Lack of fusion in welds is a planar flaw where weld metal fails to bond with the joint sidewall or a previous bead. Ultrasonic testing (UT) finds it best when the beam meets the fusion face near 90°: a 60° probe on a 30° bevel. UT usually beats radiography here, but tightly closed lack of fusion can escape every method.
This guide explains where lack of fusion forms, how to choose a probe angle from the bevel angle, what phased array (PAUT) and time-of-flight diffraction (TOFD) add, and how AWS D1.1 and the ASME codes treat a lack-of-fusion indication once it is found.
What is lack of fusion, and how is it different from lack of penetration?
Lack of fusion is a failure of weld metal to bond with an adjacent surface; lack of penetration is a failure of weld metal to reach through the joint, usually at the root. The two often share causes, such as low heat input, but their location and shape differ, and so does the inspection approach.
TWI separates lack of fusion into two forms. Lack of sidewall fusion is weld metal that fails "to fuse completely with the sidewall of the joint". Lack of inter-run fusion is weld metal that fails "to penetrate adequately the previous weld bead" (TWI Job Knowledge 41). A third position, lack of fusion at the root face, is often grouped with it.
| Imperfection | Where it sits | What failed | Typical orientation |
|---|---|---|---|
| Lack of sidewall fusion | Along the bevel face (fusion boundary) | Weld metal did not melt into the parent metal | Planar, tilted at the bevel angle |
| Lack of inter-run fusion | Between passes inside the weld | A pass did not melt into the pass below | Planar, roughly parallel to the surface or along the bead profile |
| Lack of root fusion | Root face or land | Root face not melted | Planar, near vertical |
| Lack of penetration (incomplete joint penetration) | Root of the joint | Weld metal did not extend through the thickness | Planar or a gap at the root, near vertical |
The orientation column is the one that matters for UT. A planar flaw reflects sound like a mirror, so the beam angle that sees it depends on how it is tilted. Lack of fusion, slag inclusions and porosity are compared in our guide to weld defect types.
Key takeaway: sidewall lack of fusion lies on the bevel face, so the bevel angle tells the inspector where to aim.
What causes lack of fusion in welds?
Lack of fusion in welds forms when the arc does not deliver enough heat to the surface it has to melt, or cannot reach that surface. TWI names "too narrow a joint preparation" and "incorrect welding parameter settings" as principal causes, followed by poor welder technique and magnetic arc blow, and notes that oily or scaled surfaces also contribute.
| Cause | Mechanism | What the welding inspector can check |
|---|---|---|
| Low current or high travel speed | Too little heat reaches the sidewall or the previous pass | Parameters against the qualified WPS ranges |
| Narrow groove or small bevel angle | The arc cannot reach the sidewall | Groove angle and root opening before welding |
| Wrong electrode or gun angle | Heat goes to one side of the joint | Welder technique, especially in tight positions |
| Restricted access | The welder cannot hold the correct angle | Joint location and sequence |
| Magnetic arc blow | The arc is deflected from the joint | Arc behavior; current return position |
| Oily, scaled or poorly cleaned surfaces | Oxide and contamination prevent wetting | Interpass cleaning |
| Convex beads | The next pass cannot wet the toe | Bead shape between passes |
TWI adds that these imperfections "are more likely to happen when access to the joint is restricted." For persistent arc blow, moving the current return or demagnetizing the steel is the usual fix. Parameters belong in the welding procedure; see our guide to the welding procedure specification (WPS).
Key takeaway: most lack of fusion is a heat or access problem, and a WPS that allows a narrow groove makes it more likely.
Why is lack of fusion hard to find?
Lack of fusion is hard to find because it is thin, planar and often tightly closed, so it can return little energy to any NDT method. A wide flaw does not guarantee a strong signal.
Gabriel Rihar of the Welding Institute in Ljubljana sorted lack of fusion by structure in a paper hosted on NDT.net:
| Form | What it looks like | Detectability reported |
|---|---|---|
| Lack of fusion with voids | A small gap between the faces | Detectable by UT; RT shows the voids |
| Lack of fusion with non-metallic inclusions | Oxides or slag trapped on the interface | Detectable; RT shows the inclusions, not the unfused plane itself |
| "Open" lack of fusion | Faces separated by cooling stresses | Detectable |
| "Pure" lack of fusion | Faces in contact with no measurable gap | Not detectable by NDT; found only by microscopic examination |
Rihar's paper reports that "considerably large lack-of-fusion defects give weak ultrasonic indications", which in practice are often judged minor and acceptable. He attributes this to faces that stick together in many small contact points, and to sidewall flaws on a V-groove lying at an angle that reflects the beam away. The paper also warns that a lack-of-fusion defect can open under load and make a tank leak.
