Plasma Cutting Precision for Structural Steel Weld Quality

A 20 mm column flange arrives at the fit-up bay, plasma-cut, and the fitter discovers the edge is 3 mm off square, the root face is uneven, and there is a line of grey dross along the top. Two hours later the same joint is on a second column that was cut on the same machine the same afternoon. The difference wasn’t the steel or the welder — it was the cut. Get the edge right and the weld almost welds itself; get it wrong and you pay for grinding, re-fitting, preheat, and rework.

Why Plasma Cutting Precision Decides Structural Steel Weld Quality

The argument this article rests on is simple: a plasma cut is the first weld operation. Not the first welding operation exactly, but the first process on the plate that behaves like one. It melts metal, forms a fusion boundary, leaves a heat-affected layer and creates a surface a welder must fuse to. Everything at fit-up, the arc and inspection is downstream of that cut.

So every shop eventually asks one question out loud: how do you get cut edges precise enough that the downstream groove weld lands correctly? Not precise in the abstract. Precise enough that root opening stays uniform along the joint, filler metal goes where it belongs, nobody re-fits the plate, and an inspector signs the joint without raising a repair order.

Careers hinge on the answer: fabricators pricing work by the kilo and eating the rework, welders blamed for joints that fought them from the first clamp, QC techs with squares and angle gauges at the cut table, shop engineers writing the WPS and bevel details, inspectors accepting or rejecting the result. All share one input: the edge somebody plasma cut half an hour ago.

The chain of dependency

Cut quality is a bundle of properties, each landing somewhere specific downstream.

• Angularity — a bevel off the drawing angle changes root opening along the joint.
• Kerf width — the slot the plasma jet removes, shifting hole size, nesting yield and fillet leg length.
• Dross and oxide — re-solidified metal that must come off before anyone can weld the joint.
• Root face consistency — how evenly the bevel lands across the plate thickness.
• Edge straightness — whether a straightedge sits flush along the whole edge.

Those five feed the fit-up gap, root opening, weld metal volume, preheat and interpass temperatures, distortion during assembly, and whether the joint passes or gets cut out. A 1.6 mm error at the edge does not stay 1.6 mm; it becomes a bigger error downstream, where it costs more to fix and is harder to fix.

Every trade downstream of the torch cares about three things. How fast the part fits, because a square edge closes up and clamps without shimming. How much grinding the fitter does before the arc comes on, since slag removal burns consumables, adds heat to thin material and tires the operator. And whether the inspector accepts the joint once the weld is in. Those are the yardsticks this guide measures against.

What plasma changed in structural fabrication

Three things. Repeatability: a CNC table with torch height control puts the same cut on every plate, where a hand torch leaves a human signature on every cut. Machine bevel geometry: bevels cut in machine coordinates reach fit-up with a consistent angle and root face, not a tapered, hand-swept edge. Throughput: plasma cuts heavy plate far faster than oxy-fuel, on material a laser would need cleaned first. As the AWS Welding Digest reported in May 2024, plasma systems cut rusty, mill-scale-covered or coated conductive material that would require cleanup before a fiber laser.

Cutting is also a welding-adjacent hazard, not a separate one: oxy-fuel and plasma cutting both melt metal and generate airborne metal fume, so they fall under the same federal safety regime as welding and cutting operations, under OSHA 1910.252 and in the agency’s welding fact sheet.

Precision is a system, not a number

Set expectations honestly before hunting for a figure. Precision is a system property, not a spec-sheet number. Machine, calibration, torch, consumables, gas, pressure, amperage, travel speed, material condition, operator technique: all multiplied together. A headline tolerance describes the best case on a coupon, in a controlled shop, with fresh consumables. It does not describe your Tuesday.

The code governing the weld hands you no cut tolerance. AWS D1.1/D1.1M:2025-AMD1 is the current Structural Welding Code, Steel, covering carbon steel and low-alloy plate and shapes from 1/8 in to 8 in used in building structures, as committee members explained to The Fabricator. It is voluntarily adopted in the United States, and AWS does not run a certification program the way the Canadian Welding Bureau administers CSA W47.1 (CWB comparative review). So the cut-edge tolerance you accept is a technical and commercial agreement, written into the drawing and purchase order, then defended or attacked at inspection.

This article covers the unglamorous parts: the physics of the kerf, the difference between a 0.64 mm edge and a 1.6 mm edge, the dross that must come off, the bevel geometry the code expects, and the fit-up that follows. For background on the process feeding that groove, see why stick welding still wins. When the edge itself is the problem, carbon arc gouging is still the classic way to take a bad edge back to sound metal.

How Plasma Cutting Actually Works: Kerf, Standoff and the Heat-Affected Edge

To get precision out of a plasma table, you have to understand what the torch actually does to the steel.

Plasma cutting starts with a DC arc struck between a tungsten electrode inside the torch and the workpiece. Gas pushed through a small orifice ionizes as the arc passes through it, and ionized gas is plasma: a hot, electrically conductive gas that carries the arc. The jet does not burn the steel, it melts it, and its momentum plus the oxidizing gases in the stream blow the molten metal out of the cut.

Three things must stay in balance: the arc melting the metal, the gas carrying heat into the plate, and the jet ejecting the melt. Drift apart and you get dross, drag or a tapered edge.

Kerf: the metal the jet removes

Kerf is the amount of metal removed by the plasma jet, and it is not a fixed number. Kerf widens as plate thickens and amperage rises, because more current produces a larger jet that must travel further through the material and spreads more energy sideways on the way. Hypertherm’s guide to plasma accuracy makes the same point: plate thickness and current setting are the two biggest drivers of kerf width on a given machine.

Three reasons a welder cares about kerf:

• Nesting. Nested parts only fit if the kerf is repeatable; a kerf that grows 1 mm across the plate leaves the last part off its mark.
• Bolt holes. Every hole has its own kerf, so position, size and edge-to-edge distance shift with it. A hole cut 2 mm oversize means a reamed hole or a rejected plate.
• Fillet weld legs. A plate edge 1 mm inside the line shortens every fillet leg along it, and a short leg is an undersized weld a UT technician will find.

