What Lack of Fusion Is and Why It Hides

Why this matters

Lack of fusion is the defect that does not look like one. The bead is full, the profile is right, the toes are tied in, and a shop that inspects by looking has no way to know the deposit is sitting against a surface it never melted. It is a bonding failure rather than a filling failure, it tends to be produced by exactly the habits that make a weld look good, and it is a flat crack-like discontinuity rather than a rounded one. That last point does two things: acceptance criteria treat it far more harshly than the same volume of porosity, and the inspection method you already own may be pointed in a direction that cannot see it.

Before removing a defective weld: air carbon arc gouging is the usual removal method and it is a different exposure from welding. It generates sound levels that put the work inside a hearing conservation program under 29 CFR 1910.95, ejects molten metal several feet, and produces a heavier fume load than the weld it is removing, an inhalation route needing local exhaust at the arc or a respirator issued under a written program meeting 29 CFR 1910.134. Grinding a defect out releases base metal and any coating as respirable dust by the same route, with capture at the tool. Where the defect was found by radiography, the shot area is a controlled area under 29 CFR 1910.1096 and no welding or gouging happens inside it while a source is exposed. Fire prevention and the fire watch belong to the hot work permit under 29 CFR 1910.252(a) or 29 CFR 1926.352 in construction.

What it is, stated against the three things it is not

Incomplete fusion is weld metal lying against a surface - a groove face, a previous pass, or the parent metal at the root - that did not melt. The two surfaces are in contact. They are not joined.

• Not incomplete penetration. Incomplete joint penetration means the weld did not reach the depth the joint required, so unfilled joint is left. Lack of fusion occurs in joints filled to the correct depth. One is a geometry shortfall, the other a bonding shortfall, and a report that uses them interchangeably sends the wrong correction.
• Not porosity. Porosity is gas that came out of solution in a pool that did melt everything it touched. The metal bonded; it just has voids in it.
• Not a fit-up problem, though poor fit-up causes it. A wide root opening is visible and correctable before welding. Lack of fusion is what that opening produces afterwards, once it is buried.

Where it lives

  section through the left bevel of a groove weld

  radiation shot straight down
    | | | | | | | |
    v v v v v v v v
  ----------------------------- plate top
   plate \
          \  unfused bevel face:
           \  a thin flat plane
   weld     \  lying along the bevel
    metal    \  an angled sound beam hits
              \  it nearly square, a shot
               \  from above crosses it
  ----------------------------- plate bottom

Three positions account for nearly all of it. Against the groove face, where the arc rode the centre of the joint and the pool wetted the sidewall without melting it. Between passes, where a bead was laid over a previous one whose surface was cold, slag-covered or convex. At the root, where the arc never reached the bottom with enough energy to melt both faces.

What produces it

Every entry below is a version of the same condition: energy went somewhere other than the surface that needed to melt, or the surface was not available to melt.

• Arc off the sidewall. Manipulation that keeps the arc in the middle of the pool leaves the groove faces heated by the liquid rather than by the arc, and liquid weld metal does not carry enough energy to melt solid parent metal it is only touching.
• Filler outrunning energy. Wire speed high relative to current and travel, so the pool is large and cool and floods ahead of the arc. The bead is fat and convex and the sidewalls under it are untouched.
• Short-circuit transfer on heavy sections. The mode is low-energy by design, which is what makes it useful on thin material and open roots and what makes it the classic producer of cold lap on thick joints. A mode limit, not a wire fault.
• Convex previous pass, or slag left on it. A crowned bead leaves two sharp valleys at its toes and the next pass bridges the crown without reaching into them; slag insulates, so whatever sits under it does not melt at all.
• Arc blow. An arc driven off course by residual magnetism in the work or by return lead position spends its energy off the intended face.
• Too small an included angle or too tight a root, so putting the arc on both faces is physically impossible.

Most of these produce a smooth, full, well-shaped bead. A convex overloaded bead is the one visual cue, and it is a weak one.

Why the shape of the flaw matters more than its size

A pore is roughly spherical, load flows around it, and the stress rise at its boundary is modest. A lack-of-fusion area is a flat plane with an effectively zero-radius edge, so it behaves like a pre-existing crack: it concentrates stress sharply at its perimeter and gives a fatigue crack a running start. Two discontinuities of identical volume, one round and one flat, are not comparable defects.

That is why code acceptance criteria give porosity a size and aggregate-length allowance and generally do not permit incomplete fusion at all in a code weld. The limits belong to the governing document - AWS D1.1 in the edition your specification or adopted building code invokes for structural steel, or ASME Boiler and Pressure Vessel Code Section IX with the applicable piping or vessel code for pressure work, each binding through that adoption rather than on its own.

