CNC Processing Test Preparation

• 1
  What CNC processing test preparation actually validates
• 2
  Why a dress rehearsal is cheaper than a scrap pile
• 3
  Stage 1: virtual validation before the spindle turns
• 4
  Stage 2: coupon and fixture prove the setup
• 5
  Stage 3: the first article run and its report
• 6
  When a test run is not worth the schedule
• 7
  What the numbers on a test report mean
• 8
  Is a full test protocol justified?

CNC processing test preparation is the work that happens between a released CAD model and the first full production batch. It is a short, deliberate run that forces every assumption in the process plan to prove itself on real material with real tooling. Nothing about the part changes. What changes is that the unknowns get found early.

A drawing sets nominal dimensions and tolerances. A CAM file sets toolpaths, feeds and stepovers. Neither one tells you how a 6061-T6 wall behaves when a 12 mm end mill takes 0.3 mm radial cuts at 40 mm depth, or whether the fixture you sketched in the setup sheet will hold the part without squeezing it oval.

So the test run answers four questions in order. Can the machine reach every feature? Will the material hold the geometry after cutting? Does the workholding repeat when the part is unloaded and reloaded? Does the finished part measure inside tolerance on the critical dimensions? Each one fails in a different place, and each one costs a different amount to fix.

The scope is intentionally narrow. A protocol that tries to validate everything usually validates nothing, because nobody reads the report and nobody signs off on the numbers. Pick the features that carry the function of the part and test those.

The economics are simple. A test run of 5 to 10 units consumes a small amount of material and a few hours of spindle time. A failed production run of 500 units consumes the material, the machining hours, the inspection hours, and the calendar time you needed for assembly. The test run also consumes the schedule, which is the part people forget.

There is a second cost that never shows on a quote: rework loops. When a tolerance is wrong on a production batch, the options are scrap, weld and remachine, or request a deviation from the customer. All three damage the relationship, and the third one damages it slowly.

Material behavior is the usual source of surprises. Certified aluminum is not uniform in residual stress. A 7075 plate that machines flat on day one can bow after a week when internal stress redistributes. Stainless 17-4PH in the H900 condition cuts differently from the annealed bar, and thin sections move more than thick ones.

That is why the test run uses the same heat lot, the same stock condition and the same machine as production. A test on a different lot is a test of a different problem. Reading a test report that does not name the lot number tells you almost nothing about your parts.

The first stage costs nothing but software time, so it runs on every job. The CAM programmer simulates the full toolpath, including rapid moves and tool changes, and checks for gouges, collisions and unreachable geometry. On a 5-axis job this is where the rotary table and the part blank are checked against each other.

Simulation catches interference that is invisible in a 2D drawing. A deep pocket with a Ø6 mm tool may need a tool extension that then collides with the wall on the retract move. A undercut feature may need the A axis to swing past 90°, which the fixture may not allow.

Machining strategy gets reviewed at the same time. Adaptive clearing with constant chip load keeps heat out of the cut. A light radial step with a large axial depth is usually better for deep cavities than a heavy radial cut. For thin walls, a finishing pass with a reduced radial engagement limits deflection.

The stage produces a setup sheet: tool list, holder list, zero points, work offsets, and the order of operations. If the shop floor has to guess anything from that sheet, the CAM stage is not finished.

The second stage moves to metal, but still not to the part. A coupon cut from the same stock and heat lot runs the critical operations: the deepest pocket, the thinnest wall, the tightest bore. Measure the coupon and you learn whether the tool is wearing, whether coolant is reaching the cut, and whether the surface finish matches the drawing.

Workholding is tested on the coupon as well. A custom fixture for an organic or undercut part may use soft jaws, a vacuum plate, a 3-jaw or a dedicated tombstone. The test checks that the part seats fully, that clamps do not lift it off the datum, and that the five-axis table can still reach every face.

Clamping force is the quiet failure. An over-tightened vise on a thin-walled aluminum housing closes the bore by 0.02–0.05 mm and then it springs back after unclamping. The part measures oversize in the free state and undersize in the fixture. A torque wrench, or a strain gauge on critical jobs, keeps the force repeatable.

For parts with a datum that must survive multiple setups, the fixture design decides whether the second operation can find the first one. Locating pins, shoulder stops and pre-machined reference faces all reduce the stack-up. That geometry is easier to fix on a drawing than on a fixture that already exists.

The first article is a small batch, typically 5 to 10 units, cut under production conditions: production speeds, production tooling, production coolant, production operator instructions. If the shop speeds up the test run to finish it faster, it is testing a process nobody will use.

Measure every dimension on the drawing, not just the critical ones. Full dimensional reports expose stack-up errors that a spot check misses. On tight work, the report also names the instrument: a micrometer for an outside diameter, a bore gauge or an air gauge for an inside diameter, a CMM for position and profile.

• 1
  Material certificateHeat lot and condition recorded against the drawing callout.
• 2
  Setup sheet sign-offZero points, tool offsets and work offsets verified at the machine.
• 3
  Dimensional reportEvery drawing dimension with the instrument used and the result.
• 4
  Surface and finish checkRa measured or compared against a certified roughness specimen.

Not every job earns a full protocol. A one-off bracket in 6061, with generous tolerances and simple 3-axis geometry, usually does not need a 5-piece test run. A conversational check of the CAM file plus a single first-off part is enough.

The decision turns on three factors. How tight is the tolerance relative to the process capability? How expensive is the material and the machining time per unit? How much does a failure cost the customer downstream, in assembly line stoppage or a field recall?

If tolerance is tight, unit cost is high, or failure is expensive, run the protocol. If all three are low, skip it and inspect the first part carefully. Spending a week of engineering on a part that costs $8 to make is a different kind of mistake.

Parts that always justify the protocol: thin-walled housings, 5-axis parts with undercuts, anything in titanium or Inconel, medical and aerospace components with traceability requirements, and any geometry where the fixture itself is a design problem. On those jobs the fixture often takes longer to prove than the part takes to cut.

A test report is only useful if the numbers are traceable. A dimension without an instrument is a claim, not a measurement. On a Ø20 H7 bore, a plug gauge tells you go or no-go; a bore gauge tells you 20.012 mm. Those are different levels of information, and only one of them lets you adjust a tool offset.

Process capability matters more than a single reading. If a feature measures 20.008 mm on five parts with a spread of 0.004 mm, the process is centered and stable. If the same five parts read 19.998, 20.006, 20.001, 20.009 and 19.997, the mean is fine but the spread will produce out-of-tolerance parts at volume.

Surface finish follows the same logic. Ra 0.8–1.6 μm is a normal machined finish; Ra 0.2–0.8 μm needs a finer stepover, a sharper tool and often a separate finishing pass. A test run reveals whether the finish comes from the process or from a lucky tool.

Temperature is the last variable people account for. A part measured hot off the machine reads larger than it will at 20 °C. Aluminum moves about 23 μm per meter per degree, so a 1,000 mm part that is 8 °C above room temperature measures roughly 0.18 mm long. Measure after the part stabilizes, or compensate deliberately.