50 in. | 5.00 in. |
| Leg | First (2.50 < 3.65) | Second (5.00 > 3.65) |
| Depth below scanning surface | 2.50 × 0.342 = 0.86 in. | 2.50 − (5.00 × 0.342) = 2.50 − 1.71 = 0.79 in. |
| Surface distance from beam exit point | 2.50 × 0.940 = 2.35 in. | 5.00 × 0.940 = 4.70 in. |
| Measured beam exit point to weld centerline | 2.30 in. | 4.37 in. |
| Position from centerline | 2.35 − 2.30 = 0.05 in. past it | 4.70 − 4.37 = 0.33 in. past it |
| Fusion face distance from centerline at that depth | 0.0625 + (1.25 − 0.86) × tan 30° = 0.0625 + 0.39 × 0.577 = 0.29 in. | 0.0625 + (1.25 − 0.79) × 0.577 = 0.0625 + 0.27 = 0.33 in. |
| Where it sits | Near mid-thickness, close to the weld center, well inside the weld metal | On the far fusion face |
Both depths are inside the 1.25 in. plate. If Indication 2 had been read as first-leg (5.00 × 0.342 = 1.71 in.), it would sit below the back wall, which is impossible.
The fusion-face distance uses the half-root opening (0.0625 in.) plus the bevel's horizontal offset at that depth, measured up from the root. It assumes the weld cap is centered on the joint; check the actual bevel position from the weld map.
Characterizing them. Indication 1 sits in the weld body at a pass boundary. With a broad, multi-peak echo that wanders on orbiting and is seen from both sides, it is a likely slag inclusion. Indication 2 lies on the bevel line, which is where sidewall lack of fusion forms. It should be re-scanned with a 60° probe, which meets a 30° bevel face close to perpendicular (90° − 30° = 60°). A sharp echo that jumps in amplitude at 60° points to lack of fusion; a similar, ragged echo at both angles is more consistent with slag trapped against the sidewall.
Rating Indication 1 under AWS D1.1 (static). The rating is d = a − b − c, where c = 2 × (sound path in inches − 1).
- Indication level: a = 52 dB
- Attenuation factor: c = 2 × (2.50 − 1) = 3 dB
- Rating: d = 52 − 46 − 3 = +3
- In the over 3/4 to 1½ in. row for a 70° probe, secondary sources give Class A as +2 and lower and Class B as +3, so this is Class B
- Class B rejects if longer than 3/4 in. The measured length by 6 dB drop is 1/2 in., so it is acceptable, provided no other Class B or C indication lies within 2L and it is not within 2L of a weld end carrying primary tensile stress.
The class bands come from secondary reproductions of the D1.1 static table, so confirm them against your licensed edition. Our AWS D1.1 weld inspection criteria guide works through the rating in more detail.
The same flaw in a pressure vessel. If a 1/2 in. slag indication with an amplitude above the reference level were found in a 1.25 in. ASME Section VIII Division 1 weld examined to Appendix 12, it would reject, because 1/2 in. is longer than t/3 = 1.25 ÷ 3 = 0.42 in. (secondary sources; confirm). The same flaw passes one code and fails the other.
Key takeaway: plot every indication against the weld cross-section before naming it; an indication on the bevel line is lack of fusion until shown otherwise.
How do AWS D1.1, ASME and B31.3 judge a slag inclusion?
AWS D1.1 doesn't name the flaw type in its UT tables; it classes each indication by echo rating and length, while ASME Section VIII and B31.3 radiography set slag-specific length limits tied to thickness. The codes below are paywalled. The values come from secondary sources that agree with one another and must be checked against the edition named in your contract.