A kerf wrinkle specific to bevels: the effective thickness the jet must travel through is higher than on a straight cut, and it gets higher as the bevel angle becomes shallower. That changes the amperage and travel speed needed for good cut quality, so never run a bevel on straight-edge parameters.

Standoff: the distance that changes everything

Standoff is the distance between the torch nozzle and the work surface. Too close and the nozzle can touch the plate, damaging the consumable, pitting the workpiece and stalling the arc. Too far and the arc stretches, the jet loses focus and spreads, the cut face oxidizes and roughens, and the cut angle drifts positive or negative.

On a hand torch the operator holds the distance by eye and feel. On a CNC table the torch rides on a lift, and torch height control, the Z-axis function, automatically adjusts torch-to-work distance as consumables wear, holding the squareness and quality of the cut. Without it, cut angles and edge quality get worse, and so do grinding costs at the fit-up bench.

The heat-affected edge

Plasma-cut steel has no machined edge. Three layers matter: a thin oxide and mill scale skin, a lightly melted and re-solidified layer just beneath it, and a deeper heat-affected zone where the microstructure changed. On carbon steel the skin is typically blue-black magnetite scale. Near the edge the steel can be locally harder or more crack-sensitive, depending on how quickly it cooled. That is why cut edges must be cleaned before welding and some joint faces gouged or ground back to sound metal. AWS D1.1 treats cut surface condition as part of the fabrication deliverable. The arc also makes this hot work: OSHA 1910.252 sets general requirements for welding and cutting, including fume control and fire precautions.

Consumables: the wear items in the torch

The electrode, nozzle, shield cup and swirl ring are wear items that wear at different rates. The electrode erodes and its tip shape changes, so the arc starts to wander and the cut angle follows it. The nozzle orifice enlarges, the jet loses focus, the kerf widens, dross increases and the standoff window narrows. The shield cup collects spatter until it compromises shielding gas around the nozzle.

Gas: type, pressure and flow

Gas pressure, gas flow and the type of plasma gas affect cut quality and consumable life directly. Air is inexpensive and works well on carbon steel. Oxygen adds an exothermic reaction that speeds up cutting and improves the edge on thick carbon steel. Nitrogen gives a bright, nearly oxide-free edge on stainless steel but costs more. Adding argon can improve plasma arc stability and cut quality.

Too little pressure or flow gives a weak jet, poor ejection, dross and erratic piercing. Too much gives a turbulent, unstable jet and faster consumable wear. Moisture in the line causes erratic arcs and tip erosion. For background on the gases, see Air Gas Welding Explained: Process, Equipment, Safety and AllGas Welding Supply: Gases, Services and Safety Guide.

The gas makes the arc, the arc makes the melt, the jet removes the melt through a kerf of a certain width, and standoff decides whether that kerf stays put or wanders. Change the standoff, current, gas pressure or consumable condition and you change the edge, and every deviation the fitter finds later was created here.

What Precision Actually Means: Tolerance, Angularity and Cut Quality Standards

“Within tolerance” is not a sentence a customer and a fabricator both mean the same way. A plasma-cut plate can pass every dimensional check on the drawing and still hand the welder a problem. Prevent that argument with two separate vocabularies, written down separately.

Dimensional tolerance is not cut quality

Dimensional tolerance answers where the part is: hole position, hole size, overall length, the layout of cut lines. You check it with calipers, tapes and templates, and it is a clean pass or fail.

Cut quality answers what the edge itself is like. As Hypertherm frames it, cut quality comes down to three things: angularity, how square the cut line stands to the plate; edge finish, the condition of the top edge and the bottom of the kerf; and dross, which covers high-speed dross and low-speed metal spatter.

Neither one implies the other. A CNC table can place a hole dead centre and still leave a three-degree taper on the edge: hole position is a machine axis problem, taper is a torch, gas and heat problem.

Angularity: the taper that quietly changes your weld

Angularity is the deviation of the cut face from perpendicular to the plate surface, reported as an angle. It carries a sign: positive when the face leans in the direction of travel, negative when it leans back.

On a bevel, taper changes the root opening along the joint. Where the face leans out the gap grows; where it leans in it closes. Weld metal volume follows, so pass count, preheat demand and distortion shift with it. On a 25 mm bevel, three degrees of taper is roughly 1.3 mm of root opening change from one end of the edge to the other. That is the difference between a joint that fills cleanly and one the welder bridges over with extra passes and extra heat.

The vocabulary of a thermal-cut edge

• Cut edge deviation: how far the cut face departs from nominal geometry, measured per ISO 9013:2002, Thermal cutting — Classification of thermal cuts — Quality specifications and measurements, over a stated measurement range.
• Top-edge rounding: the nose of the cut rolls over instead of presenting a sharp corner. Fine on a plate edge, unwelcome on a groove face.
• Bottom-edge drag: the cut lags behind the torch on the underside, leaving a downward-facing step.
• Spatter: low-speed metal thrown out of the kerf and adhered as loose flakes, mostly on the top edge.
• Dross: re-solidified, oxidised metal left inside the kerf, usually a hard, brittle ridge on the bottom edge at high travel speed.

ISO 9013 is the international reference for all of this. It defines how a cut is classified, how deviation is measured and over what length, and it sets quality grades tied to those ranges, so a stated deviation always carries the length it applies to.

Metric	Imperial	Where you see it
0.25 mm	0.010 in	High-definition CNC repeatability
0.51 mm	0.020 in	Heavy industrial table tolerance
0.64 mm	0.025 in	Light-industrial CNC with height control
0.77 mm	0.030 in	Entry CNC table without height control
1.6 mm	1/16 in	Hand-held plasma, cut to a guide line

Where AWS D1.1 and ISO 9013 fit together

The two standards do different jobs. AWS D1.1 governs what a fabricator must deliver for structural welding: Table 5.4 limits laminar discontinuities in cut surfaces, while Clauses 2.2 and 2.3 address lamellar tearing and requirements for welds, which is where cut edge condition turns into weldability. The CWB comparative review of AWS D1.1 against CSA W47.1 and W59 works through those differences clause by clause, including the heat treatment temperatures called out in Clause 5.8. ISO 9013 gives you the measurement language for the cut itself. Use D1.1 for the joint, ISO 9013 for the edge.

