Troubleshooting guide › Cutting problems › CNC plasma and flame cutting problems
A CNC cutting machine adds three things to the plasma or gas torch: motion, height control and a program. Each adds its own faults. Parts out of size are usually kerf or calibration; bevel is usually height; pierce failures are timing; bad holes are speed on small radii. This page is for portable CNC shape cutters and gantry machines with plasma and oxy-fuel torches, and it assumes the torch itself is cutting well (see the plasma and gas cutting pages first).
Reviewed by the Cruxweld technical team, October 2026. Settings are starting points to confirm on a test piece; acceptance is decided by the code or drawing the job is built to. Checks marked “call” are for our technical team.
The controller offsets the torch path by the kerf value. On the Cruxweld DynaCut ProX and ProGantry 2560 controller the value entered is HALF the measured kerf width (as documented by Cruxweld); some CAM programs and other controllers expect the full width, so check which convention your machine uses. If the entered value is wrong, every part is undersize or oversize by the difference, and holes are the opposite way to outside contours.
The program uses G41/G42 (left/right compensation) and the final G40 is missing, or compensation is applied twice (in the CAM and in the controller).
The controller thinks it moved 100 mm and the gantry moved 99.
Parts come out as parallelograms; diagonals of a square differ.
Dimensions vary from part to part, worse at speed.
Speed or acceleration too high, dirty rails, poor earthing letting plasma HF upset the drives.
Thin plate lifted or warped by heat; parts dropping and tipping.
| Symptom | Likely cause | What to do |
|---|---|---|
| Torch dives into the plate during the cut | Arc voltage set-point too low; divider ratio wrong (100:1 vs 50:1); THC not locked in corners so the slow-down raises the voltage and the torch dives | Raise the set-point (typical 100-150 V for air plasma); set the divider ratio to match the plasma source; enable corner lock at about 95 % speed and 5-10 mm from segment ends |
| Torch rises too high, bevel grows | Set-point too high; worn consumables raise the arc voltage so the THC lifts to compensate | Lower the set-point a few volts; replace consumables and reset the voltage |
| THC hunts up and down | Sensitivity too high; noisy arc-voltage cable | Reduce THC sensitivity (typical 30-50 on a 1-100 scale); reroute and shield the voltage cable |
| Torch does not find the plate (IHS fails) | Position-check time shorter than the descent time; proximity or ohmic sensor dirty; plate not conductive at the contact point | Lengthen the position-check time; clean the sensor contacts; check the plate surface |
| Collision alarm with no collision | Breakaway switch wiring or setting; plasma HF noise | Check the switch and its logic (normally open / closed); earth and shield |
| “Broken arc” or loss-of-arc stops | Arc protection window too tight; loss-of-arc delay too short; arc really lost on a worn electrode | Widen the protection value; set the lose-arc delay around 0.5 s; replace the electrode |
| Arc starts but the machine does not move | Arc-transfer signal not reaching the controller; arcing-check time exceeded | Check the arc-transfer wiring; lengthen the arcing check time (30 s typical); as a last resort use a time-based start |
| Wrong pierce height | Oxidised IHS contacts or arc-voltage cable | Clean and tighten |
Arc voltage rises as the torch lifts and as the consumables wear. A set-point found with new consumables will slowly lift the torch as the electrode ages. Re-check the cut height with a feeler every few hours of production, and replace consumables before the bevel appears.
The torch moves off before the arc is through; the cut starts with a gouge or the arc is lost.
Molten metal blows back onto the nozzle: double arcing and a short consumable life.
Above about two-thirds of the maximum cut thickness the pierce is unreliable and wrecks consumables.
The pierce is on the contour.
Cutting oxygen opened fully at once blows the pool out and cools the plate.
Pierce-down or pierce-up time set to zero.