AWS D1.1 UT acceptance (Clause 8)
The current edition is AWS D1.1/D1.1M:2025, which AWS announced on April 17, 2025; many contracts still cite 2020. In both editions, inspection sits in Clause 8. UT acceptance is in Table 8.2 for statically loaded nontubular connections and Table 8.3 for cyclically loaded ones (Tables 6.2 and 6.3 in 2015 and earlier).
| Discontinuity severity class | Meaning | Static (Table 8.2) | Cyclic (Table 8.3) |
|---|---|---|---|
| A | Large discontinuity | Reject at any length | Reject at any length |
| B | Medium discontinuity | Reject if longer than 3/4 in. | Reject if longer than 3/4 in. |
| C | Small discontinuity | Reject if longer than 2 in. | Reject if longer than 2 in. in the middle half of the thickness, or longer than 3/4 in. in the top or bottom quarter |
| D | Minor discontinuity | Accept regardless of length or location | Accept regardless of length or location |
The class is read from the indication rating d against the row for the weld thickness and probe angle, so there are no fixed dB bands. A lower (more negative) rating means a stronger reflector, which is why Class A sits at the low end. Evident's AWS weld inspection tutorial gives the same d = a − b − c formula. D1.1 itself doesn't require UT; the contract documents, or AISC 360 and AISC 341 for buildings, decide when it is used.
ASME Section VIII Division 1
| Method | Slag rule (secondary sources, confirm) |
|---|---|
| Full RT, UW-51(b) | An elongated indication rejects if longer than 1/4 in. (t up to 3/4 in.), t/3 (t over 3/4 to 2¼ in.) or 3/4 in. (t over 2¼ in.). A group of aligned indications rejects if their total length is more than t in a 12t length, unless they are spaced more than 6L apart (L = the longest in the group) |
| UT, Mandatory Appendix 12 | Indications over 20% of the reference level are investigated. Cracks, lack of fusion and incomplete penetration reject at any length. Other indications, including slag, reject only if the amplitude exceeds the reference level and the length exceeds 1/4 in., t/3 or 3/4 in. by the same thickness bands |
The and matters: a long slag line below the reference level, or a strong but short one, is acceptable under Appendix 12. Calling an indication "lack of fusion" instead of "slag" changes the outcome to reject at any length, which is why characterization is part of acceptance.
ASME B31.3 process piping
For girth welds in normal fluid service, an AWS forum quotation of Table 341.3.2 gives slag inclusion limits of an individual length up to Tw/3, an individual width up to 2.5 mm (3/32 in.) and Tw/3, and a cumulative length up to Tw in any 12Tw weld length. One blog summary gives the width as 2.4 mm. Treat these values as Med confidence and read the table and its notes in your edition. B31.3 UT acceptance (344.6.2) uses an amplitude and length rule like Appendix 12.
ISO
ISO 6520-1 names and groups imperfections but sets no limits. ISO 5817 sets quality levels B, C and D for fusion welds in steel, and ISO 11666 sets UT acceptance levels linked to them. Our guide to UT acceptance criteria across codes compares them with the US codes. A project specification can be stricter than any of these.
Key takeaway: the governing code, the loading or service category and the wall thickness all change whether the same slag inclusion passes; quote a limit only with its conditions.
What happens after UT finds a slag indication?
The indication is located, sized, characterized and compared with the governing acceptance criteria; only an indication that fails is repaired. A signal on the screen starts an evaluation; it doesn't decide one.
- Locate it: sound path, depth, surface distance and position against the weld centerline.
- Characterize it: echo shape, response to orbiting and swiveling, response from the other side and at another angle.
- Size it: rate it (D1.1) or compare with the reference level (ASME), then measure length.
- Compare it with the code and the contract, as in our guide to UT acceptance criteria.
- If it rejects, excavate it under the approved repair procedure, usually by grinding or air-carbon-arc gouging, and confirm removal (often with MT or PT on the excavation).
- Reweld under the WPS and re-examine the repair by the same method and extent as the original weld, unless the contract requires more.
The UT report should record the reference level, attenuation factor, rating, class, length and location of every indication evaluated, so a later reviewer can check the call.
What UT and RT can't tell you about slag
Neither UT nor RT can prove an indication is slag; both infer it from indirect evidence, and both have blind spots.
- UT can't reliably separate slag from porosity. Both scatter sound and give multi-peak echoes. Evident notes that the two can look very similar.
- Slag on the fusion face looks like lack of fusion, and vice versa. Lack of fusion often carries oxide or slag on its face, so the two can coexist in one indication.