Two more words belong on the drawing. Flatness of a cut plate governs how a gusset or stiffener seats against it; straightness of the cut line governs beam camber, bolt-hole alignment and the fit of a flange connection. Fabrication tolerances for steel buildings and bridges are set out in the AISC Steel Construction Manual resources. When an edge is out of tolerance the repair route is usually mechanical, and air arc welding and carbon arc gouging takes it back to sound metal before re-welding.

Real Accuracy Numbers: Hand Plasma vs CNC Tables vs High-Definition Plasma

Shop arguments about plasma cutting rarely end with “which machine is best.” They end with a number on a quotation, and everyone reads that number differently. A tolerance of 0.030 in sounds tight until you learn it was measured on 6 mm plate with fresh consumables on a machine levelled that morning. The figures below come from Hypertherm’s published guidance on plasma accuracy and cut quality, one of the few manufacturers that publishes straight numbers.

Hand-held plasma: accurate enough to be deceptive

An experienced operator with a hand-held air plasma torch, working to a straight guide or a soapstone line, can hold the edge to about 1/16 in (1.6 mm). That is useful on 12 mm plate, and it is why hand plasma stays alive in structural shops. But the number assumes a good operator, good consumables, a clean surface and a flat, well supported plate. The same torch on a vertical bracket or a rust-covered 25 mm flange will not repeat it. Its undeniable advantage is reach: it will follow a curve or climb a tower a small table cannot touch.

Entry-level CNC tables: repeatable, and honest about it

Entry-level CNC plasma systems can be expected to hold plus or minus 0.030 in (0.77 mm), and most ship without automatic torch height control. Without height control, the arc sets the kerf, and the kerf wanders as the plate heats and as mill scale, rust and spatter change the surface under the jet. Expect that figure on clean, flat mild steel and expect it to stretch on everything else. It is a system average, not a per-part promise.

Light-industrial CNC: where height control starts paying

Light-industrial CNC systems with full-feature torch height control reach 0.015 to 0.025 in (0.38 to 0.64 mm). Holding the torch at a set distance from the plate is what makes this possible, and it is the upgrade that most improves consistency on real shop work, where plates are never perfectly flat or clean.

Heavy industrial and high-definition plasma

Heavy industrial plasma systems cut close to square edges on plate from gauge to about 2 in (50 mm) thick, produce bolt-ready holes without secondary machining, and hold 0.010 to 0.020 in (0.25 to 0.51 mm) depending on material and thickness. High-definition plasma increases the density of the plasma gas and reshapes the nozzle so the arc leaves the torch narrower, hotter and more focused. The concentrated jet punches the kerf cleanly, so the machine can often travel faster for the same penetration and the edge arrives with less dross, less top-edge rounding and fewer slag flakes. The trade-off is consumables: high-definition electrodes and nozzles cost more and wear faster, moving real operating cost from the machine price into the consumable line.

System class	Typical published tolerance	Torch height control	Best use for structural work
Hand-held air plasma (experienced operator)	About 1/16 in (1.6 mm)	None	Field repair, curved profiles, one-off and access-limited work
Entry-level CNC plasma table	±0.030 in (±0.77 mm)	Usually none	Brackets, gussets, plate profiles, holes in clean mild steel
Light-industrial CNC with height control	0.015–0.025 in (0.38–0.64 mm)	Yes	Production profiles, bevels, moderate tolerance structural parts
Heavy industrial high-amperage CNC	0.010–0.020 in (0.25–0.51 mm)	Yes	Heavy plate to 50 mm, bolt-ready holes, contract work with tight tolerances

Read those figures as conditions, not guarantees. Exact tolerances depend on the system, the process and its setup, and gas pressure, flow and plasma gas type all change the result. A machine that flatters itself on 10 mm clean plate can disappoint on 30 mm mill-scale plate. Within CNC plasma categories, high-definition plasma is always more accurate than light-industrial and entry-level systems, but no category survives bad consumables, low air pressure or a dirty bed.

The headline tolerance is also the least interesting number on the invoice. Consumable life, machine reliability, operating cost per hour and after-sales support decide your real cost per cut far more than a few thousandths of an inch. Before buying, compare what each supplier will stand behind on parts, gases and consumables, using a practical welding equipment buyer’s guide for the questions to ask, and check who will supply gases and consumables locally through a welding supply and gas services guide. A machine with a 0.020 in spec and a two-day wait for a torch lead will cost more per job than a 0.030 in machine that never sits idle.

I’ll rewrite this section to hit the 544–816 word target while keeping every fact and link intact.

Dross, Slag, Oxide and Spatter: The Defects That Cost You Twice

Ask any fabricator what the worst part of a plasma cut is, and most will not say tolerance. They will say dross. That hardened, brittle layer of re-solidified, oxidized metal that clings to the edge of the kerf and refuses to come off. On a good cut it is a nuisance. On a bad cut it controls your schedule, because a fitter with a grinder in hand before the welder ever strikes an arc is a schedule you have already lost.

Here is the honest framing: a plasma cut is a weld in disguise. Everything the cut edge carries into the joint, the weld inherits. Dross, oxide, spatter, drag, taper. None of it disappears downstream. It just changes who pays for it and how much it costs.

What Dross Actually Is

Dross is the re-solidified, oxidized metal left behind in the kerf when the plasma jet does not fully eject the molten material. Instead of being blown clear, it oxidizes in the airstream, cools, and welds itself to the cut edge as a dark, brittle crust.

It comes in two flavours, and telling them apart saves a lot of guessing.

• High-speed dross sits on the bottom edge of the cut. It forms when the jet is angled upward or the travel speed outruns the jet’s ability to clear the kerf. It is heavy, dark, vitrified, and genuinely hard to remove. Sometimes it has to be ground off or chipped with a pneumatic chipping hammer before the edge is fit-up ready.
• Low-speed dross and spatter collect on the top edge as loose, flaky adhered droplets. This is the more forgiving form. A quick rasp or a light grinder pass usually clears it, and many shops simply dress it off as part of normal prep.

