CNC treatment of deformation: causes and mitigation strategies
Deformation is not one problem. It is at least five, and each one needs a different fix. This page maps the symptom you see on the machine to the cause behind it, then to the process change that stops it. Written for engineers and buyers who sign off on a drawing with tight tolerances and need to know where the risk sits before the first chip is cut.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Deformation symptom, likely cause and what to change
Read this as a lookup table. Find the symptom closest to what you measure, then check the cause column before you touch the program.
| Symptom you measure | Most likely cause | What to change |
|---|---|---|
| Part bows after unclamping | Clamping stress released | Softer jaws, lower clamp pressure, rough then finish |
| Thin wall springs back | Cutting force deflects the wall | Lighter radial depth of cut, more axial passes |
| Bore oval after cooling | Heat from the cut, uneven cooling | Flood coolant, lower surface speed, semi-finish pass |
| Long part twists over hours | Residual stress from the mill | Stress relief before machining, rough and rest |
| First article good, part 20 drifts | Tool wear and thermal growth | Tool life limits, in-process probing, warm-up cycle |
| Flat face dishes in the middle | Fixture holds only at the edges | More support points, vacuum or low-melt fixturing |
| Slot closes after anodizing | Surface layer growth and heat | Account for coating, mask tight slots, stress relief |
| Part moves during a second op | Inconsistent datum between setups | Single setup on 5-axis, or a machined soft jaw |
Where the real fix usually sits
If the part moves after unclamping, the problem is stress and clamping, not the finish pass. If it changes during the run, the problem is heat and tool wear. Fix the first one before you spend time chasing the second.
Residual stress is the reason the part moves after the cut
Most deformation blamed on the machine is really stress already inside the material. Rolled plate, extruded bar and castings all carry internal stress from the way they were made. When you remove material from one side, you unbalance that stress and the part bends to find a new balance. The cut is not wrong. The material was already loaded.
You see it as a part that measures in tolerance on the machine and moves after it comes off the vise. Sometimes it moves in minutes. Sometimes it moves overnight. A 6061-T6 plate pocketed on one face will often bow 0.05 to 0.15 mm across a 300 mm length, and no amount of finishing will bring that back if the stress was not dealt with first.
The practical answer is to release stress before the finish cuts. Rough the part, leave 0.5 to 1.0 mm of stock on all faces, then let it rest or stress relieve it. For aluminium, a low-temperature stress relief cycle before machining and a rest period between roughing and finishing both work. For steel, normalizing or annealing before the first op removes most of the problem.
One more point. Symmetry helps. If you remove material from both sides of the part in roughly equal amounts, the stress balance stays closer to where it started. Asymmetric pockets on one face only are the fastest route to a banana-shaped part.
- 1Rough with stockLeave 0.5–1.0 mm on all faces before finishing.
- 2Rest between opsLet the part sit after roughing so it moves before you finish.
- 3Keep it symmetricBalance material removal on both sides where the design allows.
Clamping and fixturing force the part into a shape it does not want
A vise is a press. When you clamp a thin or irregular part, you bend it into contact with the jaws. The cut is then made on a bent part. Release the clamp and the part springs back to its own shape, and the geometry you just cut goes with it. The error is real, but it is not in the toolpath.
The clue is repeatability. Clamp the same part with less pressure and the error changes. Measure the part in the fixture and it looks good. Measure it on the surface plate and it does not. That is clamping stress, not a machine problem.
Fix it by spreading the force. Soft jaws machined to the part profile contact more area, so you need less pressure for the same hold. Use a torque wrench on the vise and record the setting that works. For thin walls, support the wall from behind with a filler, a low-melt alloy, or a vacuum fixture. On a 5-axis machine, machine all accessible faces in one setup so the part never gets re-clamped in a different position.
If a part is delicate enough, do not clamp it at all. Vacuum plates, adhesive fixturing and low-melt alloys hold a part flat without side pressure. They cost more setup time, which is why they are used on the parts where clamping is the dominant error.
