What Thermal Cutting Does to the Edge You Are About to Weld

Why this matters

A thermally cut edge is not a clean surface waiting for weld metal. It is a surface that was melted or burned away and then quenched by the cold plate behind it, so it arrives at the joint carrying a thin hard layer, an oxide skin, whatever gas the cutting jet dissolved into it, dross on the underside, and a taper that means the angle you set is not the angle you got. Every one of those changes what the first pass does. The allowance for removing them is a metallurgical requirement sized to the cutting process, and a shop that treats it as tidying up produces defects it will later blame on the welder.

Before cutting anything: never cut or weld on a drum, tank or pipe that has held a flammable or its residue until it has been cleaned or inerted, which is the requirement at 29 CFR 1910.252(a)(3), and fire prevention, spark containment and the fire watch for the cut belong to the hot work permit under 29 CFR 1910.252(a) in general industry or 29 CFR 1926.352 in construction. Plasma cutting generates a heavy fume plume plus ozone and nitrogen oxides, an inhalation route that needs a downdraft table, a water table or local exhaust at the cut, with a respirator issued under a written program meeting 29 CFR 1910.134 where capture is not achievable; cutting galvanized coating adds zinc oxide, old paint can add lead under 29 CFR 1910.1025 or 29 CFR 1926.62, and stainless adds hexavalent chromium under 29 CFR 1910.1026 or 29 CFR 1926.1126. Plasma cutting is also loud enough to put the work inside a hearing conservation program under 29 CFR 1910.95. A plasma torch carries a high open-circuit voltage, so change consumables only with the machine off and isolated per the equipment manufacturer's instructions, and keep gloves and the work area dry. Cutting throws molten dross several feet and it goes under things, so clear the floor below and check it after.

What the cut face is made of

   plasma kerf through plate, section view

   torch above
   top of kerf   |<-- wider here
                 \
                  \    what is on the face:
                   \    thin hardened layer
                    \   gas-enriched skin
                     \  oxide film
   bottom of kerf     |<-- narrower here
                      dross clinging underneath

A hardened layer. The cut face reached melting and then had the whole cold plate behind it acting as a quench. On plain low-carbon steel the layer is usually harmless. On steel with a higher carbon equivalent, and on anything already hardened, that thin layer can transform to a hard structure under tension, which is why the same detail cuts fine on one grade and shows edge cracking on another.

An oxide film. Oxyfuel cutting is an oxidation reaction, so an iron oxide film is not a side effect of the process, it is the process. Oxide does not fuse the way clean metal does.

Dissolved gas from the cutting jet. Air and nitrogen plasma leave a nitrogen-enriched skin on the cut face. Nitrogen dissolved in the pool and rejected on freezing is a porosity source, and the sibling card on porosity owns how to read the resulting pores. It also affects the metallurgy of stainless and aluminium weld metal.

Dross. Re-solidified metal and oxide hanging on the underside of the cut. It is a physical obstruction to fit-up and a contaminant if it ends up in the joint.

Taper. A plasma kerf is wider at the top than the bottom. That is inherent to the process, not a machine fault, and it means the bevel angle you cut differs from the bevel angle you set.

What each process leaves you

Two entries in that table surprise people. Waterjet is the only one on the list that hands you an edge with no thermal history at all, which matters most on materials where any heat is a problem. And a machined edge is not automatically clean: cutting fluid is oil, and oil in the arc is a hydrogen source that the hydrogen card owns.

The prep allowance, and where the number comes from

The removal allowance is a depth of metal taken off the cut face before welding, and its owner is the governing document plus the material. Where the code or the specification states a surface roughness limit, a notch depth limit or a required removal on cut edges, that document owns the number in the edition your project invokes, and AWS D1.1 in the adopted edition is the usual home for structural steel. Where it does not, the practical driver is which of the four conditions above actually exists on your material: dross always comes off, a gas-enriched skin comes off wherever nitrogen pickup matters, and a hardened layer comes off wherever the grade can harden.

