How to optimize the CNC machining parameters of high-quality parts
Cutting speed, feed rate, depth of cut and tool condition decide surface finish, dimensional accuracy and cycle time. This page explains how those four variables interact, which ranges suit aluminium, stainless and titanium, and when you should stop chasing a parameter and change the setup instead. It is written for engineers and buyers who review a process before a part is released to production.

In this article
- 1
- 2
- 3
- 4
- 5
What you actually change to optimize the CNC machining parameters
A CNC program only moves a tool along a path. The physics of the cut comes from four numbers: cutting speed (surface speed at the tool edge), feed per tooth, axial and radial depth of cut, and the geometry and sharpness of the cutter. Change one and the other three respond. Push feed without raising speed and the chip thins, heat goes into the tool, and the edge breaks down early.
The goal is never the fastest single number. It is a stable cut that holds tolerance across a batch. A process that runs 20 percent faster but drifts 0.03 mm over 200 parts costs more than the slow one, because scrap and rework eat the gain.
Material drives the window more than anything else. Aluminium 6061 and 7075 tolerate high surface speed and deep radial cuts. 316L stainless work-hardens if the tool rubs, so feed must stay high enough to cut under the hardened skin. Ti-6Al-4V conducts heat poorly, so most of the heat leaves with the chip and coolant pressure matters as much as the number.
Start from the tool maker's recommended surface speed, then adjust for rigidity. A part held in a Ø400 mm rotary table with a short overhang behaves differently from a thin wall clamped at one end. On flexible setups, reduce radial engagement first, keep feed per tooth constant, and accept a lighter axial pass.
- 1Cutting speedSets edge temperature and tool life.
- 2Feed per toothControls chip thickness and work hardening.
- 3Depth of cutTrades cycle time against deflection.
- 4Tool conditionA dull edge raises force and heat at the same settings.
Cutting speed, feed rate and the chip that carries the heat
Surface speed (Vc) is the speed of the material past the cutting edge, in m/min. It is set on the spindle as rpm using the cutter diameter. In aluminium, Vc of 300–600 m/min is normal with carbide and flood coolant. In 304 or 316 stainless, 120–200 m/min is a safer band. Titanium Ti-6Al-4V usually sits at 40–80 m/min.
Feed per tooth (fz) sets the chip thickness. Too thin a chip rubs instead of cutting. That is the classic failure in stainless: the tool polishes the surface, the material work-hardens, and the next pass breaks the edge. A practical floor for stainless is around 0.05 mm per tooth; many shops run 0.08–0.12 mm per tooth for roughing.
Feed rate in mm/min is fz × number of teeth × rpm. When you raise rpm to improve finish, feed must rise with it or the chip thins and the edge rubs. This is the mistake we see most often in customer-supplied programs: spindle speed increased, feed left alone.
Coolant changes the numbers. High-pressure through-tool coolant lets you keep speed higher in deep pockets and in titanium because it clears chips and cools the edge. Without it, drop the speed 20–30 percent and shorten the axial depth.
- 1Aluminium 6061 / 7075Vc 300–600 m/min, fz 0.10–0.25 mm, air blast or flood.
- 2Stainless 304 / 316LVc 120–200 m/min, fz 0.05–0.12 mm, never let the tool rub.
- 3Ti-6Al-4VVc 40–80 m/min, fz 0.05–0.10 mm, high-pressure coolant.
- 4POM / PEEKVc 200–400 m/min, sharp uncoated edges, strong chip evacuation.
Depth of cut, radial engagement and tool deflection
Axial depth (ap) and radial width (ae) decide how much of the flute is in the material. Traditional roughing uses a small radial step and deep axial pass, often 0.5–1 × diameter axially with 25–40 percent radial engagement. High-efficiency milling goes the other way: 5–10 percent radial engagement with 1–3 × diameter axially. Both remove material fast; they stress the tool differently.
Deflection is the limit. A 12 mm carbide end mill at 3 × diameter overhang bends measurably under load, and that bend shows up as a taper or a wall that is not square. If a bore must hold ±0.005 mm, finish it with a short, stiff tool and a light pass rather than a long reach tool at high feed.
For finishing, keep radial engagement low and consistent. A constant 0.2–0.5 mm radial step on a 10 mm cutter gives a predictable Ra 0.8–1.6 μm on aluminium. If you need Ra 0.2–0.8 μm, plan a separate finishing pass at higher speed, lower feed and a fresh edge.
Thin walls are their own case. Below about 1 mm wall thickness on aluminium, cutting force pushes the wall away from the tool, and the finished wall springs back oversize. Support it with sacrificial material, or take the last passes with near-zero radial engagement from both sides.
- 1RoughingDeep axial, light radial, constant chip load.
- 2FinishingLight radial, higher speed, fresh edge.
- 3Long reachReduce ap before you reduce feed per tooth.
- 4Thin wallsSupport or machine from both sides.
Tool condition, runout and how to prove the parameters work
Tool runout is the quiet parameter. A cutter with 0.02 mm of runout loads one flute more than the others, and that flute fails first. Check runout at the holder with a dial indicator before blaming speeds and feeds. Hydraulic and shrink-fit holders hold runout tighter than a standard collet.