What each inspection method can actually reach

Worked example: a splice that passed radiography and failed ultrasonics

A shop welded a full-penetration groove in 3/4 inch plate, single V, bevel face tilted 30 degrees off vertical. The joint was radiographed with the source directly over the weld, shooting straight down to film underneath, and it passed. A later ultrasonic examination with a 60 degree angle beam reported a sidewall discontinuity about 0.3 inch across the bevel face.

Why the film could not have shown it. Radiography detects a flaw by the change in absorption path the beam sees. Take the unfused plane as an illustrative 0.002 inch opening, 0.3 inch across the face, which are ordinary figures for a tight fusion flaw rather than measured ones. A beam running straight down crosses that plane at 30 degrees, so the void length along the beam is the opening divided by the sine of that angle:

• 0.002 divided by sin 30 degrees, which is 0.5, gives 0.004 inch of void along the beam.
• Against 0.75 inch of plate, that is 0.004 divided by 0.75, or about 0.53 percent of the radiographic path.

Correction, printed: the detectability benchmark is already a percentage of thickness. Radiographic sensitivity is set by the image quality indicator the technique uses, and a common requirement is 2 percent sensitivity, meaning a change of 2 percent of part thickness must be visible on the film. That 2 percent is not a target to add anything to, it is the threshold the 0.53 percent is compared against. At 0.53 against 2, the flaw produces roughly a quarter of the smallest change the technique was qualified to show. The film is not defective and the radiographer did nothing wrong; the technique was pointed the wrong way for this flaw.

What a differently aimed shot would have seen. Rotate the source so the beam runs along the bevel face rather than across it, and the void length along the beam approaches the full 0.3 inch width. Against 0.75 inch of plate that is about 40 percent of the path, far above the 2 percent benchmark and unmissable. Same flaw, same film, same energy, purely a geometry difference.

Why the angle beam found it. A 60 degree shear wave, whose direction matches the normal to a face tilted 30 degrees off vertical, strikes that plane square. A planar flaw struck square is the best reflector in ultrasonic testing, which is the inverse of radiography's weakness and the reason ultrasonic examination is the normal method for fusion-face flaws in thick joints.

Check against the sibling rules, with the figures printed. The heat-input card states that arc energy figures compare two procedures only at constant thickness and joint geometry; this example quotes no heat input and attributes the flaw to arc placement on the sidewall rather than to an energy figure. The undercut and overlap card owns toe geometry at the weld face; this flaw is on the fusion face inside the joint, so no toe claim is made. The 0.53 percent against 2 percent comparison above concerns radiographic detectability, not acceptance, and the porosity allowance is cited rather than restated.

The failure mode this produces. Not a leak and not a break on the day. A splice like this carries static load for a long time. Cycling is what kills it: the flat flaw's perimeter is a fatigue initiation site, the crack grows from it into sound metal over months, and by the time anything shows on the surface the remaining section is already reduced. That is why code treats planar flaws as unacceptable rather than as allowable up to a size.

What would flip the recommendation. Take the same shop, the same welder, and a 3/16 inch lap joint. Angle-beam ultrasonic examination of thin material is unreliable because near-field effects and multiple reflections swamp the signal, and radiography works better on thin sections because 2 percent of a small thickness is a small absolute change. On thin work the answer moves back toward radiography plus fit-up control plus a procedure that keeps short-circuit transfer inside its useful range.

How to verify you are not producing it

• Watch the arc, not the pool. If a welder can describe where the pool is but not where the arc is landing on the groove face, the arc is riding the middle. That is a coaching item with a specific correction and it is worth more than any downstream inspection.
• Reject a convex pass before the next one goes on it. A crowned bead is the one visible precursor. Grind it flat, with capture at the tool for the dust, rather than burying the valleys.
• Match the inspection method to the flaw you are worried about, and say which flaw that is. An inspection specified as "NDE per the specification" with no method reasoning is how a shop gets a clean report on a joint carrying a planar flaw.
• Confirm the ultrasonic technique was written for your joint. Angle, scanning surface and access belong to a written procedure, and an angle beam that cannot reach the far bevel face reports nothing while looking identical on paper to a clean joint.
• Where the joint is thick, restrained and critical, say on the report that a surface-only inspection is an incomplete answer.

References

• AWS D1.1 structural welding code for steel, in the edition invoked by your project specification or the adopted building code, for incomplete fusion acceptance criteria and inspection requirements
• ASME Boiler and Pressure Vessel Code Section IX and the applicable piping or vessel code, as adopted by your jurisdiction, where the work is pressure-retaining
• 29 CFR 1910.95 for occupational noise exposure during air carbon arc gouging; 29 CFR 1910.134 for respiratory protection; 29 CFR 1910.1096 for radiographic controlled areas
• 29 CFR 1910.252(a) and 29 CFR 1926.352 for hot work fire prevention during weld removal and repair
• See related: Why Porosity Appears and What It Tells You; How Undercut and Overlap Each Form; What a Weld Actually Is and What It Is Not