Position matters because it tells you the cause. Bottom-edge dross points at pierce quality, travel speed and standoff. Top-edge spatter points at arc energy, consumable condition and gas mix. Once you learn to read the edge, diagnosis gets much faster.

Why Dross Is a Welding Problem First

Dross is not a cosmetic issue, and this is the point most missed in the shop. On a joint face, dross becomes a lack-of-fusion and porosity risk. It is oxide. It does not melt. It floats on the puddle surface, and a torch angle that would normally wet the joint face now sits on top of a crust that repels weld metal. The result is a bead that looks fine and is not, with the defect buried under the cap. That is a reject waiting for an inspector with a UT set.

It gets worse in practice because dross is porous. It traps moisture, mill scale and dirt. Even on a dry day, that moisture finds its way into the weld and contributes to hydrogen cracking risk on restrained or thick joints. And a dross edge is uneven, so the fitter either grinds it back and loses the root face, or leaves a variable gap that changes heat input.

Grinding the dross off creates a second problem. On thin material, a grinder will burn straight through the edge, and on thicker sections the operator will work-harden the surface. Work-hardened steel is harder to grind, and it resists fusion just as badly as the dross did. You end up with a worse joint than the one you started with.

Oxide: The Layer You Cannot See at Arm’s Length

Dross is visible. Oxide is not always. Plasma-cut carbon steel comes away with a blue-black magnetite scale over the surface, thin but continuous. It has to be removed at the joint face before welding, and AWS D1.1 expects clean, sound metal at the joint. The CWB comparative review of AWS D1.1 puts it plainly: lamellar discontinuities in cut surfaces are a fabrication defect to be assessed, not accepted by default.

Stainless steel is a different story. Plasma-cut stainless picks up a chromium-rich oxide tint that is chemically tenacious and carries a higher risk of corrosion and weld contamination if it is not cleaned. On stainless, cleaning means dedicated stainless-grade abrasive and a dedicated wire brush. Reusing a carbon steel brush contaminates the joint and creates exactly the problem you were trying to avoid. If your shop handles stainless and aluminium, prepping aluminium and stainless for welding properly is worth a read before the next job.

What Dross Actually Costs

Add it up honestly. Dross means extra grinder time and consumable discs. It means a second fit-up because the edge moved when it was dressed. It means a welder working a tired, sore body on a joint that is now out of position. It means a possible rejection at inspection, which means the whole plate comes off the fit-up bench. It means the schedule slips by hours, and nobody on the schedule chart has ever written “dross” in the reason column.

That is why dross gets paid twice. Once on the cutting table, and once at the fit-up bench.

Causes to Symptoms: The Diagnosis Table

Symptom	Likely cause	Fix
Sudden dross on a machine that cut cleanly yesterday	Gas type, pressure or flow changed	Check gas type, regulator delivery pressure and filter/dryer first
Heavy bottom-edge dross	Travel speed too fast for the amperage, or torch standoff too high	Slow the travel, raise amperage, drop the standoff
Top-edge spatter and flakes	Worn electrode or nozzle, or wrong consumable set for the material	Change consumables, verify the correct set for the alloy
Dross only on the first 50 mm of a cut	Poor pierce, no pierce delay	Add pierce delay and set the pierce height
Round top edge, drag on one side	Excessive standoff, worn height control	Re-zero height control, verify torch height sensor
Dross and poor edge on rusty or painted plate	Contaminated surface, low gas quality	Clean the surface or raise air quality; check for moisture in the line
Intermittent arc, cut stops partway	Moisture or oil in the gas line, loose torch lead	Purge and dry the line, inspect leads and connections

Hypertherm’s guidance is worth repeating here because it saves hours: gas type, pressure and flow are the first variables to check when dross appears suddenly on a machine that was cutting cleanly the day before. Settings rarely change on their own, but gas delivery does.

One more point that eases a lot of frustration. Plasma is far more forgiving than oxy-fuel on dirty, rusty or painted surfaces. As the AWS Welding Digest notes in its plasma and fiber laser comparison, the same rust, mill scale and contaminants that demand bead blasting or chemical washes before a fiber laser will cut are largely no obstacle to plasma. So when dross shows up on dirty stock, the answer is usually a settings problem, not a material problem. That is good news, because settings you can fix in minutes.

Next time a plate arrives at the fit-up bench with a crusty edge, do not just reach for the grinder. Check the gas first, then the consumables, then the travel and amperage settings. Fix the machine, and the dross stops arriving in the first place. To understand exactly how those settings interact, it is worth knowing what the plasma jet is doing in the kerf.

Beveling and Groove Prep to AWS D1.1 for Structural Steel

A plasma cut edge is not finished until it is prepared for welding. The bevel is the contract between the cutting bay and the welding bay, and it is where cut tolerance turns into weld volume, preheat demand and distortion. Get the angle, root condition and edge cleanliness right and the fitter can weld straight from the plate. Get them wrong and the job pays for grinding, extra filler metal and rework.

What a Fabricator Needs From a Bevel

Five things should be true before a prepared edge leaves the shop.

• Correct bevel angle. Usually 30, 35 or 45 degrees, as the detail calls for, held within a degree or so along the joint.
• A uniform root face. The root face is the land of uncut steel at the bottom of the groove. The root opening is the gap left when two members are drawn together. A constant root face gives a constant root opening; a random one gives one that varies metre by metre.
• No sharp notches. A nick in the edge becomes a crack starter and a filler metal trap.
• No dross, slag or heavy oxide on the joint face. Anything left on the fusion face has to come off, or it goes into the weld.
• Edges that meet within groove-weld tolerance. Two plates that meet cleanly are a joint; two plates that meet 3 mm off square are a fitting problem.

