CNC Milling Savannah: Precision Parts

• 1
  What actually happens inside a CNC mill
• 2
  What three-axis milling can and cannot hold
• 3
  Where five-axis milling earns its cost
• 4
  Material behavior changes the plan
• 5
  How tolerance and finish are verified
• 6
  Choosing the right setup for the part

Milling is subtractive. A spinning cutter with multiple flutes moves through a block of metal and shears away material in chips. The CAD model becomes a toolpath, the toolpath becomes G-code, and the machine follows that code without a hand on a crank. Repeatability comes from the machine, not from the operator's eye.

Every pass has a cutter, a feed rate, and a spindle speed. Those three values decide whether you get a clean wall or a chatter mark. Aluminium 6061 is often cut at 3,000–10,000 rpm with a two or three flute carbide end mill. 17-4PH stainless runs far slower, often 200–500 rpm on the same feature, because the material work-hardens if the tool rubs instead of cuts.

Heat is the quiet problem. Roughly 80% of the heat in a cut leaves with the chip, so chip evacuation matters more than coolant volume. Deep pockets with poor evacuation dull tools fast and push dimensions around. That is why a part that looks simple on screen can be slow to make.

• 1
  CutterCarbide end mills, ball nose, and face mills cover most features
• 2
  AxesThree linear axes, or five when the tool must reach underneath
• 3
  SetupEach new face usually means a new workholding position

A three-axis mill moves in X, Y and Z only. The tool always approaches from one direction. That is fine for plates, brackets, housings with open pockets, and most parts you can reach from six sides with a few setups. It is the cheapest way to remove material and the easiest to inspect.

The limits show up in three places. Deep cavities with small corner radii need long, thin tools that deflect. Undercuts cannot be reached at all. Features on five or six faces need a separate setup each time, and every setup adds a small position error. A part with four setups can accumulate 0.02–0.05 mm of stack-up even when each setup is dialed in well.

On a three-axis machine we hold ±0.005 mm on critical features when the geometry cooperates: rigid part, short tool, shallow depth of cut, and a single setup for the tight tolerances. When the geometry does not cooperate, more setups do not fix it. The part gets rotated to a five-axis machine instead.

• 1
  Good fitPlates, manifolds, brackets, open pockets, flat sealing faces
• 2
  Poor fitUndercuts, deep thin ribs, features on five or more faces
• 3
  Setup costEach additional face adds fixturing time and position error

A five-axis machine adds two rotary axes. The tool can tilt and the table can turn, so the cutter reaches the part from almost any angle without re-fixturing. On a medical implant or a turbine bracket, that removes three or four setups and the errors that come with them.

The second benefit is tool life and surface finish. Tilting the tool so it contacts the flank rather than the tip spreads the load and lets you use a shorter, stiffer cutter. Long tools chatter; short tools cut. On deep cavities, that single change often turns a scrapped part into a good one.

The third benefit is geometric freedom. A curved channel, a blended fillet, or an impeller blade can be cut in one continuous pass instead of being approximated by a series of flat cuts. The surface is smoother and the part is stronger because the toolpath does not leave stair steps.

Five-axis is not automatically better. Programming takes longer, machines cost more per hour, and simple prismatic parts gain nothing. If a part can be made on three axes in one setup, that is the cheaper route.

• 1
  Best fitImpellers, medical implants, aerospace brackets, complex housings
• 2
  Main gainFewer setups, shorter tools, continuous contoured surfaces
• 3
  Not worth itFlat plates and simple blocks that fit one three-axis setup

The same drawing in aluminium and in titanium is two different jobs. Aluminium 6061 and 7075 cut fast and hold tight tolerances easily. Titanium Ti-6Al-4V conducts heat poorly, so the cutting edge stays hot and tools wear quickly. Feeds and speeds drop, cycle time rises, and cost follows.

Stainless is the middle ground with a catch. Grades like 304 and 316 work-harden if the cutter rubs, so a light pass at the wrong feed does more damage than a heavier one. Grade 17-4PH in the H900 condition is harder again and usually needs carbide with a coating and a rigid setup.

Plastics behave differently. POM and PEEK machine cleanly but move with temperature, so a part measured hot may be out of tolerance cold. ABS and PC soften if the cutter dwells. For thin walls in any plastic, light finishing passes and sharp tools matter more than spindle speed.

• 1
  Easy6061, 7075, brass C36000, POM, ABS
• 2
  Moderate304, 316L, 4140, 17-4PH, PEEK
• 3
  DifficultTi-6Al-4V, Inconel, magnesium AZ31B

A tolerance only means something if it is measured. We check incoming raw material, monitor dimensions during the run, and inspect the finished part before it ships. Reports are available on request. On a ±0.005 mm callout, that usually means a coordinate measuring machine rather than calipers.

Finish is specified as Ra, an average roughness value. As-machined surfaces sit around Ra 1.6–3.2 μm. A finer finishing pass gets Ra 0.8–1.6 μm, and a polished or lapped surface reaches Ra 0.2–0.8 μm. Each step adds time. Ask for the finish the function needs, not the finest value available.

Surface finish and tolerance are linked. A wall that measures in tolerance on a rough surface may drift once a finishing pass removes another 0.05 mm. Finishing stock has to be planned into the roughing toolpath, not added at the end.

For parts that must seal, mate, or slide, call out the finish on the functional face only. Blanket-finishing an entire part to Ra 0.2 μm multiplies cost for surfaces that never touch anything.

• 1
  As-machinedRa 1.6–3.2 μm, general structural faces
• 2
  Fine finishRa 0.8–1.6 μm, sealing faces, bearing seats
• 3
  PolishedRa 0.2–0.8 μm, sliding surfaces, optical housings