That finding is why "UT is the best method" needs a qualifier. UT is usually the most capable method for this flaw. No method is reliable for a tightly closed one.
Which NDT methods can detect lack of fusion in welds?
For embedded lack of fusion, UT is normally the first choice, radiography (RT) is a partial second, and surface methods only work if the flaw breaks the surface. TWI's own guidance is explicit:
"Ultrasonic inspection is normally more effective than radiography in detecting lack of inter-run fusion imperfections." (TWI FAQ)
TWI adds that effectiveness depends on the material, and that surface-breaking imperfections can be found with penetrant or magnetic particle testing.
| Method | Finds lack of fusion when… | Misses it when… |
|---|---|---|
| Visual testing (VT) | It breaks the surface at a toe or the cap; also catches poor bead shape that predicts it | It is buried, which is most of the time |
| PT | It is surface-breaking and open | It is subsurface or closed |
| Magnetic particle testing (MT) | It is at or just below the surface of ferromagnetic steel | It is deeper than near-surface |
| RT | The radiation beam runs along the plane of the flaw, or the flaw carries voids or inclusions | The plane is tilted to the beam, or the faces are tight |
| Conventional UT | One beam strikes the fusion face near normal incidence | The beam is off-angle, coverage is incomplete, or the faces are in tight contact |
| PAUT | A sectorial or multi-angle scan includes a near-normal angle | Same physics limits as UT, plus setup errors |
| TOFD | Tip diffraction from the flaw edges is above noise | The flaw is in the near-surface or back-wall dead zone |
For RT, the British Institute of NDT advises directing the radiation beam down the angle of the fusion face to improve detection of sidewall flaws (BINDT). That is rarely practical on a production weld, which is one reason UT is preferred for planar flaws. RT remains the better tool for volumetric flaws such as porosity, while slag inclusion sits between the two because slag often lines the same fusion planes.
How is the UT probe angle chosen from the bevel angle?
Choose a refracted angle that puts the beam at, or as close as possible to, 90° to the fusion face: for a single-V or single-bevel groove, that angle is about 90° minus the bevel angle. BINDT notes that EN ISO 17640 requires one of the probe angles to examine the weld fusion faces at, or as near as possible to, normal incidence, because that maximizes the echo from sidewall lack of fusion.
The geometry is simple. Probe angles are measured from the normal to the plate surface. A bevel angle is measured from the same normal (the plate's thickness direction). A fusion face tilted β from vertical has its own normal tilted (90° − β) from vertical, so a beam refracted at (90° − β) meets that face square on.
| Bevel angle (each side) | Beam angle for normal incidence | Practical choice |
|---|---|---|
| 45° (single bevel) | 45° | 45° |
| 37.5° (common pipe and plate bevel) | 52.5° | 45° and 60°, or PAUT across both |
| 30° (60° included single-V) | 60° | 60° |
| 20° to 22.5° | 67.5° to 70° | 70° |
| 0° to 10° (narrow gap, J-prep wall, square edge) | 80° to 90° | Not reachable with a single probe; use tandem, PAUT or TOFD |
The ISO wording means "near normal" is the target, not an exact match, and a second angle is still used for flaws in other orientations, such as inter-run lack of fusion and transverse cracks. In AWS D1.1, the probe angle and scanning pattern for each thickness and joint come from a testing-angle table in the UT part of the code (Clause 8, Part F); check the table number in your edition. Read more on refraction and wedges in our guide to shear wave ultrasonic testing.
Key takeaway: the bevel angle sets the probe angle; a 70° probe on a 30° bevel is 10° off the face normal.
Which leg reaches which face?
On a single-V weld scanned from one side, the near fusion face (on the probe's side) faces away from the incoming direct beam. A beam aimed at that face square on has to come from below, after bouncing off the back wall: the second leg. The far fusion face is hit in the first leg, but only its lower part, because the upper part sits under the weld cap and the probe cannot get close enough. That is why procedures scan from both sides, or use both legs from one side, to cover both fusion faces.