The Code Framework: AWS D1.1 Clause 5 and Its Tables

Most shops never read D1.1 end to end. They read the joint detail, the procedure and the inspection results. But AWS D1.1 lays out the geometry rules that decide whether an edge is acceptable, and the fabrication clause is where they live.

CWB’s comparative review of AWS D1.1:2015 against CSA W59 is a useful public map of how the two codes line up. It highlights several references worth keeping at the fit-up bench:

Reference	What it governs for the fitter
Clause 5 (Fabrication)	General fabrication requirements: cutting, welding and heat treatment; Clause 5.8 sets heat treatment temperatures
Prequalified joint details	Bevel angle, root face, root opening and groove dimensions for common joint types
Table 5.1	Allowable atmospheric exposure of low-hydrogen electrodes
Table 5.3	Prequalified minimum preheat and interpass temperature
Table 5.4	Laminar discontinuities in cut surfaces
Table 5.8	Acceptable and unacceptable weld profiles
Table 5.9	Weld profile schedule
Clauses 2.2 and 2.3	Lamellar tearing and requirements for welds

One difference matters: D1.1 specifies a minimum weld size for bevel, V, U and J prequalified joints, while CSA W59 specifies a minimum depth of size for the same joints. Different wording, same practical consequence. For plain-language background, this Fabricator explainer is a solid read.

Why Two Degrees Off Costs Real Money

Push a bevel angle off the drawing value and everything downstream moves with it. On a 25 mm plate with a 20 mm root face, losing three degrees of included angle changes required weld metal volume, adds a fillet pass, raises deposited heat and therefore preheat and interpass demands, and shifts the shrinkage balance that causes distortion. Small angle error, large downstream consequence. AISC publishes the Manual of Steel Construction alongside standards such as AISC 303, which covers fabrication tolerances for steel buildings and bridges.

Multi-Pass, U-Groove and Back-Gouging

Thick joints get multi-pass grooves, and fillet passes then have to blend into the root, which is why an inconsistent root face shows up as a lumpy, hard-to-weld first pass. U-groove preparations reduce deposited metal for the same joint strength, but need a machine that can hold a curved profile consistently across the plate.

Back-gouging to sound metal is where plasma competes well: air plasma gouging and carbon arc gouging strip out slag, undercut or a cracked pass quickly on heavy sections. Air arc and carbon arc gouging stays the standard answer for clean-up where plasma would be too aggressive. One caution: plasma is a poor choice for the root face of a very deep groove. The jet angle wanders, the root land ends up scalloped, and grinding or oxygen gouging gives far better control on that last millimetre.

Laminated Plate: The Layer Nobody Checks

Plasma cuts through the layers of a laminated plate cleanly. The cut severs the metallurgical bond at the edge, so a plate that looks perfect on the machine can open up later. Every laminated member needs its cut edge checked for delamination and any lifting repaired before welding. Table 5.4 in D1.1 addresses laminar discontinuities in cut surfaces, and Clauses 2.2 and 2.3 connect edge condition directly to weld soundness.

A 60-Second Bevel Acceptance Check

1. Bevel angle within tolerance using a bevel gauge or machinist square.
2. Root face width consistent along the groove.
3. Root opening uniform when the joint is pulled up.
4. Joint face free of dross, slag and heavy scale after grinding.
5. No sharp notches or undercut at the edge.
6. No visible lamellar tearing or edge separation on laminated members.
7. Match marks match the drawing and bolt holes line up.

From Cut Edge to Weld: Fit-Up, Preheat, HAZ and Distortion

So far the job was geometry: angles, holes, lengths. Now it is metallurgy. The edge you hand the fitter sets the grinding, the heat and the restraint the member carries. Cut precision pays for itself here, or quietly costs you money.

Fit-Up: Where Cut Precision Pays Off First

Uniform root opening is the clearest payoff from a straight, square cut. A consistent bevel means the gap measures the same at top, bottom and both ends, so the fitter bolts up, gauges once and welds.

Taper removes that certainty. The fitter can accept a gap that changes along the joint, pack it, or grind back to sound metal with a wheel or an air arc gouging process. On thick plate a backing bar is often the practical answer, adding a welding operation, more heat and another surface to clean. Taper also pushes the joint toward end-loaded behaviour, and AWS D1.1 cuts the allowable length of an end-loaded fillet weld by formula once the weld runs 100 to 300 times longer than the weld size. Fit-up is not neatness. It is code.

Strongbacks, clamps and tack order matter just as much. Tack welds set off the joint line, or tacked in the wrong sequence, lock asymmetry into the member before the real weld starts.

The Thermally Altered Edge

Every plasma-cut edge carries a heat-affected layer: temper colours, a decarburised skin on carbon steel, and magnetite scale left as the jet oxidises the surface. It is narrow, but not neutral. For ordinary structural work D1.1 does not generally require removal of the heat-affected zone before welding, but it does require a joint fit and welded without defects: an edge free of scale, dross and oil.

Edge condition matters most where hydrogen cracking risk climbs: thick sections, higher-carbon steels, low-temperature service, heavily restrained joints. Hydrogen from electrode moisture and plate surface condition, with tensile stress in the weld and HAZ, can produce a delayed crack. The three levers are edge cleanliness, preheat and electrode storage. D1.1 Table 5.1 sets allowable atmospheric exposure of low-hydrogen electrodes per the manufacturer’s storage and re-exposure instructions, so an opened container has a clock running on it. For shop-level consumable handling, the case for SMAW still holds, because low-hydrogen rods simplify storage and preheat control.

Fume belongs here too. Cutting and welding stainless or other chrome-bearing alloys releases hexavalent chromium, a documented carcinogen, and OSHA’s limit for Cr(VI) is 5 micrograms per cubic metre as an 8-hour time-weighted average, per NIOSH’s documentation of the hazard. That is why source extraction at cut and weld matters, and why fume control basics in OSHA’s welding fact sheet belong in the quality conversation, not only the safety one.