- Amplitude isn't size. A rough slag line returns only part of its energy to the probe, so a long inclusion can rate as a minor Class D indication. A smooth flaw at a poor angle can also rate low.
- D1.1's attenuation factor is an approximation. The 2 dB per inch model can be several dB off at long 60° and 70° sound paths (Holloway, CINDE Journal, 2017), enough to move an indication across a class boundary on thick welds.
- RT doesn't give depth from a single exposure and can miss tight planar flaws that are not aligned with the beam.
- Geometry echoes mislead. Root and cap reinforcement, mismatch and a backing bar can all give echoes near where slag forms. Plot every indication before calling it.
- Neither method replaces procedure. Without a written UT procedure, calibrated reference levels and a qualified technician, the characterization in the table above is opinion.
How EMA Industries applies this
EMA Industries performs AWS D1.1 structural weld UT, MT and PT, provides NDT Level III services and supports WPS and PQR work. We are based in Southern California and mobilize nationwide. On weld UT jobs, the governing code, edition, loading category and NDT extent should be confirmed before scanning, and suspected slag on or near a bevel should be checked from both sides and with a second angle before it is reported. See our ultrasonic examination services or contact us to discuss a project.
Related reading: Weld defect types · Lack of fusion in welds · Porosity in welding · Angle-beam testing in UT · AWS D1.1 weld inspection criteria
Frequently asked questions
What happens if you don't remove slag from previous weld passes?
The next pass covers it and traps it inside the joint as a slag inclusion. Once buried, it can't be found by visual inspection and needs UT or RT. Depending on its size and the governing code, it can be accepted, or it can force a repair, a re-examination and a delay.
How do you remove slag from a weld?
Chip off the bulk of the slag with a chipping hammer, then wire brush the surface. Use a needle scaler or light grinding at the bead toes, corners and sidewalls, where residue sticks. Grind back a convex bead or undercut that would trap slag under the next pass, and look at the surface before welding over it.
Is it okay to leave slag on a weld?
Not between passes, because the next layer can bury it and form an inclusion. On a finished weld, surface slag should still be removed so the weld can be visually inspected and so it doesn't hold moisture against the steel. Some coatings and inspections require a fully cleaned surface.
What are slag inclusions in welding?
Slag inclusions are pockets or layers of solidified flux residue trapped within a weld, usually along pass boundaries or against the groove sidewall. They are volumetric and often elongated. Their significance depends on size, location and distribution, judged against the governing code.
How can you tell slag from lack of fusion on UT?
Slag usually gives a lower, ragged, multi-peak echo that changes as the probe is orbited and appears from both sides of the weld. Lack of fusion gives a sharp, strong echo at the angle facing the bevel, often from one side only, and sits on the fusion line. Location against the weld cross-section is the strongest clue.
Is a slag inclusion always a reason to reject a weld?
No. Most codes allow small slag inclusions. AWS D1.1 accepts a UT indication based on its severity class and length, and ASME Section VIII radiography rejects an elongated indication only above a thickness-based length (1/4 in. for welds up to 3/4 in. thick). Cracks, by contrast, are rejected at any size.
Sources
- Defects/imperfections in welds: slag inclusions (Job Knowledge 43), TWI (definition, causes, prevention, RT appearance, acceptance context)
- Basic waveform interpretation, Evident (slag, porosity and lack of fusion echo signatures)
- AWS weld inspection tutorial, Evident (D1.1 indication rating d = a − b − c)
- AWS announces release of D1.1/D1.1M:2025, American Welding Society, April 17, 2025
- Holloway, P., analysis of the AWS D1.1 attenuation factor, CINDE Journal, 2017
- AWS D1.1/D1.1M:2020 and :2025, Clause 8, Tables 8.2 and 8.3; ASME BPVC Section VIII Division 1, UW-51 and Mandatory Appendix 12; ASME B31.3, Table 341.3.2 and para. 344.6.2; ISO 6520-1:2007; ISO 5817; ISO 11666:2018 (all paywalled; values above from agreeing secondary sources, to be confirmed against the licensed text)