Worked example: a 60° probe on a 30° bevel
Setup. 1 in. (25 mm) plate, single-V groove with a 30° bevel each side (60° included), zero root opening for simplicity. Target: sidewall lack of fusion on the near-side face at mid-thickness (0.50 in. deep). Probe: 60° shear wave.
Step 1: where the face is. At depth 0.50 in., the face is (1.00 − 0.50) × tan 30° = 0.50 × 0.577 = 0.29 in. from the weld centerline.
Step 2: sound path in the second leg. The beam travels down to the back wall, then back up 0.50 in.:
- first leg: 1.00 ÷ cos 60° = 2.00 in.
- second leg to the flaw: 0.50 ÷ cos 60° = 1.00 in.
- total sound path: 3.00 in.
Step 3: probe position. Surface distance = 3.00 × sin 60° = 2.60 in. So the beam index point sits 2.60 + 0.29 = 2.89 in. from the weld centerline.
Step 4: depth check. In the second leg, depth = 2t − (sound path × cos θ) = 2.00 − (3.00 × 0.50) = 0.50 in. That matches the target, so the geometry is consistent. Full skip distance is 2 × 1.00 × tan 60° = 3.46 in., which confirms the flaw lies inside the second leg.
What changes with a 70° probe. The 70° beam meets the same face 10° off its normal. A smooth planar reflector sends the echo back about 20° away from the probe, so far less energy returns. The flaw is still there; the screen may simply show less of it.
Why that matters for acceptance: an AWS D1.1 rating
AWS D1.1 rates each UT indication with d = a − b − c, where a is the indication level, b the reference level and c = 2 × (sound path in inches − 1) dB. A lower (more negative) d means a stronger reflector. Take a 3/4 in. statically loaded weld scanned at 70°, with a reference level b = 50 dB, and an indication at 2.0 in. sound path (c = 2 dB):
| Case | Indication level a | Rating d | Class (static, 5/16 to 3/4 in., 70°) | Result |
|---|---|---|---|---|
| Beam near normal to the flaw | 55 dB | 55 − 50 − 2 = +3 | A (+5 and lower) | Reject at any length |
| Same flaw, beam well off normal | 61 dB | 61 − 50 − 2 = +9 | D (+8 and higher) | Accept at any length |
The class bands in that row are taken from secondary reproductions of the D1.1 static table and must be confirmed against the licensed edition. The point stands without them: the same flaw can rate in the most severe class or the least severe class depending on how squarely the beam hits it. Probe angle selection is part of acceptance, not just detection. The full rating method is worked through in our guide to AWS D1.1 acceptance criteria, and the formula checks are collected in UT formulas.
What do PAUT and TOFD add?
PAUT sweeps many beam angles from one probe position and records an image; TOFD sizes the through-wall height of flaws from tip diffraction, largely independent of their angle. Neither replaces the need to think about the bevel.
Phased array (PAUT)
A phased array probe fires its elements in timed groups to steer the beam. A sectorial scan of, say, 40° to 70° includes the near-normal angle for most common bevels in a single pass, so a 37.5° bevel no longer forces a choice between 45° and 60°. Encoded scans also give a permanent record that a second reviewer can re-read.
- AWS D1.1:2020 added a PAUT annex. A vendor reproduction of that annex (Annex H) quotes it as "mandatory requirements that shall apply when phased array ultrasonic testing (PAUT) is used", calling for written procedures, advanced personnel qualification and PAUT-specific calibration (Sonatest summary).
- The same source quotes the acceptance clause as requiring that welds have "no planar indications" and no indications exceeding the annex's amplitude and length tables. Confirm the wording and the D1.1:2025 numbering before relying on it.
- For general weld PAUT practice, ASTM E2700-26 covers contact UT of welds with phased arrays.
Time-of-flight diffraction (TOFD)
TOFD uses a transmitter and receiver on opposite sides of the weld and times the weak signals diffracted from a flaw's upper and lower tips. Depth comes from diffraction geometry, not the pulse-echo formula. Because tip diffraction depends much less on flaw tilt than a mirror reflection does, TOFD is useful for exactly the off-angle planar flaws that conventional UT under-calls.
- ASTM E2373/E2373M-19(2024) states that TOFD is "not based on amplitude response", has proven effective on 9 to 300 mm (0.375 to 12 in.), and can be used with pulse-echo weld examination to improve sizing and to separate flaws from geometric reflectors.
- TOFD has dead zones near the scan surface and the back wall, so it is normally paired with pulse-echo or PAUT.