Preheat, Interpass and the Heat You Put In

Preheat slows cooling through the critical range so hydrogen can escape and the microstructure can transform without hardening and cracking. AWS D1.1 Table 5.3 sets minimum preheat and interpass temperatures for prequalified joints, driven by thickness and carbon content. Outside that envelope, preheat follows your qualified procedure. Interpass temperature is the half nobody measures: a second pass onto a cold first pass adds to the quench. Post-weld heat treatment is a separate decision, and D1.1 Clause 5.8 sets its requirements and temperatures.

Heat input is the fabricator’s lever on distortion, and D1.1 caps it for prequalified joints. More heat input means faster travel and a larger heat sink to distort around; less means more passes, more arc time and a longer sequence.

Distortion Starts at the Groove, Not the Arc

Distortion gets blamed on heat input and sequence, and both matter. The root cause is often upstream: an asymmetric bevel angle and an uneven root face put more metal on one side of the joint than the other, so shrinkage pulls unevenly and the member bows. A groove two or three degrees off the drawing angle changes not just weld volume but the direction the member moves as it cools.

Once the joint is symmetrical, sequence does the rest. Long continuous welds distort more than broken ones, so back-step, balanced and double-up techniques matter more than most operators expect: welding from the ends toward the centre, alternating sides and sharing heat where the procedure allows. The vertical and out-of-position options in the all-position rod guide are worth reviewing before a large assembly goes on the table.

What Inspection Actually Sees

Inspection runs visual first, then the nondestructive methods the contract calls for: MT or PT for surface-breaking defects, UT for internal ones. The cut edge changes what the inspector finds: a rough, drossy, oxidised edge masks surface indications, while a ground or gouged edge can expose a subsurface flaw hidden as-cut. It also makes UT near the edge harder to read, because residual stress and local geometry both play a part.

D1.1 requires filler metal selection to match base metal strength, and material identification and traceability are covered in Clause 3.3 with Tables 3.1 and 3.2. A plasma-cut plate carries a heat number or a mark-up, so keep it matched to the member through fit-up and welding. The Fabricator’s guide to D1.1 is a solid refresher on how the clauses fit together.

None of this is separate from welding. The cut edge is the first weld: it sets the gap, the heat, the restraint and what the inspector sees. To prove the value of precision on your own floor, measure one member end to end, from cut tolerance to arc time to final straightness, and compare it with the same member cut sloppily.

Common Plasma Cutting Mistakes and a Field Troubleshooting Playbook

Every plasma problem surfaces later as a symptom the fitter finds: a bottom edge full of dross, a hole that will not take a bolt, a bevel leaning off square. Work backwards from the symptom instead of guessing at settings. These eight faults come up most often.

Symptom	Likely Causes	Corrective Action
Dross on one side only	Travel speed mismatched to amperage; worn electrode or nozzle; height control drift; dirty plate	Rebalance amperage and travel; fit fresh consumables; recalibrate height control; clean the plate
Cut angle leaning, tapering along the plate	No torch height control; dragging torch lead; worn drive rack or belt; backlash; gantry out of square	Fit or repair height control; add a lead guide; inspect rack, belt, backlash and squareness
Drag or rounded top edge	Excessive standoff; torch angle not perpendicular; rising arc voltage	Set standoff to the manual value, re-zero height control, square the torch
Pierce failure or oversized hole at pierce	Low gas pressure; moisture in the line; worn shield cup; dirty plate; wrong gas	Check pressure and flow; service filter and dryer; replace the shield cup; clean the pierce area
Hole undersize, oversize or out of position	Kerf compensation off or wrong offset; consumables drifting standoff; backlash or squareness error	Enable kerf compensation and verify the offset; fit fresh consumables; recalibrate height control
Cut will not start, arc extinguishes, or arcs over	Low gas pressure; moisture; loose torch leads; failing torch or power supply; wrong pierce settings	Restore pressure and flow; dry the line; reseat leads; test the power source; correct pierce settings
Poor penetration or cut stops partway	Amperage too low for thickness; travel too fast; torch lead grounding the arc; worn electrode	Raise amperage or slow the feed; free the lead and check grounds; fit a new electrode; verify the program
Rough, heavily oxidized cut surface	Wrong gas mix; air cutting where oxygen-assisted gas is specified; dirty plate; low arc energy	Switch to the specified gas; clean the surface; raise amperage or slow travel

Two patterns explain most of the list. Without automatic torch height control, the odds of negative or positive cut angles and poor cut quality rise, and the shop pays for it in grinding away excess molten metal and surface scratch, according to Hypertherm. Worn consumables also drift the torch-to-work distance, so height control must be recalibrated whenever the electrode and nozzle set changes. Most shops skip that recalibration, which is why a machine that cut well at 8 a.m. starts tapering edges by lunch.

Preventive maintenance that actually prevents

• Daily: Inspect the electrode and nozzle for wear, and change them on a fixed interval rather than when the cut fails. Check gas inlet pressure and flow, the filter and dryer for moisture, and confirm water-cooling flow.
• Weekly: Inspect the table bed and slats for spatter buildup. Check gantry squareness against the manual’s tolerance. Lubricate the drive rack, rails and bearings. Confirm height control holds set height after a consumable change.
• Monthly or program change: Cut a test coupon in the same material and thickness as the job, measure squareness and root face, record consumable hours, and keep both as your reference.

That baseline turns arguments into data. When a new operator claims the machine is cutting badly, pull the coupon from the same material and thickness and compare. Match the reference and the machine is fine, so the setup is at fault. Drift and you have a machine problem, and you know it early. Storing consumables dry and dated, the way you manage stick electrodes, matters too: damp gas lines cause the same bad-cut symptom.

New operator or new machine

These are different problems and shops confuse them regularly. A new operator needs a reference, not freedom: the baseline coupon, the manual settings, and a short list of the parameters they may touch. They should not adjust amperage, travel speed, gas pressure and height control in one hour. A new machine needs commissioning instead: verify gas flow and pressure, confirm height control calibration against the manual, check gantry squareness and backlash, then run the baseline coupon before it touches production work.

One more habit pays off fast: log every cut defect by machine, shift, material and consumable hours. Within a month the log shows which variable is drifting, and guesswork becomes a scheduled change. Treat plasma cutting as a measured process and the fitter’s grinding time drops before anyone notices.