- 1Soft jawsMachined to the part profile to spread contact area.
- 2Lower pressureUse a torque wrench and record the number that holds without bending.
- 3Support thin wallsFiller, low-melt alloy or vacuum behind the wall.
Heat from the cut changes the part while you measure it
Cutting generates heat, and heat expands metal. Aluminium grows about 23 μm per metre per degree Celsius. A 300 mm aluminium part that is 10 °C warmer than the inspection room is roughly 0.07 mm longer than it will be when it cools. If you finish the bore hot and measure it hot, you will chase that number all afternoon.
The bigger problem is uneven heat. If one face gets flood coolant and the other sees only air, the part warps as one side expands and the other does not. Deep pockets, long slots and thin ribs all show this. You get an oval bore, a twisted rail, or a face that dishes in the middle.
Control it with coolant, not with speed. Flood coolant on both sides of the cut, directed at the contact zone, keeps the part temperature closer to the room. Take a light semi-finish pass before the final pass so the part has time to settle at the temperature it will be measured at. Let parts cool on the bench before final inspection, and record the room temperature on the inspection report.
On steel, the same logic applies with a different number. Steel expands about 11 to 13 μm per metre per degree Celsius, so the effect is smaller but still larger than a ±0.005 mm tolerance on a long part. Thermal stability is a process decision, not a machine specification.
- 1Flood both sidesCoolant on both faces of the cut keeps the part even.
- 2Semi-finish firstA light pass before the final cut lets the part settle.
- 3Cool before inspectionMeasure at room temperature and record it.
Thin walls, long parts and hard materials need their own approach
Thin-wall parts fail differently. The wall deflects away from the cutter, so the tool takes a lighter chip than programmed and rubs instead of cutting. That rub adds heat and work hardening, and the wall gets thinner and more springy as you go. The usual answer is the opposite of what feels natural: reduce radial depth of cut, increase axial depth, and run a higher feed per tooth so the tool cuts rather than rubs.
Long parts need support along their length, not just at the ends. A 4,000 mm rail held only at two points will sag and chatter in the middle. Support it with adjustable stands or a steady rest, and machine in sections with overlap so the tool pressure stays constant.
Hard materials like 17-4PH, Inconel and Ti-6Al-4V deform less from stress but more from heat, and they work harden quickly. Keep the tool engaged, avoid dwelling in the cut, and use high-pressure coolant where the machine supports it. A dwell of half a second in Inconel can raise the local hardness enough to blunt the next insert.
Magnesium and some castings deform from the fixture as much as the cut. They also carry more internal stress than the drawing suggests, so the rest-and-rough sequence matters more, not less.
- 1Thin wallsLess radial depth, more axial depth, higher feed per tooth.
- 2Long railsSupport along the length and machine in overlapping sections.
- 3Hard alloysDo not dwell; keep the tool engaged and cool the contact zone.
Why the first part passes and part twenty drifts
A process that is stable at 9 a.m. is not always stable at 3 p.m. Tool wear changes the cutting force, which changes the deflection, which changes the size. On a light finishing pass in aluminium, a worn tool can push the wall 0.02 to 0.04 mm more than a fresh one. On a hard material, the effect is larger and arrives faster.
Thermal growth in the machine adds to it. A spindle that has run for four hours is longer than one that started cold. If you set the tool offset at the start of the shift and never touch it, the last parts of the shift will not match the first ones.
The fix is to make the process self-checking. Set a tool life limit in the program and change on count, not on feel. Probe a datum between operations so the machine can correct for drift. Measure the first part, the middle part and the last part of a run, and keep those numbers with the setup sheet. If the spread is larger than a third of the tolerance, the process is not capable yet, no matter what the first part said.
- 1Tool life limitsChange on count, not when the finish starts to look dull.
- 2In-process probingRe-datum between operations to correct accumulated drift.