Coating matters for life, not for geometry. TiAlN and AlTiN coatings help in steel and stainless at higher temperatures. Uncoated polished carbide is often better in aluminium because it resists built-up edge. In titanium, a sharp uncoated or AlCrN edge with high-pressure coolant usually outlasts a coated general-purpose tool.
Verification is simple and cheap. Cut one part, measure the critical dimensions, then cut five more and measure again. If the trend is stable inside tolerance, the parameters are good. If dimensions drift with tool wear, tighten the process or plan a wear offset in the program.
We inspect 100 percent of parts before shipment, with raw material checks, in-process monitoring and final inspection. Reports are available on request. For parts that must hold ±0.005 mm, that record matters more than any single speed value in the program.
- 1Check runout firstDial indicator at the holder, target under 0.01 mm.
- 2Match coating to materialPolished carbide for aluminium, TiAlN for steel.
- 3Run a six-part trendMeasure twice, watch for drift.
- 4Log the setupSpeed, feed, tool, holder and offset in one sheet.
Starting parameters by material and operation
Ranges are starting points for carbide tooling on rigid setups. Adjust for overhang, wall thickness and coolant.
| Material | Surface speed | Feed per tooth | Typical finish |
|---|---|---|---|
| Aluminium 6061-T6 | 300–600 m/min | 0.10–0.25 mm | Ra 0.8–1.6 μm |
| Aluminium 7075 | 250–500 m/min | 0.08–0.20 mm | Ra 0.8–1.6 μm |
| Stainless 304 / 316L | 120–200 m/min | 0.05–0.12 mm | Ra 0.8–1.6 μm |
| Stainless 17-4PH | 100–160 m/min | 0.05–0.10 mm | Ra 1.6–3.2 μm |
| Steel 4140 | 150–250 m/min | 0.08–0.15 mm | Ra 0.8–1.6 μm |
| Titanium Ti-6Al-4V | 40–80 m/min | 0.05–0.10 mm | Ra 1.6–3.2 μm |
| Brass C36000 | 200–400 m/min | 0.10–0.20 mm | Ra 0.8–1.6 μm |
| POM / PEEK | 200–400 m/min | 0.10–0.25 mm | Ra 0.8–1.6 μm |
The trade-off you have to pick
If the part is rigid and the material is aluminium, push speed and axial depth and accept the tool wear. If the part is thin, deep, or made of stainless or titanium, cut the radial engagement and keep feed per tooth high; a slower spindle with a real chip beats a fast spindle that rubs.
Questions engineers ask before releasing a program
Should I optimize for cycle time or for surface finish?
Pick one per operation, not per part. Roughing is for cycle time, so use deep axial passes and a constant chip load. Finishing is for finish and tolerance, so use a light radial step, higher surface speed and a fresh edge. Mixing the two goals in one pass usually gives you neither.
If the part has one critical bore and everything else is clearance, run the whole part as roughing and give that bore its own finishing pass with a dedicated tool.
Can I use the same parameters for a prototype and a 10,000-part run?
The numbers can start the same, but the plan should differ. On a prototype you can stop and measure after each operation and adjust offsets by hand. On a production run the process has to hold tolerance while the tool wears, so you need a wear offset strategy, a tool-life count and a check interval written into the setup sheet.
We run from one prototype to 10,000+ part runs with no minimum order quantity, so the same job often moves from a hand-adjusted first article to a controlled production cell.
How much does coolant pressure change the numbers?
A lot, in deep pockets and in titanium. Through-tool high-pressure coolant clears chips and cools the cutting edge, which lets you keep surface speed 20–30 percent higher than with flood coolant alone. In Ti-6Al-4V, chip evacuation is the limiting factor more often than tool strength.
In aluminium, air blast or mist is often enough and avoids the thermal shock of heavy flood cooling on thin walls.
What tolerance can these parameters realistically hold?
On a rigid setup with a stable process, ±0.005 mm is achievable on critical features, which is ±0.0002 in. That does not mean every dimension on the drawing gets that band. It means the features you call out as critical can be held when the setup, tool and inspection plan are built for them.
Surface finish follows the same logic: Ra 0.2–0.8 μm is available where it is specified, not as a blanket condition across the part.
Which materials are the hardest to dial in?
Titanium Ti-6Al-4V and Inconel, because they hold heat at the edge and work-harden. Stainless 316L is next, for the same work-hardening reason. All three punish a light feed far more than a high one.
Magnesium AZ31B and AZ91D cut easily but need chip control for safety. We machine all of these routinely, along with 6061, 7075, 17-4PH, 4140 and engineering plastics such as PEEK.
Do you share the parameter sheet with the customer?
Yes, on request. We quote with a free DFM analysis within 12 hours and can discuss the process plan before production starts, which can begin within 24 hours. Parts ship in 3–5 days for typical work.
Uploads are secure and confidential, and an NDA is available on request if your drawings are sensitive.
Send us the part and the tolerance callouts
We review your geometry, pick the process and quote with a free DFM analysis within 12 hours.
12-hour quote100% inspectionNo minimum order quantityNDA on request