Plasma vs Oxy-Fuel vs Fiber Laser vs Oxy-Acetylene Gouging

Choosing a cutting process is not about the prettiest edge in a brochure photo. It is about matching the process to the material, the tolerance the drawing demands, and what your shop can afford to run eight hours a day. Four processes cover most structural fabrication work.

Oxy-Fuel: Still the Standard for Long, Straight, Heavy-Plate Cuts

Oxy-fuel has been the backbone of structural steel fabrication for over a century. For a long straight cut through heavy plate it is still the cheapest route to a clean edge: one large oxygen cylinder and a smaller fuel gas bottle serve several torches, there is no capital equipment to buy, and consumables are a tip and a length of hose.

Oxy-fuel is slow and needs a clean preheat before the cut starts, so the operator heats the plate before any cutting happens. The edge comes out heavily oxidized and scaled, meaning more grinding before the joint is weld-ready, a cost that never shows up in a per-inch quote. Hoses and tips wear fast, so maintenance and downtime climb.

Plasma: The Generalist That Earned the Table

Plasma is faster, cuts a narrower kerf than oxy-fuel, and leaves a cleaner, less oxidized edge that needs far less dressing before fit-up. It also runs on dirty, coated or rusty conductive material, which matters when plate arrives with mill scale, primer and grime on it. Plasma handles bevels, tapers and CNC table work that would be tedious with a torch, and it is the process behind most modern laser-era cutting cell layouts.

On cost, plasma pulls decisively ahead in the 12 to 25 mm (0.5 to 1 in) range and thicker, holding the advantage in productivity and operating cost. Even at the tight end of tolerance, Hypertherm notes that modern CNC plasma systems with torch height control hold accuracy of 0.38 to 0.64 mm (0.015 to 0.025 in), more than enough for structural fit-up.

Fiber Laser: Precision You Pay For

Fiber laser earns its place on thin sheet and fine features. The kerf is narrower, cutting can be faster on light material, and small holes and intricate profiles come out clean with little secondary work. That matters for stainless trim, aluminum panels and high-volume parts work.

Structural plate is a different question. The AWS Welding Digest reported in May 2024 that fiber laser cutting systems cost two to five times more to purchase than comparable plasma machines, and that plasma systems can often recover their capital expense within two years or less. A fiber laser also needs enclosures and dedicated material handling for safety, while a plasma table can hit similar throughput by lengthening the cutting bed.

Environment matters. Fiber laser is more sensitive to a dirty shop: contaminants affect optical components and precision drives, accuracy and uptime. Fiber lasers have fewer consumable parts but higher maintenance costs from complex cutting heads and motion control, while plasma systems are less complex, more robust and better suited to harsh industrial floors. And rust, mill scale and coatings still have to come off before the beam works. Past 12 to 25 mm (0.5 to 1 in), plasma is the cheaper machine to own and run.

Gouging and Clean-Up: Oxy-Acetylene vs Plasma

When a weld has to be removed, back-gouged to sound metal, or a defect dug out, the classic answer is still oxy-acetylene. It is controllable, works at any thickness, and the flame chemistry is well understood. Shops needing a torch primer should read our guide to air acetylene welding torch basics, brazing and safety, plus the process notes on air arc welding and carbon arc gouging for heavy carbon removal.

Plasma systems can also gouge and bevel, which keeps everything on one machine and one power supply. The caveat: a bevel increases the effective thickness the arc must push through. A 25 mm plate cut with a 30 degree bevel behaves like heavier plate at the root, so plasma beveling demands more amperage, more gas and a slower travel speed than a square cut on the same material.

A Practical Decision Guide

Situation	Best choice	Why
Long straight cuts, heavy plate	Oxy-fuel	Lowest cost per cut, no capital gear
Rusty, primed or dirty plate	Plasma	Ignites through contamination, cleaner edge
Precision geometry, bevels, CNC tables	Plasma	Tight tolerances, low running cost
Thin sheet, fine features, high counts	Fiber laser	Narrow kerf, speed, no consumables
Back-gouging, defect removal	Oxy-acetylene	Controllable, works at any thickness

The decision comes down to six variables: material condition, thickness, the edge tolerance the drawing demands, the condition of the shop floor, the throughput needed, and true total cost including consumables, power, gas, maintenance and the grinding you avoid. Tight tolerance on dirty plate means plasma; clean thin material and enough volume means a laser earns its place. Whatever you buy, hot work safety comes from the code, with precautions in OSHA 1910.252.

A Practical Quality-Control Workflow and What to Check Before You Weld

Everything in the previous nine sections collapses into one habit: catch the problem while the steel is still on the table. A layered QC system does exactly that. Four layers, each cheap, each catching a different class of error, turning cut precision from something you argue about into something you measure. This version needs only a clipboard, a machinist square, pin gauges and a log sheet.

Layer 1: At the Machine

The operator owns this layer. The check takes three minutes, before the part leaves the table, when the cause is still obvious and the fix is a consumable change, not a scrapped plate.

Check	Tool	What Acceptable Looks Like
Edge squareness	Square or angle gauge	Even contact along the full edge; taper visible immediately
Root face width	Calipers or bevel gauge	Constant from end to end
Dross	Visual, both sides	None on the joint face; light scale elsewhere
Top edge condition	Visual	No rounded or melted lip
Hole size	Pin gauge or plug gauge	Bolt slides in without forcing
Edge straightness	Straightedge and chalk line	No bow across the plate
Consumables and gas	Log card	Electrode and nozzle hours, gas pressure, filter and dryer status

Two habits matter most. Record consumable hours and gas pressure on the same sheet, because most dross traces back to a worn electrode or a wet air line. Sample squareness randomly through a nest, not only the first part, since a torch that drifts as consumables wear looks fine on piece one and out on piece twenty.

Layer 2: At the Fit-Up Bench, 60 Seconds

The fit-up bench is where cut quality becomes visible. Six quick checks let the fitter flag a bad edge before it becomes grinding, preheat and rework.