- Our rail-specific comparison is in PAUT vs TOFD for rail weld inspection.
| Technique | Main role for lack of fusion | Main weakness |
|---|---|---|
| Conventional angle-beam UT | Detection, with a probe chosen for the bevel | Single angle per pass; A-scan only; operator dependent |
| PAUT (sectorial or multi-group) | Detection across many angles, imaging, encoded record | Setup and focal law errors; still amplitude based |
| TOFD | Through-wall height sizing; detection less sensitive to tilt | Surface and back-wall dead zones; geometric noise |
| PAUT + TOFD together | Detection plus sizing in one encoded scan | Cost, procedure qualification, data review time |
How do AWS D1.1 and ASME treat lack of fusion?
The codes differ: visual and some UT rules reject lack of fusion by type, while AWS D1.1 Part F and ASME B31.3 UT rules judge it by amplitude and length like any other indication. This is a summary; the full comparison is in our guides to AWS D1.1 acceptance criteria and UT acceptance criteria across codes.
| Code and method | How lack of fusion is judged | Confidence |
|---|---|---|
| AWS D1.1, Table 8.1 (visual) | Complete fusion is required between adjacent layers and between weld and base metal | Secondary sources; confirm |
| AWS D1.1, Clause 8 Part F (conventional UT), Tables 8.2 static and 8.3 cyclic | Rating d = a − b − c sets Class A to D by thickness and probe angle; each class has a length limit. The flaw type is not named | Rating method verified; class bands from secondary sources |
| AWS D1.1:2020, PAUT annex | Welds acceptable only with "no planar indications" and within the annex amplitude and length tables | Vendor quotation; confirm |
| ASME Section VIII, Div. 1, Mandatory Appendix 12 (UT) | Indications above 20% of reference are investigated; cracks, lack of fusion and incomplete penetration are rejected at any length | Secondary sources; confirm |
| ASME B31.3, para. 344.6.2 (UT) | Rejected only if amplitude exceeds the reference level and length exceeds a wall-thickness limit; lack of fusion is not singled out | Forum quotations of the code text; confirm |
| ASME Section V | How to examine; contains no acceptance criteria | ASME committee page |
Three practical points follow.
- Under D1.1 Part F, an off-angle lack of fusion can pass. The worked example above shows why probe angle selection is an acceptance issue.
- Under ASME VIII Appendix 12, characterization decides the outcome. Calling an indication "lack of fusion" rather than "slag" rejects it at any length, so the UT procedure and the technician's characterization skills matter as much as sensitivity.
- B31.3 practitioners disagree on low-amplitude lack of fusion. In an NDT.net forum thread on B31.3 UT acceptance, experienced Level IIIs noted that a lack-of-fusion indication at 80% of reference passes the UT clause as written, while the radiography table rejects lack of fusion. The contract can impose stricter rules; the code as written does not.
D1.1 itself also doesn't require UT; the contract documents, AISC 360 or AISC 341 do. Check the code edition named in your contract before applying any table above: the current edition is AWS D1.1/D1.1M:2025, but many contracts still cite 2020.
What happens after a lack-of-fusion indication is found?
The indication is first evaluated against the governing code; only a rejectable indication is excavated, re-welded and re-examined. A UT signal is not by itself a reject.
- Confirm and characterize. Re-scan from the other side and with a second angle. Lack of fusion on a sidewall usually peaks sharply from one direction and fades from the other; porosity responds from most directions.
- Rate and size it under the governing code (D1.1 rating and length, ASME amplitude and length, or TOFD height where the contract allows).
- Report it. Record location, depth, length, rating or amplitude, probe angle and scan side, so the repair crew can find it.
- Excavate. TWI's guidance is removal "by localised gouging, or grinding, followed by re-welding as specified in the agreed procedure."
- Confirm removal before welding, often with MT or PT on the excavation.
- Re-weld and re-examine with the same method and procedure that found it.
- Fix the cause. If the defect recurs and arc blow is not to blame, TWI recommends amending the welding procedure or retraining welders.
