CNC Machining Shaped Parts: How Geometry Decides the Process

• 1
  What matters before you quote
• 2
  Why CNC machining shaped parts is a tool-reach problem
• 3
  How setup count controls tolerance on shaped components
• 4
  Wall thickness and deflection in shaped CNC parts
• 5
  Material behavior on shaped CNC machining parts
• 6
  Inspecting shaped surfaces after machining
• 7
  Which machine fits which shape

Every shaped surface has a normal direction. To cut that surface cleanly, a tool axis has to roughly line up with that normal, or the cutter skids and leaves chatter marks. On a flat plate, all normals point up, so one spindle orientation handles the whole face. On a shaped part, normals fan out in three dimensions. A turbine-style housing might have surfaces pointing in 40 different directions.

That fan of normals is what drives machine choice. A 3-axis mill can only present the tool from above. If a surface normals points sideways, the 3-axis machine has to tilt the part instead, which usually means a second or third fixturing. A 5-axis machine tilts the tool, so the same part can often be finished in one setup. The trade is programming cost and a slightly softer machine structure.

Tool reach also sets the floor on internal radii. If a pocket has a 2 mm corner radius, the largest tool that fits is a Ø4 mm cutter. That small cutter has to remove all the stock in the pocket, at low depth of cut and slow feed, unless you pre-drill or use a larger tool for the open area first. Designers who increase corner radii to 6 mm or 8 mm often cut cycle time by 30 to 50 percent without changing function.

None of this is about machine brand or operator skill. It is geometry. Once you know the normal directions and the smallest internal radius, the process narrows itself down. That is the first thing we check when a shaped part arrives for DFM review.

• 1
  Count the surface normalsSurfaces pointing more than 45° from vertical usually need a 4th or 5th axis.
• 2
  Find the smallest internal radiusThat radius sets the maximum tool diameter, which sets the achievable feed rate.
• 3
  Check for undercutsAny feature wider below than above needs a tilted tool or a split design.

Each time a part leaves the vise and comes back, it picks up error. Chuck runout, jaw wear, chip trapped under a locating face, thermal drift between operations: they all stack. A three-setup job on a ±0.005 mm part is asking for trouble unless the datums are designed for it. A single-setup job on the same part has one stack-up, and that stack-up is usually easier to control.

The standard fix is to design datums that survive the whole process. Pick one primary face and one pair of holes, and cut every critical feature from those. If the CAD model uses a different datum than the fixture, the machinist has to translate, and translation is where mistakes live. On shaped parts, we often ask for a small flat pad on the casting or forging specifically so the first op has something solid to sit on.

5-axis work is not automatically more accurate. A trunnion table adds a rotary axis with its own backlash and thermal growth. The gain is that features cut in the same setup share the same error field, so their relative position is usually tighter than if they were cut in separate ops. Absolute position to a far-away datum can actually be looser than on a well-fixtured 3-axis machine.

The practical rule: use the fewest setups that still give the tool clear access. If a feature can be reached from the primary direction with a longer tool, and the tool is stiff enough, do it there. Save the tilting for surfaces that genuinely cannot be reached.

A wall is a beam. Its stiffness rises with the cube of its thickness and falls with the cube of its unsupported length. That is why a 2 mm wall at 10 mm tall machines cleanly and a 1 mm wall at 30 mm tall sings. The cutting force does not change much between the two; the wall's ability to resist it collapses.

There are three practical responses. First, leave more stock and take lighter finishing passes, so the wall is thick while the heavy cutting happens. Second, support the wall with sacrificial material or a wax or low-melt filler, then remove it after. Third, tilt the tool so the cutting force pushes along the wall instead of across it. On a 5-axis machine that third option is nearly free; on a 3-axis machine it usually is not available.

Designers can help before the part is ever quoted. A rib that runs to a corner is stiffer than a free-standing wall. A wall that tapers from 3 mm at the base to 1.5 mm at the top machines better than a constant 1.5 mm. And a small fillet where the wall meets the floor removes the stress riser that often causes the wall to crack during clamping.

If the function truly needs a 0.8 mm wall, that is doable. We would run it on a 5-axis center with a small-diameter tool, light radial engagement and a finishing pass at Ra 0.8–1.6 μm. It just costs more time, and the quote will show it.

Aluminum 6061 and 7075 cut freely and hold a shape well. 7075 is stronger but more prone to stress relief movement after heavy stock removal, so a shaped aerospace bracket in 7075 may need a rough, a stress-relief pause, and a finish. 6061 is more forgiving and is the default for most shaped housings.

Stainless 304 work-hardens. A shaped part with a lot of small-radius corners in 304 will wear out small tools quickly because the material gets harder every time the tool rubs instead of cuts. 303 and 17-4PH behave better. For a shaped medical instrument in 316L, we plan for slower speeds and more frequent tool changes, and we quote accordingly.

Titanium Ti-6Al-4V and Inconel are in a different category. They hold heat at the cutting edge, so tool life drops sharply. Shaped parts in these alloys usually need high-pressure coolant, conservative depth of cut and a process plan that accepts longer cycle time. The geometry is not the hard part; the thermal load is.

Plastics are the opposite problem. POM and PEEK machine cleanly but move with temperature, so a shaped part with a ±0.05 mm tolerance may need to be measured after it stabilizes. ABS and PC are soft enough that a shaped surface can be polished by hand if the finish spec allows it. Carbon fibre is abrasive and dulls tools fast.

A shaped surface is hard to measure with calipers. The normal direction changes across the surface, so a contact point can slide. The practical approach is to fixture the part on the same datums used for machining and measure with a CMM or a portable arm, taking points along known cross-sections. If the CAD model defines the surface, the CMM can compare to it directly.

For freeform surfaces, we usually report profile tolerance rather than point dimensions. A profile callout of 0.05 mm across a blended surface is more honest than a dozen linear dimensions that only sample a few spots. If the drawing shows linear dimensions on a shaped face, ask whether a profile tolerance would express the intent better.

Surface finish is measured where it matters. A shaped part might need Ra 0.2–0.8 μm on a sealing face and Ra 1.6–3.2 μm everywhere else. Polishing a whole shaped part to a fine finish adds cost without adding function. Mark the critical surface on the drawing and let the rest stay as machined.

We inspect 100% of shaped parts before shipment, with raw material check, in-process monitoring and final inspection. Reports are available on request. For shaped parts, the in-process check is often the one that catches a problem early, before the finishing passes lock in a bad surface.