Notches are the separate case. A gouge from a torch stall or a plate slip is a stress raiser in the finished member, not just a cosmetic mark. Governing documents limit their depth and require them to be repaired by grinding to a gradual taper, or by welding and grinding flush where deeper. Route the acceptable depth and the repair method to the adopted code, and never dress a notch by simply blending its mouth, which leaves the root.

Worked artifact: an edge acceptance record, three edges, one job

Same job, three edges, the record filled in as each was inspected before fit-up.

Edge A, oxyfuel-cut bevel, 3/4 in carbon plate. Oxide film present. Dross present along the full underside. No notches, no visible taper. Grade is ordinary structural carbon plate with a low carbon equivalent, so no hardened layer of consequence. Allowance applied: dross removed by chipping and grinding, cut face wire-brushed and lightly ground to bright metal. Accepted.

Edge B, air-plasma-cut bevel, 1/2 in carbon plate. Dross light. Gas-enriched skin present, since the cutting gas was air. Taper measured on the kerf at 3 degrees. Allowance applied: dross removed, cut face ground back nominally 1/32 in to take out the gas-enriched skin. Accepted after grinding.

The taper is the finding on Edge B, and it is a dimensional one. A 3 degree taper across a 1/2 in cut depth is a width difference of tan 3 degrees times 0.5 in, which is 0.0524 times 0.5, or 0.026 in top to bottom. On a single bevel that is a 3 degree error in the groove face angle. Bring two of those together into a double-bevel joint and the included angle is off by up to 6 degrees from what the torch was set to, before anyone has looked at a fit-up gauge. The joint preparation card owns the four numbers that a groove is judged by and the tolerance each one carries; the point here is that the groove angle on a plasma-cut bevel has to be measured on the part rather than read off the machine setting.

Edge C, sheared edge, 1/4 in plate. Burr present, cold-worked band along the sheared face, no thermal layer. This edge was going into a load-carrying fillet weld, so the allowance applied was to grind the sheared band off to bright, undistorted metal and remove the burr. Accepted after grinding.

What the record is for. Three edges, three different conditions, three different allowances, and none of them is "grind it a bit." The record ties each allowance to the condition that justified it, so when a defect shows up two operations later somebody can tell whether the edge was the cause or was already ruled out.

Check the record against the rules the siblings state

• Taper arithmetic re-derived. tan 3 degrees is 0.0524; times a 0.5 in cut depth is 0.026 in. Doubled for a two-sided joint, the included angle error is up to 6 degrees, which is 2 times the 3 degrees measured, not an independent figure.
• Each allowance is tied to a condition that was actually observed. Edge A: dross observed, oxide observed, no hardened layer expected on a low carbon equivalent grade, so no depth removal claimed. Edge B: gas-enriched skin from air plasma, nominal 1/32 in removal. Edge C: cold-worked band, no thermal layer, grind to bright metal. No edge got an allowance for a condition it did not have.
• No number in the artifact was introduced that the sections above did not state. The 3 degree taper is measured on the part, the 1/32 in is a nominal removal stated as nominal, and the code-owned figures (roughness limit, notch depth) are routed rather than invented.
• Direction check against the siblings. The porosity card has nitrogen pickup producing pores and names base metal condition as a constant source. Edge B's gas-enriched skin is the same relationship at its origin, and the correction here is removal of the skin rather than more shielding gas, which is the correction the porosity card explicitly warns against reaching for first.

References

• AWS D1.1, structural welding code for steel, in the edition invoked by your project specification or the building code your authority having jurisdiction has adopted, for thermal cut surface quality, notch limits and repair of cut edges
• 29 CFR 1910.252(a) and 29 CFR 1926.352 for hot work fire prevention, including 1910.252(a)(3) on containers that have held flammables
• 29 CFR 1910.134 (respiratory protection), 29 CFR 1910.1026 and 1926.1126 (hexavalent chromium), 29 CFR 1910.1025 and 1926.62 (lead), 29 CFR 1910.95 (hearing conservation)
• See related: Why Joint Preparation Decides More Than Technique; Why Porosity Appears and What It Tells You; Why Grinding and Gouging Are Part of the Weld