Lack of fusion is a bonding failure, not a missing-metal failure. Filler metal is present and
the joint can look completely full on a cross-section; what is missing is the metallurgical
bond, because the surface it landed on — a joint sidewall, the root, or the previous pass —
never reached fusion temperature before the new metal solidified against it. The result is a
planar, often very tight discontinuity that behaves like a pre-existing crack under load rather
than like a rounded, less severe imperfection.
The two conditions that produce it both come down to arc energy arriving at the wrong place.
Heat input below the qualified window fails to raise the sidewall or
the prior pass to fusion temperature at all. Travel speed above the qualified rate lets the
weld pool run ahead of the melting front, so new filler is laid over
material that looks wetted from above but was never actually melted. Torch angle, weave
pattern and, on mechanised work, joint-tracking drift can produce the same failure locally even
when average current and voltage readings look correct, because they aim the arc’s energy away
from the surface that needed it.
Because the joint face and the parent material can appear fully bonded from the top, visual
inspection is the weakest route for finding lack of fusion — it is the reason ultrasonic and
phased array testing lead detection wherever the imperfection can occur, in contrast to
surface conditions such as undercut that a visual check finds directly.
Root-region lack of fusion is closely related to insufficient penetration:
both trace back to arc energy per unit length that was too low for the joint actually being
welded, which is why procedures compute arc energy from measured current, voltage and travel
speed rather than trusting the panel settings, and why
interpass temperature is checked alongside the arc settings
on multi-pass work.