How cnc machining temperature control drives tolerances and surface finish
Thermal error is usually the largest single error source in a machine tool. This page explains where the heat comes from, what it costs in microns, and which controls pay back. Written for engineers and buyers who need to judge a process, not a sales sheet.
Why cnc machining temperature control is a tolerance problem
A machine tool is calibrated at 20 °C. That number is not a comfort setting. It is the reference at which the builder measured positioning accuracy, and the whole error budget is tied to it. Move the structure a few degrees away from that reference and the geometry changes.
The change is small per degree, but it accumulates over the length of the part and the length of the cycle. A machine quoted at 5 μm over 300 mm at 20 °C can lose most of that allowance after a 5 °C ambient drift over a long cycle. The error lands where it is hardest to see: on a bored hole, a flat face, or a long milled slot.
In a shop without active control, thermal error is commonly estimated at 40 to 70 percent of total volumetric error. That is why machinists chase the room, not only the ballscrew. Fixing the room often buys more accuracy than changing the control strategy.
The practical question is not whether the machine moves. It is whether it moves in a way you can predict and correct before the finish pass. That is what the rest of this page is about.
Thermal expansion numbers to keep in the program
Linear thermal expansion of steel is roughly 11 to 12 μm per meter per °C. Aluminum is about 22 to 24, and cast iron about 10 to 11. These are the numbers you use when a drawing calls a tight tolerance on a long part.
Run the arithmetic on a real part. A 500 mm aluminum part that rises 6 °C above its inspection temperature grows about 0.066 mm. That is already three times a ±0.02 mm tolerance. A 1 m steel ballscrew that warms 4 °C grows roughly 0.046 mm. Neither part is out of control. The temperature is.
The direction of the error matters too. A part machined warm and inspected cold will read small. A part machined cold and inspected warm will read large. Both can pass on the bench and fail at the customer.
This is why we write the inspection temperature on the report. If the drawing says 20 °C, measure at 20 °C. If it does not say, agree on a number before the first article.
- 1SteelAbout 11–12 μm/m/°C.
- 2AluminumAbout 22–24 μm/m/°C.
- 3Cast ironAbout 10–11 μm/m/°C.
Where the heat actually comes from
The machine generates most of its own heat. Spindle bearings at 10,000 to 15,000 rpm can put several hundred watts into the spindle housing. Ballscrew nuts, servo motors, hydraulic power units, and way lubrication all add heat. The structure is not a static frame.
Cutting energy also goes into the part and the tool. Roughing aluminum at 20 kW of spindle power leaves a lot of heat in the chip, but a finishing pass on hardened steel puts more into the tool and the part. The cycle matters as much as the spindle load.
The shop adds the rest. Roller doors, direct sunlight, and roof heat gain move ambient faster than the machine can follow. A machine sitting in a 12 °C shop in the morning and a 28 °C shop after lunch will not hold the same dimensions.
You cannot remove all of these. You can rank them. Spindle and coolant usually come first, ambient second, cutting heat third. Spend money in that order.
Shop-level and coolant-level controls
The cheapest effective control is a stable, conditioned shop space. Holding a machining cell within ±1 °C of 20 °C is realistic with a properly sized HVAC system. ±0.5 °C is achievable but costs noticeably more in equipment and energy. Keep the machine away from roller doors, direct sunlight, and roof heat gain.
Coolant temperature matters as much as air temperature because the coolant wets the part, the tool, and often the ballscrew and spindle housing. A chiller holding coolant at 20 ±1 °C stabilizes the cutting zone and slows heat soak into the casting.
Through-spindle coolant at 50 to 80 bar is standard on production horizontals. It improves chip evacuation and heat removal in deep pockets, and it keeps the tool at a more constant temperature. On a deep bore, that can be the difference between a straight hole and a tapered one.
None of these controls remove the need for a warm-up. They only make the starting state more repeatable.
In-process probing and warm-up cycles
On-machine probing is the practical bridge between temperature theory and accepted parts. Touch probes with 1 μm repeatability let the machine measure a datum or a critical feature and update work offsets before finishing. This corrects for slow drift rather than predicting it.
The limitation is that probing measures at the current thermal state. If the part is probed warm and inspected cold, the offset will be wrong by the growth between the two states. Probe close to the finish pass, and keep the inspection temperature in mind.
A disciplined warm-up cycle is cheap and often overlooked. Running the spindle through its speed range for 20 to 40 minutes and exercising the axes before the first cut brings the machine to a repeatable thermal state. Production shops that run the same part all day get this for free after the first hour. Job shops that start cold every morning do not.
For tight work, either warm up or probe every part. Doing neither is a bet that the morning and the afternoon will behave the same. They rarely do.
Which control pays back for which part
| Part and tolerance | Useful control | Expected result |
|---|---|---|
| 500 mm aluminum bracket, ±0.02 mm | Conditioned shop at 20 ±1 °C | Growth held to about 0.011 mm per °C |
| 1 m steel ballscrew housing, ±0.01 mm | Coolant chiller plus warm-up | Screw drift cut to a few microns |
| Deep bore in stainless, ±0.02 mm | Through-spindle coolant at 50–80 bar | Straighter hole, less taper |
| Short steel pin, ±0.05 mm | No active control needed | Shop ambient is enough |
| Prototype, one or two parts | Probe every part | Drift corrected before finish |
| High-volume run, same part all day | Warm-up plus probing | Stable offsets after first hour |
When tight temperature control is the wrong investment
If the tolerance is looser than 0.05 mm on a part under 200 mm, spend the money on inspection and fixtures first. If the tolerance is tighter than 0.02 mm over 500 mm, control the room and the coolant before you buy a new machine.
Common questions
What shop temperature should a precision CNC cell hold?
20 ±1 °C is realistic with a properly sized HVAC system. ±0.5 °C is achievable but costs more in equipment and energy.
Keep the machine away from roller doors, direct sunlight, and roof heat gain. Those three move ambient faster than any control can follow.
Does thermal compensation replace a warm-up cycle?
No. Compensation models work well for repeatable, slow drift in a stable environment and can recover 50 to 80 percent of the thermal error. They lag when the duty cycle changes abruptly.
A 20 to 40 minute warm-up brings the structure to a repeatable state. Use both when the tolerance is tight.
Should the part be probed warm or cold?
Probe close to the finish pass so the offset reflects the state the tool will cut in. If the part is probed warm and inspected cold, the offset will be wrong by the growth between the two states.
Agree on an inspection temperature before the first article and write it on the report.
How much does a 5 °C ambient drift cost in accuracy?
On a machine quoted at 5 μm over 300 mm at 20 °C, a 5 °C drift over a long cycle can consume most of that allowance.
The loss shows up on long features: bored holes, flat faces, and milled slots.
Which materials are most sensitive to temperature?
Aluminum moves about twice as much as steel per degree. A 500 mm aluminum part that rises 6 °C grows about 0.066 mm.
Steel and cast iron are more forgiving, but a 1 m steel ballscrew that warms 4 °C still grows roughly 0.046 mm.
Can GreatLight hold ±0.005 mm on temperature-sensitive parts?
Yes, with 100 percent inspection before shipment and reports on request. We run conditioned cells and probe critical features before finishing.
Send the drawing and we will confirm the process and the inspection temperature in the quote.
Send the drawing, get a process review
Quotation and free DFM analysis within 12 hours. We will tell you which controls the part needs and which it does not.
12-hour quote100% inspectionNo minimum order
More from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
