Master On-Site CNC Processing: How It Works and When to Use It

• 1
  What On-Site CNC Processing Actually Means
• 2
  How the Cutting Process Behaves On the Floor
• 3
  Matching the Machine to the Part
• 4
  Material Behavior and Surface Finish
• 5
  When On-Site Machining Pays Off and When It Does Not
• 6
  Inspection, Documentation and Process Control
• 7
  On-Site Machining vs Outside Shop: Which Fits

On-site CNC processing is subtractive machining performed at the point of use: the machine sits in your building, cuts your part, and the part never enters a shipping crate. The spindle, the fixture and the operator are all inside the same four walls as the assembly line that consumes the output.

The phrase is often stretched to cover two different setups. In the first, a shop installs a machining center inside a customer plant and runs it there. In the second, a supplier keeps the machine at its own plant but schedules production so tightly that parts arrive as if they were made next door. Only the first removes transport from the process; the second just compresses the calendar.

That distinction matters because the two setups fail in different ways. A machine installed in your plant is limited by the floor space, power and chip handling you can give it. A supplier running tight schedules is limited by freight, customs and how well the drawings travel. Both can hold ±0.005 mm. Neither can fix a bad datum.

For most buyers, the practical question is not which label applies. It is whether the part needs to be machined within walking distance of the assembly station, or whether a qualified shop three time zones away can hold the same tolerance once the process is locked.

Every on-site CNC processing job comes down to the same loop: the controller reads a toolpath, the servo drives move the axis, the tool removes material, and the resulting geometry is measured against the drawing. The loop is mechanical, so the errors are mechanical too. Thermal growth, tool wear, fixture deflection and spindle runout all show up in the finished part.

On-site work adds one more variable: the environment is whatever your plant already is. A machining center rated at ±0.005 mm will hold that number in a temperature-controlled room. Put the same machine next to a loading dock in August and the bed grows. Aluminum expands roughly 23 μm per meter per °C, so a 1,000 mm part and a 10 °C swing move the number more than the machine tolerance does.

This is why on-site shops usually ask about ambient conditions before they quote a tight tolerance. If the part is 50 mm long, thermal drift is noise. If it is 1,000 mm long and the drawing calls ±0.02 mm across the length, the room becomes part of the process.

Chip evacuation is the other floor-level constraint. A 16-machine cell generates a lot of swarf, and an on-site installation has to route conveyors, coolant and mist collection through an existing building. Retrofitting that into a leased space is often the largest hidden cost of the whole project.

Three-axis work covers most prismatic parts: plates, housings, brackets and fixtures with features reachable from one direction. If the part needs holes on four faces, a three-axis machine means multiple setups, and every setup adds a datum error. That is where four-axis and five-axis machines earn their cost.

Five-axis simultaneous machining lets the tool stay normal to a curved surface while the table rotates. Impellers, turbine blades, medical bone plates and complex automotive brackets fall into this group. The trade-off is programming effort: a five-axis toolpath takes longer to prove out, and the first article usually costs more than the same part on a three-axis machine.

Size drives the choice as much as geometry. Our largest travel is 4,000 × 400 × 150 mm, which suits long extrusions, rails and structural profiles. Mid-size work sits in the 750 × 1,150 × 550 mm and 600 × 600 × 600 mm envelopes. Compact parts run on 500 × 500 × 450 mm and 500 × 310 × 200 mm machines, where the smaller envelope usually means faster acceleration and better surface finish.

A Ø400 mm rotary table covers most round and indexed work, including mill-turn parts that need turning and milling in one setup. Mill-turn removes a handling step, and handling is where round parts lose concentricity.

Aluminum is the default for on-site work because it cuts fast and holds tolerance without much fuss. The 6061 and 7075 families cover most brackets, housings and prototypes; 2024 and 5052 show up in aerospace and sheet-derived parts. Aluminum also anodizes cleanly, so a machined surface can become the finished surface.

Stainless and steel change the calculus. Grades 303, 304 and 316 machine at lower feeds and generate more heat, so coolant and rigidity matter more than speed. Tool steel and 4140-class alloys often need pre-hardened stock or a heat-treat step after machining, and heat treatment moves dimensions, which means the finishing pass has to come after.

Titanium and Inconel sit at the difficult end. Ti-6Al-4V conducts heat poorly, so the cutting edge absorbs it and tool life drops. Inconel work-hardens if the feed is too light. Both are machinable at ±0.005 mm, but cycle times are long and the process window is narrow.

Surface finish follows the same logic. As-machined surfaces land around Ra 1.6–3.2 μm. A high-quality finish at Ra 0.8–1.6 μm is a normal production target for sealing faces and bearing bores. Fine finishing down to Ra 0.2–0.8 μm usually needs a separate pass with a smaller stepover, which adds time but not new fixturing.

On-site work pays off when the part and the assembly line are coupled. If a fixture has to be trimmed to fit a weldment that is still on the floor, or a prototype bracket has to be tested the same shift it is drawn, a machine within walking distance removes days of freight and rework. Iteration speed is the real product being bought.

It also pays off when the part is too large or too awkward to ship safely. Long rails, large frames and one-piece structures that exceed a standard pallet are cheaper to machine where they will be installed than to crate, insure and move twice.

It does not pay off when the part is a stable, repeatable design. A bracket that has not changed in three years, running at 10,000 pieces a year, belongs on a dedicated machine in a shop that does nothing else. Moving that work on-site adds floor space, staffing and maintenance without improving the part.

Nor does it pay off when the tolerance depends on environment you cannot control. If the drawing calls for a tight tolerance across a long dimension and your building swings 10 °C between shifts, an outside shop with a controlled room will hold the number more reliably, even with shipping in the middle.

On-site machining only works if measurement keeps pace with cutting. A part that is machined on the floor and then sent out for inspection has lost most of the benefit, because a failed dimension becomes a second trip instead of a quick correction. Inspection has to be local.

Our own flow runs raw material check, in-process monitoring and final inspection, with 100% inspection before shipment and reports on request. That structure exists because the failure modes are predictable: a wrong tool offset, a loose clamp, a worn insert. In-process checks catch all three before the part is finished.

Documentation matters more when the machine is remote from the engineering team. Setup sheets, tool lists and inspection records turn a one-off job into a repeatable process, which is what allows a part to move between machines without requalification.

Certification supports this rather than replacing it. We hold ISO 9001:2015, IATF 16949:2016, ISO 13485:2016 and ISO 27001:2022. The first covers general quality, the second automotive, the third medical devices, and the fourth information security for customer data and drawings.