• Root opening uniform along the joint, or any taper packed out and noted.
• Joint face free of dross and scale after grinding, down to bright metal.
• No sharp notches or undercut-looking steps along the prepared edge.
• No delamination or lamellar tearing on the visible lamination lines.
• Bolt holes align with the drawing and the mating part.
• Match marks match, and member identification is legible against the drawing.

Layer 3: At the Weld

Confirm the AWS D1.1 qualified procedure is on the wall and matches the joint. Verify preheat and interpass temperature with a calibrated instrument, not by feel, and record it. Check electrode storage: low-hydrogen consumables have limited allowable exposure to the shop atmosphere, so time out of the holding oven is a controlled parameter. This welding rod guide is worth posting at the station. Inspection scope comes from the contract, not from habit.

Layer 4: The Feedback Loop

Log every cut defect by machine, shift, consumable hours, grade and thickness. Trends are obvious once written down: dross clustered on the night shift at a given consumable life points to something other than material quality. Adjust settings, the change interval or the training, then watch the column change.

Safety Is a Quality Issue

Plasma and oxy-fuel cutting are hot work, and OSHA 1910.252 sets general requirements for welding, cutting and brazing. Where the object to be cut cannot be moved, movable fire hazards must be moved to a safe place; where they cannot be removed, guards must confine heat, sparks and slag.

OSHA’s welding fact sheet identifies what drives worker exposure to welding fume: the type of process, base and filler metals, location (outside or in an enclosed space), welder work practices, air movement, and ventilation controls. That same fume contains metals including aluminum, antimony, arsenic, beryllium, cadmium, chromium, cobalt, copper, iron, lead, manganese, molybdenum, nickel, silver, tin, titanium, vanadium and zinc, plus gases such as argon, helium, nitrogen, carbon dioxide, nitric oxide, nitrogen dioxide, carbon monoxide and ozone. The composition details are laid out in this overview of welding fume compounds.

Safety belongs in a quality workflow because a poorly controlled hot-work area produces poor work: poor visibility, distracted operators, hot work cooling in the wrong sequence and rework. Carbon arc gouging shows how fume control and cut quality are the same conversation.

The Whole System, One Cut at a Time

Plasma cutting precision is not a number on a spec sheet. It is a system property: machine, torch, consumables, gas, settings and operator working together. Tolerance is the shared language that lets a shop and a client talk honestly about it, and the cut edge is the first weld operation whether or not anyone lit an arc.

One last exercise. Pick a real job and measure it end to end: grinding minutes removing dross, extra weld metal deposited, preheat minutes added, hours of re-fit and rework. That figure is usually bigger than the argument you have been having, and the easiest justification your next cut-quality programme will ever need.

Frequently Asked Questions

How accurate is plasma cutting compared with fiber laser cutting?

For dimensional accuracy on a CNC table, plasma holds roughly 0.010–0.030 in (0.25–0.77 mm) depending on the machine class and setup, while fiber laser’s narrower kerf gives finer features on thin material. The bigger difference is material condition and shop environment: plasma reliably cuts rusty, scaled or coated conductive steel, whereas fiber laser needs cleaner stock and is more sensitive to shop dust and optical contamination (AWS Welding Digest, May 2024).

Is dross on a plasma-cut edge acceptable for structural steel welding?

Not on the joint face. Dross is re-solidified oxidized metal that must be removed before welding because it traps moisture, promotes porosity and can cause lack of fusion; loose dross near a joint is also a burn and injury hazard. Off the joint face, light dross is normally tolerated as long as it is not loose and the surface is free of scale and contamination.

Does AWS D1.1 require the heat-affected zone to be removed from plasma-cut edges?

For ordinary structural work, AWS D1.1 does not generally require removal of the heat-affected zone from thermally cut edges; it addresses cut surfaces through requirements such as Table 5.4 (laminar discontinuities in cut surfaces) and the Clause 5 fabrication provisions. What the code does require is that the joint be prepared and welded to the specified groove details, with preheat and interpass temperatures meeting Table 5.3.

What tolerance should I write into a purchase order for plasma-cut steel?

State the tolerance class, the material and thickness range, the edge quality expectation and the measurement method — for example ‘edges within 0.030 in of square on material up to 1/2 in thick, measured per ISO 9013’. Vague phrases like ‘machine cut’ invite disputes, because entry-level CNC tables typically hold about ±0.030 in (±0.77 mm) while heavy industrial systems with torch height control can hold 0.010–0.020 in (0.25–0.51 mm).

How do I prepare a plasma-cut bevel for a groove weld?

Grind or dress the joint face back to bright, sound metal, remove all dross and mill scale, verify a uniform root opening along the joint, and confirm the bevel angle against the drawing or the prequalified joint detail in AWS D1.1. Check for lamellar tearing or lamination on the plate edge and repair it before welding; a groove that is 2–3 degrees off drawing changes weld volume, pass count and distortion.

What causes excessive dross and a leaning cut angle on a plasma cutter?

Dross usually means amperage is too low for the travel speed, the consumables are worn, the standoff is too high, or gas pressure and flow are off; a leaning angle usually points to a torch height control problem, a worn or wrong consumable set, or a gantry squareness issue. Work through them in that order, and check the machine’s gas filter and dryer for moisture before changing settings.

When is fiber laser the better choice than plasma for structural fabrication?

Fiber laser wins on thin material where cycle time and a narrow kerf matter, on fine features and small hole work, and in shops with clean air, dust control and enclosed cutting. Plasma wins on thick plate (roughly 12–25 mm and up), on rusty or coated material, on dirty shop floors, and on cost: fiber laser systems cost two to five times more than comparable plasma machines.

Is plasma cutting safer than oxy-fuel cutting on a structural steel job?

For fire risk it is generally better, because plasma cuts far faster with less heat and produces far less slag and spatter, though both are hot work under OSHA 1910.252 and need the same fire precautions. For fume it is not automatically safer: OSHA notes that oxy-fuel and plasma cutting both generate airborne metal fume, so local exhaust ventilation and, where controls are insufficient, respiratory protection are still required.