CNC Multi-Surface Processing: How Multiple Faces Get Machined in One Setup

• 1
  What CNC multi-surface processing actually means
• 2
  How 3+2 and simultaneous 5-axis change the surface result
• 3
  Fixturing and datums: where multi-surface accuracy is won
• 4
  When multi-surface processing is the wrong choice
• 5
  Holding ±0.005 mm across multiple faces
• 6
  Comparing setup strategies for multi-surface parts

CNC multi-surface processing is not one machine feature. It is a planning decision: several faces of the same part must be cut to a shared set of datums, and the choice of how many setups you spend decides the accuracy you can hold. A part with five functional faces can be made on a 3-axis mill, a 4-axis mill, or a 5-axis center. The geometry does not care. The tolerance stack does.

Every workholding change introduces a new source of error. Clamp the part in a vise, flip it, and re-touch off, and you add the vise repeatability, the parallelism of the second face, and any burr left on the locating surface. For a bracket, that is fine. For a hydraulic manifold with a 0.02 mm positional callout between ports on opposite faces, it is not.

This matters because drawing dimensions are usually given feature-to-feature, while a machine holds them feature-to-datum. Multi-surface work is the discipline of keeping the datum visible for as long as possible, so the operator never has to guess where the part actually is.

The practical result: fewer setups usually means tighter feature relationships, but not always cheaper. Long 5-axis cycles cost spindle time. The trade is setup error against cycle time, and it changes with part size, material, and quantity.

In 3+2 positioning, the rotary and tilt axes index the part into a fixed orientation, lock, and then cut with three linear axes. The tool axis is constant through the cut. This is the workhorse for multi-surface processing: five faces of a prismatic part can be reached in one setup, and each face is cut under rigid, predictable conditions. Most housings, manifolds, and brackets never need anything else.

In simultaneous 5-axis, all five axes move together. The tool axis follows the surface, which lets a ball or barrel cutter reach a contoured face at a constant lead angle. This is what makes a blade root, a turbine vane, or a deep sculpted pocket machinable without a long, thin tool. The cost is that feed rate is now a vector, not a number, and the controller must keep the tool tip speed and surface finish consistent as the axes interpolate.

The failure mode of simultaneous motion is chatter from poor dynamic stiffness at extreme rotary angles. When the table tilts 60°, the part is cantilevered away from the trunnion and the whole loop gets softer. A shop that knows this will often rough in 3+2 and finish in simultaneous, which keeps the heavy cuts rigid and only pays for five-axis interpolation on the last passes.

For a part with a single angled pad, 3+2 is almost always faster and cheaper. For a part whose surface normal changes continuously over the cut, simultaneous is the only way to avoid a series of facets.

• 1
  Choose 3+2Flat faces, drilled and tapped holes, bores, and any feature whose axis is fixed relative to a table position.
• 2
  Choose simultaneousContoured surfaces, deep cavities with short tools, and features that must be cut without a tool-axis change leaving a witness mark.

A 5-axis center removes setup error only if the part is held in a way that keeps the datum reachable. On a trunnion table, that usually means a tombstone, a self-centering vise, or a custom soft-jaw fixture that locates on a machined face rather than on raw stock. Raw stock is never a datum. Its surface is saw-cut, and its flatness varies from bar to bar.

The first operation should establish the primary datum and machine a reference face or bore that every later cut is measured from. If the part is 300 mm long and the tolerance between opposite faces is 0.03 mm, machined locating pads beat clamping on a sand-cast surface every time.

Thermal growth is the second enemy. Aluminium expands about 23 μm per meter per degree C. A 500 mm part that warms 5 °C during a long cycle moves roughly 0.06 mm, which is larger than a ±0.005 mm tolerance. Shops running tight work control coolant temperature and let the part stabilize before the finishing pass. This is not marketing; it is arithmetic.

Clamping force also deflects thin walls. A vacuum plate or low-pressure hydraulic clamp holds a 2 mm wall without crushing it, and the cut itself is lighter with a smaller radial engagement.

Single-setup 5-axis is not automatically better. A 4,000 mm long extrusion with features on two faces does not fit a trunnion machine, and trying to force it onto one will cost more than two clean setups on a large 3-axis mill with a right-angle head. Size sets the boundary before tolerance does.

Very high quantities also push work away from 5-axis. A 50,000-piece sensor housing with three faces is usually cheaper as a die casting or a progressive stamping, with machining only on the sealing face and bore. Multi-surface CNC is a fit for low to medium volume and for geometry that cannot be cast or molded without a tool that costs more than the parts.

Materials cut this both ways. Titanium and Inconel punish long 5-axis cycles because tool life drops, so a shop may deliberately split operations to keep the expensive cutter engaged only where it is needed. Free-machining brass and aluminium tolerate long cycles, which makes single-setup work more attractive.

There is also an inspection limit. If a feature on the back face cannot be reached by a CMM probe in the final orientation, you will have to inspect it in a separate setup, and the measurement uncertainty of that setup may exceed the tolerance you are trying to prove.

At ±0.005 mm, the machine is rarely the limiting factor. A modern 5-axis center with a Ø400 mm rotary table repeats well inside that. What moves is the part, the fixture, and the temperature. The sequence that works is: rough everything with stock left, stress-relieve if the material demands it, then finish all critical faces in one continuous operation without unclamping.

In-process probing helps on parts with a true position callout. Touch off the datum bore after roughing, update the work offset, and cut the finishing passes from that measured position. This absorbs the difference between where the part was modeled and where it actually sits after the first cuts.

Cutter choice matters more than people expect on multi-surface work. A long reach tool needed to clear a tall wall will deflect, and deflection shows up as a tapered bore or a mismatched surface across a split. Where possible, use the shortest tool that reaches, and accept a tool change rather than a long, weak setup.

Finally, document the datum on the traveler. If the setup sheet says 'locate on face A' and the operator locates on face B because A has a burr, the whole tolerance chain is void. A deburring step before the finish setup is a process control, not housekeeping.