- 3Sample through the runFirst, middle and last part measurements go on the setup sheet.
Step by step: how we keep deformation inside tolerance
This is the order we apply on a part where deformation is the main risk. Skip a step and the later ones get harder.
- 11. Review the drawing for deformation riskMark thin walls under 2 mm, unsupported spans over 10× the thickness, asymmetric pockets and any tolerance tighter than ±0.02 mm. Flag the datums you will actually hold.
- 22. Check the material conditionConfirm the temper or heat-treat state. 6061-T6 behaves differently from 6061-O. Ask whether the stock was stress relieved, and if not, add a relief cycle before machining.
- 33. Rough with even stockLeave 0.5–1.0 mm on all faces and remove material from both sides where the design allows. Keep the part cool and do not push the roughing pass into the finish allowance.
- 44. Rest or stress relieveLet the part sit at room temperature, or run a controlled low-temperature relief cycle. This is where most of the movement happens, so let it happen before the finish cut.
- 55. Semi-finish, then measureTake a light semi-finish pass, let the part cool, and measure the key features. Use those numbers to adjust the finish offsets rather than guessing.
- 66. Finish with light passesUse smaller radial depth, higher feed per tooth and flood coolant on both sides of the cut. Avoid dwell. Keep the tool path continuous around thin walls.
- 77. Release the clamp before the last checkUnclamp, let the part settle, then measure. A part that only measures well in the vise is not finished.
- 88. Inspect at room temperatureRecord the room temperature with the inspection report. Measure first, middle and last parts on a run, and note tool changes.
Deformation questions we get from engineers
Can stress relief guarantee a part stays flat?
No. Stress relief removes most of the internal stress from the stock, but the machining itself introduces new stress as material is removed unevenly. It reduces the movement and makes it predictable.
In practice, a rough-rest-finish sequence with a low-temperature relief cycle between roughing and finishing brings most aluminium and steel parts into a stable state. You still measure after the part has cooled.
Does a 5-axis machine reduce deformation?
It reduces the deformation caused by re-clamping, because more faces are cut in one setup. The part stays in the same position and the same clamp pressure from the first cut to the last.
It does not remove residual stress or cutting heat. Those still need the same roughing, rest and coolant discipline. The machine changes the setup count, not the metallurgy.
What tolerance should I expect on a thin-wall part?
It depends on the wall thickness, the material and the unsupported length, so we quote it per feature rather than per part. A rigid feature on a stable part can hold ±0.005 mm. A 1 mm wall standing 30 mm tall will not.
Send the drawing with wall thicknesses called out and we will tell you which features are realistic before you order.
Is cold working or annealing better before machining?
Annealing removes stress and softens the material, which makes it easier to cut but changes the mechanical properties. Cold working keeps the strength but leaves stress in the stock.
For parts that need high strength after machining, we machine in the final temper and manage the stress with the rough-rest-finish sequence instead of annealing. For parts where the final properties allow it, annealing before machining is the simpler route.
How do you check for deformation after machining?
Parts are measured after they have cooled to room temperature, out of the fixture. We check raw material on receipt, monitor dimensions during the run, and inspect 100% before shipment, with reports on request.
For parts with a history of movement, we measure the first, middle and last part of the run and keep those records with the setup.
Does surface finishing change the dimensions?
Yes. Anodizing grows the surface layer, and it can be several micrometres per side depending on the coating type. Tight slots, fine threads and press fits need to account for that growth.
Heated finishing processes can also relax stress that survived machining, so a part that measured well before anodizing can move afterwards. Tell us which features matter and we will mask or adjust the pre-finish size.
Send the drawing before the part moves
We review deformation risk on your drawing and reply with a quotation and DFM notes within 12 hours. Production can start within 24 hours of approval, and parts ship in 3–5 days.
12-hour quote100% inspectionNo minimum order quantityNDA on request
More process notes from GreatLight
We publish setup notes, tooling trials and inspection data from the factory floor.
