Defect guides

Running too cold

The only defect on this site that can look good and be worthless.

Read this one differently. Every other guide here describes something you can see. A cold weld's defining feature is that its worst consequence — lack of fusion — is invisible from above. A skilled welder running too cold can lay a bead that looks tidy, even attractive, and is barely attached to the plate. This is the failure mode that a photograph, an app, and your own eyes are all worst at catching.

What it looks like — when it looks like anything

The visible signs, when they're present at all:

And on a bad day, none of the above. A cold bead run with steady hands can have even ripples, consistent width and a clean surface. That is the trap.

What causes it

Amperage or voltage too low

The direct cause. Often the result of deliberately backing off after burning through something, or of settings copied from a chart written for different thickness. Worth checking against the manufacturer's range for your electrode diameter and material thickness before assuming technique.

Travel speed too fast

Correct machine settings, insufficient dwell. The arc passes over the joint too quickly to melt the parent metal, so filler is deposited onto steel that never reached melting point. This produces some of the most convincing-looking cold welds, because the bead itself can be perfectly regular.

Excessive stickout (MIG/flux-core)

Amperage in wire processes is a consequence of wire feed and arc length, not something you dial in directly. Long stickout raises resistance in the wire and drops the current actually reaching the arc, so the weld runs cold even though nothing on the machine changed. If your welds went cold when the joint got harder to reach, this is why.

Wrong joint access and gun angle

If the arc can't reach the root of the joint — poor prep, insufficient gap, an angle that buries the arc against one face — filler stacks up in a corner it never fused into. Common on fillet welds with a tight included angle.

Heat sinking into the parent metal

Thick section, cold ambient temperature, or clamping to a large steel bench all draw heat out of the joint. Settings proven on 6mm plate in summer can genuinely be too cold on 20mm in an unheated workshop in February. Thick material may require preheat as part of the procedure.

How to tell — properly

Since the visible signs are unreliable, use destructive testing on practice coupons. This is what practice coupons are for, and it is the only honest way to calibrate whether your settings are producing fusion:

Break one coupon from every session where you changed settings. It costs one plate and tells you what no photograph can.

Why this is the dangerous one

A weld that is too hot is ugly, and ugly gets fixed — someone sees it, it gets ground out and rerun. A weld that is too cold can pass a casual visual check, get painted, get installed, and hold nothing. The joint has a fraction of its designed strength and no outward sign of it.

This is also precisely why nobody should treat a photo-based score — from this app or any other — as evidence that a weld is sound. Fusion is the property that matters most and the one least visible from above.

What a photo can and can't tell you

A photograph can flag the correlates of a cold weld: a high narrow profile, a steep toe angle, visible overlap, clinging slag. Those are real signals and worth acting on when they appear.

It cannot detect lack of fusion, and no amount of image quality changes that — the defect is at the boundary between the bead and the plate, underneath metal that is in the way. A clean-looking cold weld will score well on surface appearance and be a bad weld. That is a limit of the method, not a bug to be fixed in a later version.

If the joint carries load, it needs a bend test, RT or UT, and a Certified Welding Inspector. Not a photo. Not a score. Not an app.

Related