What UT can't do: the limits
UT is the most capable common method for lack of fusion in welds, but it cannot guarantee that a weld is free of it.
| Limit | Why it happens | What helps |
|---|---|---|
| "Pure" lack of fusion (faces in contact) | Too little gap to reflect sound; Rihar reports it can't be detected by NDT | Welding control and visual checks during welding |
| Large but weak indications | Faces stick together in places; flaw tilted away from the beam | Correct probe angle; PAUT sectorial scan; TOFD |
| Vertical sidewalls (narrow gap, J-preps) | No single probe reaches 80° to 90° | Tandem techniques, PAUT, TOFD |
| Near-surface flaws under the cap | Probe can't get close enough; TOFD surface dead zone | Scan from both sides; second leg; surface methods |
| Thin material below 5/16 in. | Outside the D1.1 Part F range | A procedure qualified for that thickness |
| Coarse-grained or austenitic welds | Scatter and beam skew | Specialized probes and procedures |
| Operator dependence | Conventional UT leaves no image | Encoded PAUT records; Level III procedure review |
The honest summary is that UT is usually the right first method for lack of fusion, and the best results come from the right angle, both scan sides and, on critical welds, PAUT and TOFD together.
How EMA Industries applies this
EMA Industries performs AWS D1.1 structural weld UT, MT and PT, and provides NDT Level III services and WPS/PQR support. We're based in Southern California and mobilize nationwide. On a groove-weld job, the scan plan should record the bevel angle and the probe angles chosen for it, both scan sides, and the code edition and loading category, so a reviewer can see that both fusion faces were covered. To scope an inspection, see our ultrasonic testing services or contact EMA Industries.
Related reading: Weld defect types · Porosity in welding · Slag inclusion in welding · NDT for welding inspection · What is ultrasonic testing?
Frequently asked questions
What does lack of fusion mean in welding?
Lack of fusion means the weld metal did not bond with the base metal along the joint sidewall, with a previous pass, or at the root face. It leaves a thin, planar unbonded area inside the joint that can act like a crack under load.
What is the difference between lack of fusion and lack of penetration?
Lack of fusion is a failure to bond with an adjacent surface, such as the sidewall or a previous bead. Lack of penetration (incomplete joint penetration) is a failure of weld metal to extend through the joint thickness, usually at the root. Both are planar, but they sit in different places and need different beam angles.
Which probe angle is best for detecting sidewall lack of fusion?
The angle that meets the fusion face square on, which is about 90° minus the bevel angle. A 30° bevel calls for a 60° probe, and a 45° bevel for a 45° probe. Where no single angle fits, PAUT sectorial scans or TOFD fill the gap.
Can radiography detect lack of fusion?
Radiography can detect lack of fusion when the radiation beam runs along the flaw plane or when the flaw contains voids or inclusions. Tight, tilted sidewall lack of fusion is often missed, which is why TWI rates UT as normally more effective for inter-run fusion.
Is lack of fusion always a reject?
It depends on the code and method. ASME Section VIII Appendix 12 rejects lack of fusion at any length, and AWS D1.1 visual criteria require complete fusion. AWS D1.1 Part F UT and ASME B31.3 UT judge it by amplitude and length, so a weak indication can pass.
How is lack of fusion repaired?
The flaw is removed by grinding or gouging, the excavation is checked, and the joint is re-welded to the qualified procedure. The repair is then re-examined with the method that found the flaw, and the welding cause is corrected if it recurs.
Sources
- Weld defects/imperfections: lack of sidewall and inter-run fusion (Job Knowledge 41), TWI
- How can lack of fusion defects be detected and remedied?, TWI FAQ
- Lack of fusion in welded joints, G. Rihar, Welding Institute, Ljubljana, hosted on NDT.net
- LOSWF: lack of side wall fusion, British Institute of Non-Destructive Testing (cites EN ISO 17640 normal-incidence requirement)
- AWS announces release of D1.1/D1.1M:2025, American Welding Society, April 17, 2025
- PAUT of welds to 2020 AWS D1.1, Sonatest (quotes D1.1:2020 Annex H)
- ASTM E2373/E2373M-19(2024), Use of the Ultrasonic Time of Flight Diffraction (TOFD) Technique, ASTM International
- ASTM E2700-26, Contact Ultrasonic Testing of Welds Using Phased Arrays, ASTM International
- ASTM E164-24, Contact Ultrasonic Testing of Weldments, ASTM International
- ASME B31.3 UT acceptance criteria: lack of fusion accepted?, NDT.net forum (practitioner quotations of para. 344.6.2; not the code itself)
- AWS D1.1/D1.1M:2020 and :2025, Clause 8 and Tables 8.1 to 8.3; ASME BPVC Section V and Section VIII Div. 1 Mandatory Appendix 12; ASME B31.3 para. 344.6.2 (paywalled; values above from secondary sources and to be confirmed against the licensed text)