ART CNC Machining: What Determines Part Quality

Most sculpted-part jobs fail at the setup stage, not at the cutter. A part with a contoured face, a blended fillet and engraved detail cannot be reached in one orientation. We start by splitting the part into the minimum number of setups that still expose every surface, then check whether each setup has a stable datum. If a datum only exists on a curved surface, we add a machining tab or a sacrificial boss so the second op has something flat to sit on.

Five-axis work earns its place when the surface normal changes faster than the tool can tilt on a three-axis machine. A long tapered wall, an undercut boss, or lettering that wraps around a cylinder all fall into that group. When the geometry stays on one face and the depth is shallow, three-axis cutting with a ball nose tool holds the same result at lower cost.

The fixture matters as much as the toolpath. Soft jaws machined to the part contour spread clamping load over a wide area, which keeps thin walls from springing. For one-off pieces we sometimes cut a dedicated pocket in a block of aluminium and hold the part with two bolts through the tab. That costs an extra hour of setup and saves a scrapped part.

A blanket ±0.005 mm callout on every dimension is not a specification, it is a wish. Machine geometry, thermal drift and tool wear all consume part of the band. On aluminium and brass with stable setups we hold ±0.005 mm on a limited number of critical features. On a 4,000 mm part the same number is unrealistic because the thermal expansion of the stock alone moves the surface.

The practical approach is to split dimensions into three groups: fits that pair with another part, cosmetic features, and free dimensions. Fits get the tight band and a gauge. Cosmetic surfaces get a profile callout and a surface finish number instead of a linear tolerance. Free dimensions can live at ±0.1 mm or looser, and loosening them cuts cycle time because the tool can run at a higher feed.

Material behavior sets the floor. Titanium Ti-6Al-4V and 17-4PH stainless move during and after cutting, so a feature that measures in tolerance on the machine can leave the band after the part cools. We rough, stress-relieve where the geometry allows it, then finish. On thin aluminium plates we take lighter finishing passes rather than one deep pass, which keeps the heat out of the part.

A sharp edge is the first thing to disappear in finishing. Bead blasting rounds a 0.2 mm edge, and anodizing builds a layer that blurs 1.5 mm lettering. If the drawing depends on crisp lines, we mask those areas or move to a chemical conversion coat that adds less thickness. The finishing step is chosen with the feature size, not after the fact.

For aluminium, clear or coloured anodizing at 10–25 μm keeps engraved detail readable. Hardcoat adds wear resistance but changes the dimension by a larger amount, so critical fits need masking. On stainless, bead blasting followed by passivation gives a uniform matte without touching the geometry. Polishing brings stainless to a mirror look but softens every corner.

Laser marking is the last step because it cannot be undone. We ask for a minimum character height of 1.5 mm and a vector file, not a raster image. Smaller text is possible on flat faces with a good contrast, but it fades quickly on curved or textured surfaces.

Machining is the wrong answer for some shapes. A lattice of 2 mm struts, a hollow shell with internal channels, or a part with fifty identical features is usually faster and cheaper as a casting or an injection-moulded part. Machining earns its cost when the quantity is low, the tolerance is tight, or the geometry needs to be revised after the first fit check.

A hybrid route often makes sense. We machine a prototype, the customer tests the fit, and once the design freezes we move the same geometry to die casting or vacuum casting for the production run. That keeps the early parts accurate and the later parts cheap. Nothing about the early machining work is wasted.

Send the 3D file and the critical dimensions, not just a 2D drawing. A STEP file lets us check tool reach and wall thickness before quoting, and we return a DFM note within 12 hours. If a wall will chatter or a hole depth exceeds what the tool can reach, we say so at that point rather than after the first cut.