| What you see | Cause | Fix |
|---|---|---|
| Contour not cut, or the machine lifts and re-pierces in the middle of a shape | Open contour: the lines in the DXF do not meet (gap or overlap at a corner) | Join or close the polyline in CAD; most CAM software has a “close gaps” tolerance; preview the toolpath before cutting |
| Machine cuts the same line twice | Duplicate lines on top of each other in the DXF (copied layers, exploded blocks), or text and dimension layers exported | Purge duplicates and delete text, dimension and hatch layers before export; export only the cut layer |
| Shape cut as hundreds of tiny segments, machine shakes | Arcs exported as polylines with many short segments | Export arcs as arcs (R12 or 2000 DXF with arcs), or raise the arc tolerance; enable short-line fitting on the controller |
| Part 25.4 times too big or too small | Inch / mm units mismatch between CAD and CAM | Set the DXF units and the CAM import units the same; check a known dimension on the preview |
| Drawing right, program wrong on the controller | Post-processor for another controller; kerf or lead-in applied twice (CAM and controller); origin or axis direction | Use the post-processor for your controller; compensate kerf in one place only; set the origin and preview |
| Cut order wrong: large part cut before its holes | Nesting sequence not set inside-first | Sequence holes and inside contours first, outside last; cut small parts before large ones |
| Small parts tip up and the torch hits them | Part dropped through the slats before the cut finished | Micro-tabs (0.5-1 mm) on small parts, cut small parts last on a skeleton, or cut them where the slats support them |
| Skeleton warps and the last parts are out | Heat built up in a thin skeleton | Alternate cutting positions across the plate, keep the skeleton web at least 10 mm, cut remnant-plate jobs with the plate clamped |
Cut recovery after a power or arc failure: note the breakpoint on the controller before anything else, do not move the plate, use the breakpoint recovery (or return-to-break) function to resume from the recorded point, and when restarting inside a contour move the start back 5-10 mm so the arc re-fuses the cut. If the plate was moved, re-reference the plate corner before resuming.
| Symptom | Cause | Fix |
|---|---|---|
| Part cut mirrored or rotated | Axis direction or program origin | Check the axis setting and the program start point; use the preview |
| Scaled part has wrong lead-in length | Scaling applies to lead-in and lead-out too | Re-check lead lengths after scaling |
| Nest overlaps or runs off the plate | Plate size or origin wrong; nesting calculation | Set the origin at the plate corner; preview the nest; use the offset function to move the whole nest |
| Cut restarts in the wrong place after a stop | Breakpoint not recorded, or the plate moved | Use the breakpoint recovery function; re-reference the plate corner |
| Controller shows “no file” or cannot read the USB | File not a plain G-code or DXF the controller accepts; USB formatted wrongly | Export as the controller specifies; FAT32 USB |
| Speed changes by itself | Rate override keys pressed; automatic corner and small-arc limits | Reset the rate to 100 %; this is normal behaviour at corners |
| When | Task |
|---|---|
| Every shift | Clean rails, racks and the cross-beam guide of slag and dust; thin film of oil on racks only |
| Every shift | Check plasma consumables and flame nozzle; drain the air water separator; test the emergency stop |
| Every shift | Pre-start checks: end stops, drag chain, hoses, earth and work lead, gas leaks, extinguisher |
| Weekly | Inspect hoses, cables and drag chain for wear and burns; replace damaged hoses at once |
| Weekly | Check gas connections for leaks; check guide wheels for play; check the torch is vertical (and after any collision) |
| Monthly | Measure a test square and a 1 m line; correct kerf and pulse equivalent; check gantry squareness |
Plasma always cuts one side better because of the swirl in the arc. Set the cut direction so the good side is on the part: clockwise around outside contours, anticlockwise inside holes, for most torches. If both sides are bevelled, the torch is too high or the consumables are worn.
Kerf compensation is applied in the wrong direction for inside contours, or the kerf value is set from the outside cut with worn consumables. Check G41/G42 on holes and re-measure the kerf.
Almost always worn consumables changing the kerf, or a collision that knocked the torch out of vertical. Replace, square the torch, and re-cut the test square.
With the machine set up and the kerf entered, about plus or minus 0.5 mm on general fabrication parts from a well-maintained machine; holes and small features are the limit. A sample of your own part on your plate is the honest answer, and a trial confirms it.
Lamps, fault codes and the checks an operator can do outside the cabinet, for the machines this page applies to.
For clean CNC results the plasma source must give a stable arc voltage signal and an arc-transfer signal to the controller, and the controller must have arc-voltage height control with corner lock. On flame cutting, staged pierce oxygen and a preheat timer make thick-plate piercing repeatable. A trial cut on your own drawing and plate, with the measured kerf entered, is the right way to judge a machine.
Send a photo of the weld or cut, your machine model, consumable and plate thickness. Our technical team replies on WhatsApp with the likely causes and the next checks to make.
45+ years of welding expertise · GeM Registered OEM Manufacturer · Supported Worldwide
