When a plasma cut goes wrong, the usual reaction is to fit a new nozzle, then fiddle with the current, and lose most of a day before producing a handful of acceptable parts. In reality there are only four root causes behind every edge defect: speed, gas, standoff and consumable condition. This guide gives you a repeatable diagnostic routine — read the dross shape to tell whether you are too fast or too slow, read the bevel direction to tell whether the arc is deflecting or the torch is too high, then work the seven-step table until the cut is back in spec. Properly applied it turns a day of trial and error into about an hour.
1. The four dimensions of cut edge quality
ISO 9013 is the reference for thermal cut surface quality, and it boils down to four measurable dimensions. Build a set of visual reference coupons for the shop floor:
| Dimension | Definition | Acceptable range | Consequence if out of tolerance |
|---|---|---|---|
| Dross | Re-solidified residue on the lower edge | None, or fine beads that flake off | Extra grinding hours, poor assembly fit-up |
| Squareness / bevel | Angle between cut face and plate normal | ≤ 3° (high-precision reaches 1°–2°) | Bevel parts will not seat; hole position shifts |
| Surface roughness Ra | Microscopic profile of the cut face | Ra 6.3–12.5 μm | Lower fatigue life, poor coating adhesion |
| Heat-affected zone | Narrow band of altered microstructure | 0.3–1.0 mm on carbon steel | Hardening, cracking, accelerated tool wear |
Terminology and grading are defined in ISO 9013; process guidance is also available from the American Welding Society at www.aws.org.
2. Dross: decide first whether it is fast dross or slow dross
This single judgement carries more information than anything else in the routine. The shape of the dross tells you which way to move the speed.
| Sign | High-speed dross (too fast) | Low-speed dross (too slow) |
|---|---|---|
| Location | Continuous fine bead along the lower edge | Clumps at the lower edge climbing upward |
| Colour and hardness | Bright silver, extremely hard, metallurgically bonded | Dull grey/black, porous, knocks off |
| Drag lines | Strongly raked backwards, steep angle | Near vertical or raked forward |
| Top edge | Normal | Roll-over, rounded, widened kerf |
| Action | Reduce speed by 5 %–10 % | Increase speed by 5 %–10 % |
If speed changes do not move the needle, the cause is not speed. Shift to gas and standoff: check dynamic pressure at the torch, confirm the torch is not so low that spatter is bouncing back, and inspect whether the nozzle orifice has gone out of round.
3. Bevel: the direction tells you the cause
The orientation of the bevel is the diagnostic clue, and it is the one most often overlooked:
- Symmetrical waist on both sides — classic over-slow travel or excessive current. Too much heat input over-melts the mid-thickness.
- Constant single-sided lean (one face square, one tapered) — arc deflection, usually from an out-of-round nozzle orifice. Replace nozzle and swirl ring immediately.
- V-shape, wide at the top — torch too high and the arc spreading. Lower the standoff or re-check whether the THC arc-voltage setpoint has drifted.
- Inverted V, wide at the bottom — insufficient gas pressure or flow. Check filter differential pressure and look for line leaks.
On CNC work, bevel is also frequently a kerf compensation problem rather than a machine problem. If the offset appears systematically on clockwise or counter-clockwise contours only, fix it in CAM: StarCAM nesting software applies direction-aware kerf compensation automatically.
4. Double arcing: the number one nozzle killer
Double arcing means the arc is no longer burning only between electrode and workpiece — it is also breaking down between electrode and the inside wall of the nozzle. Signature symptoms:
- Two or more irregular notches at the nozzle orifice, or pitting on the bore wall;
- An audible change in arc sound mid-cut with an abrupt drop in edge quality;
- A nozzle destroyed in ten or fifteen minutes.
| Trigger | Mechanism | Fix |
|---|---|---|
| Pressure or flow too low | Gas film no longer insulates the nozzle bore | Verify dynamic pressure while cutting; clean or replace filter elements |
| Pierce height too low | Blowback bridges nozzle to workpiece | Set pierce height to 1.5–2 × cutting height |
| Nozzle touching the plate | Direct short circuit and puncture | Enable THC; fit a DJC-01 torch guard for collision protection |
| Damaged or blocked swirl ring | Swirl lost, arc runs off-axis | Replace the swirl ring together with the nozzle |
| Inadequate cooling | Nozzle overheats and loses insulation | Check coolant level, temperature and flow rate |
5. Pierce points, lead-ins and hole quality
In production, most scrap is not generated on long straight cuts — it is generated at hole entries. Controls that matter:
Pierce height and dwell
Thick plate requires raising the torch and holding a dwell until the pool breaks through, then dropping to cut height. Most machines expose pierce height and pierce delay; allow 0.3–1.5 s of dwell on heavy plate.
