CNC machining in product development: how it actually works

• 1
  What happens when a cutter meets the blank
• 2
  How tolerance stack-up shapes your drawing
• 3
  Workholding decides how many setups you pay for
• 4
  Material choice sets the process window
• 5
  Why prototype rounds get faster, not slower
• 6
  When to machine a prototype, and when not to

The process is subtractive. A rotating cutter follows a toolpath generated from your CAD model, and material leaves the blank as chips. The machine does not care whether the shape is a bracket or a housing. It only cares whether the cutter can reach the surface without hitting the part or the fixture.

Speed comes from three places, not one. Toolpath generation is software. Metal removal rate is the combination of spindle speed, feed and depth of cut. Setup is how fast a blank can be located and clamped. A shop that programs slowly but cuts fast still ships late.

The practical ceiling on any job is stiffness. A long, thin cutter deflects under load, and the deflection shows up in the wall as taper, chatter or a size that drifts from part to part. Short, rigid tools with a large core diameter cut more predictably than long reach tools, even at lower feed.

Heat is the second ceiling. Aluminium 6061 and 7075 conduct heat away quickly, so they tolerate aggressive parameters. Titanium TC4 and Inconel do not. They hold heat at the cutting edge, which shortens tool life and can work-harden the surface if the cutter rubs instead of shearing.

Chip evacuation decides whether a deep pocket is a ten-minute job or a two-hour job. Chips that recut under the tool double the load and wreck the finish. Through-spindle coolant and a toolpath that lifts chips out solve most of it. On deep ribs, the toolpath matters more than the machine.

A drawing that says ±0.005 mm everywhere is not a better drawing. It is an expensive one. Each tight callout adds inspection time, may add a second setup, and forces the shop to hold the feature while the rest of the part is cut. Most designs need tight control on two or three features, not twenty.

Stack-up is the real question. If a bore and a mating shaft each carry ±0.05 mm, the assembly can see 0.1 mm of variation before any thermal effect. Chasing ±0.005 mm on the bore only helps if the mating part is held to a matching band. Otherwise you paid for accuracy the assembly cannot use.

Datums deserve the same scrutiny. A feature dimensioned from a rough cast surface inherits the casting variation. A feature dimensioned from a machined datum does not. Choose functional datums first, then assign tolerances to the features that control fit, and leave cosmetic surfaces loose.

Surface finish and tolerance are separate purchases. A Ra 0.8–1.6 μm finish on a sealing face is worth the cost. The same finish on a hidden internal wall is not. Specify finish only where a seal, a bearing or a sliding contact touches the part.

When in doubt, ask for a DFM review before you release the drawing. A machinist reading the model will spot a corner radius that needs a 2 mm cutter, or a pocket depth that exceeds five times the tool diameter, and those two details decide whether the part is easy or painful.

Every additional setup adds a new source of error and a new queue. A part that can be cut from three sides in one vise is cheaper than the same part split across two fixtures and a re-clamp. Designers rarely see this because the CAD model hides the fixturing.

Five-sided access changes the math. On a simultaneous 5-axis center, a Ø400 mm rotary table lets the tool reach undercuts and angled faces without releasing the part. That preserves the datum relationship between features that would otherwise be re-established.

Thin walls are a workholding problem before they are a cutting problem. A 0.8 mm wall on a 100 mm aluminium housing will move when the vise releases, even if it measured perfectly while clamped. Adding a sacrificial web, a boss, or a light finishing pass after stress relief keeps the wall where the drawing says.

For long parts, travel matters. A machine with 4,000 × 400 × 150 mm of travel handles long extrusions and rails that would otherwise need repositioning. Repositioning a 4,000 mm part mid-cut is a tolerance risk that is easy to avoid at the quoting stage.

Ask your supplier how the part will be held. If the answer is vague, the first article will reveal it. A short conversation about fixtures before cutting saves a revision cycle later.

Aluminium 6061-T6 is the default for most housings, brackets and fixtures. It cuts fast, holds a good finish, and anodizes predictably. Grades 7075 and 2024 are stronger but less forgiving: 7075 machines well yet can be stress-corrosive if the grain direction is wrong, and 2024 needs a protective finish.

Stainless 303 is the free-machining grade and the easiest to turn. Grade 304 and 316 resist corrosion better but work-harden quickly, so the cutter must keep moving. 17-4PH gives high strength after heat treatment and is common on medical and aerospace parts where a specific hardness range is required.

Plastics behave differently from metals in ways that surprise first-time buyers. POM and PEEK hold tight tolerances well. ABS and PP deflect under clamping and expand with heat, so the finished size depends on how the part was held and cooled. On long plastic parts, allow a day for the material to settle before final inspection.

Material availability can set the schedule more than the machine does. A standard 6061 plate is often in stock; a specific titanium grade may not be. Ask about stock before you commit to a material callout, especially on a revision that must ship in the same week.

The material list is not a menu to shop from casually. Each grade carries its own tooling, coolant and finishing implications. Pick the grade that meets the function, then let the process window follow.

The first round tests whether the concept fits together. The second round tests whether it can be assembled by a person on a bench. The third round tests whether the tolerances hold when ten units are built instead of one. Each round changes what the shop needs to control.

Rapid turnaround is not a single number. It is the sum of quoting, programming, cutting and inspection. When the quote and DFM feedback arrive within 12 hours and production can start within 24 hours, the calendar time between two design revisions shrinks to days rather than weeks.

Do not change everything between rounds. If round two changes the material, the wall thickness and the mounting pattern at once, and the part fails, you learn nothing about which change caused it. Hold two variables steady and move one.

Inspection data belongs in the next revision. If the first article shows a bore running 0.03 mm over nominal, correcting it in the model before round three is cheaper than tightening the tolerance and hoping the shop holds it.

Keep the same shop across rounds when you can. Fixtures, tooling and setup notes carry over, and the second article often costs less than the first because the learning is already paid for.