Necessity of Pre-production Samples for 5-axis CNC Machining

• 1
  What Pre-production Samples for 5-axis CNC Machining Prove
• 2
  Why 5-axis Geometry Makes Sampling Different
• 3
  Fixtures, Datums and the First-article Inspection
• 4
  Where Pre-production Samples Fall Short
• 5
  Planning a Pre-production Run That Answers Questions
• 6
  Step by Step Through a 5-axis Sample Run
• 7
  When a Pre-production Sample Is Worth It

A pre-production sample is a complete part, machined on the production machine, held in the production fixture, using the released CAM program and the released stock size. It is not a render, not a 3D print, and not a part cut on a smaller machine with a promise that the big machine will behave the same way.

The value comes from what it exposes. On a 5-axis job, most scrap is not caused by a wrong toolpath in the flat areas. It comes from the interaction between the rotary axes, the workholding, and the way the part sits in space after each index. A sample puts that interaction on the table while the material cost is still one piece.

Take an aluminium housing with bores on four faces. The CAM simulation shows a clean sweep because the model assumes a perfect blank. The real blank arrives 0.4 mm oversize on one edge. Trunnion clearance shrinks, the tool shanks start to graze the fixture, and the operator has to re-datum mid-cycle. None of that appears in the software.

So the sample is a measurement of the process, not of the part alone. It answers three questions. Can we hold the datum across every index? Do we have tool access at the angles the drawing demands? Does the cycle finish without an operator intervention that changes the setup?

On a 3-axis machine, the part never moves relative to the cutter frame after the setup. Errors stay roughly constant across the part, and a single offset shift can often pull the whole feature set back into tolerance. Five-axis work breaks that assumption. The part rotates, so any error in the rotary centerline, the fixture seat, or the blank size maps onto the part at a different angle every time.

Consider a rotary table with a centerline error of 0.02 mm. On a face milled at A0, that error shows up as a simple offset. Rotate to A90 and the same 0.02 mm becomes a combination of position and angular error across the feature. Rotate again to A180 and the sign flips. The part can pass on one face and fail on the opposite face, from one root cause.

Thermal drift adds to this. A spindle running at high speed for two hours grows, and so does the workpiece. On a short 3-axis cycle the drift is small. On a long 5-axis cycle with many indexes, the first face and the last face are cut under different thermal conditions. Sampling with in-process checks shows whether that drift matters for a given part.

Tool access is the other hard limit. A deep undercut that looks clear in the model may need a long reach tool that deflects 0.03 mm under load. The sample tells you whether the finish and the position hold at that reach, or whether the feature needs a different approach angle or a shorter tool in a second setup.

The fixture is usually the largest single risk in a 5-axis job, and it is also the cheapest thing to change before a run starts. Soft jaws, a dovetail block, or a vacuum plate all have to locate the part on surfaces that the drawing actually controls. Locating on a cast surface with 0.5 mm variation is a quiet way to lose a whole batch.

Datum strategy has to match the inspection plan. If the quality team measures from a bore and the machinist set the part from an outside edge, the two numbers will disagree even when the machine is perfect. The sample run is the moment to fix one datum chain and write it down.

First-article inspection on the sample should be a full dimensional report against the drawing, not a spot check of three features. Mark each reported value with the datum it was taken from. When a value is out, the report shows whether the cause is the machine, the program, or the fixture.

Reports are available on request at GreatLight, and every part gets raw material check, in-process monitoring and final inspection before shipment. On the sample, ask for the raw numbers rather than a pass or fail stamp. Raw numbers tell you how much margin is left for the production run.

A sample is one part, and one part is a weak statistic. It cannot tell you the standard deviation of a process, and it cannot prove that a run of 10,000 pieces will hold the same tolerance. It shows that the process is capable of one good result, which is a different claim.

Samples also mislead when the material differs. A sample cut from 6061-T6 will not predict the behavior of a production lot in 7075 or 17-4PH. Harder alloys push tool deflection and heat into the cut, and the feeds that worked on aluminium can chatter on stainless. Match the sample material and temper to the production specification.

Another gap is volume. A sample is cut slowly, with an operator watching, and often with a single tool change strategy. Production runs use optimized feeds, longer tool life settings, and sometimes a different workholding to cut cycle time. If the sample and the run use different setups, the sample proves the design, not the process.

Finally, a sample cannot validate a finish that depends on a long cycle. Anodizing, hardcoat and plating results depend on surface condition and rack positions. If cosmetic appearance is critical, the sample should go through the same finishing line as production, and the finish should be judged on the sample, not on a coupon.

A sample run should be planned like a test, not like a small order. Write down the questions before the machine starts. Which feature is most likely to fail? Which dimension has the tightest tolerance? Which face is hardest to reach? The sample then has a purpose beyond producing one part.

Material selection is part of the test. Aluminium 6061-T6, 7075, 2024 and 6082 all machine differently. Stainless 303, 304, 316L and 17-4PH each need their own feeds and speeds. Titanium TC4 (Ti-6Al-4V) and Inconel move the problem to heat and tool wear. Pick the alloy that production will use, not the one that is easiest to cut.

Tool selection follows the same rule. A sample cut with a four-flute carbide end mill that is not on the production tool list proves nothing about the production process. Use the tools, the holders and the stick-out that the run will use. If a long reach tool is needed, the sample must use it.

Document the setup while it is fresh. Fixture position, torque values, probe routines, and any deviation from the CAM file belong in the setup sheet. A 5-axis setup that lives only in the operator's memory will not survive a shift change or a repeat order six months later.