Lead-in and lead-out
Never start on the finished contour. Use an arc or straight lead-in so the start notch lands in the skeleton, and leave 3–5 mm of overlap at the lead-out to avoid an uncut tail.
Small holes
Below roughly 1.5 × thickness in diameter, do not pierce at full current. Use reduced-current piercing or switch to edge start with a slow ramp-in; either approach can more than double nozzle life.
6. The seven-step diagnostic routine (print it and pin it to the machine)
| # | Check | Pass criterion | If it fails |
|---|---|---|---|
| 1 | Consumable condition | Hafnium pit depth < 1.5 mm; nozzle orifice round | Replace electrode + nozzle + swirl ring as a set |
| 2 | Dynamic pressure at torch | Matches the process chart, measured while cutting | Adjust regulator, change filter element, find leaks |
| 3 | Gas quality | Pressure dew point ≤ 2 °C, oil-free and water-free | Service the dryer and three-stage filtration |
| 4 | Standoff and THC | 4–8 mm; arc-voltage setpoint matched to current | Re-set arc voltage, re-calibrate initial height sensing |
| 5 | Single-variable speed tuning | Drag lines near vertical | Adjust in 5 %–10 % steps |
| 6 | Current vs thickness | 70 %–85 % duty loading | Avoid running permanently at maximum output |
| 7 | Edge verification | Dross flakes off; bevel ≤ 3° | Record the parameters and lock them into the process library |
Cut one test coupon after each correction and change only one variable at a time; otherwise you will never identify the true cause. Once dialled in, save the parameters to the machine's process library or the CAM cut chart so quality does not depend on one operator's memory.
7. Frequently asked questions
Why is the edge still bad with a brand-new nozzle?
A fresh nozzle cannot compensate for wrong pressure, standoff or speed, and it cannot repair a worn electrode. Work the seven steps from the top and pay particular attention to hafnium pit depth — an old electrode will destroy a new nozzle quickly.
The dross knocks off easily. Do I still need to remove it?
It depends on what happens next. For any welding or coating operation, remove it even when it flakes easily, otherwise you get slag inclusions in the weld and voids under the coating. For blanking parts that will be fully machined later, leave it.
Why is one side of the cut good and the other bad?
Almost certainly arc deflection: an out-of-round nozzle orifice or a damaged swirl ring has pushed the arc column off the torch axis. Replacing nozzle and swirl ring together usually fixes it. If it recurs, check the gas for water and oil contamination.
Is CNC better than handheld only because the machine costs more?
Not really. The decisive differences are standoff stability — THC holds constant arc voltage for the whole cut — and constant travel speed, neither of which a human hand can reproduce. A handheld operator using a drag shield and accepting minor speed variation can still reach a middling finish.
How do I quantify edge quality instead of eyeballing it?
Make three reference coupons — good, acceptable, reject — and keep them at the machine. Periodically measure bevel with a protractor and roughness with comparison blocks. Well-tuned high-precision plasma holds 1°–2° of bevel; for that class of work consider the MS-4000AYF FineCut plasma torch.
8. Where to start
Nine times out of ten, edge defects are located by three moves: read the shape, judge the direction, verify with a single variable. No trial-and-error marathon required. Reference coupons and a locked-down process library stabilise quality far more effectively than buying a more expensive machine.
If dross, bevel or abnormal nozzle consumption is eating into your shop's output, contact the Icyrobot technical team — we can run a process audit with you. Related reading: plasma consumable life and failure signs and gas selection for stainless steel. Recommended equipment: Max-200HPR high-precision plasma and the LGK-100IGBT inverter air plasma cutter.
