CNC Machine Tool Online Detection System: How It Works on the Shop Floor
A CNC machine tool online detection system measures the part or the cutting tool while it is still clamped on the machine. This page explains the probe types, how the measurement loop closes, and where the accuracy limits sit. It is written for engineers and buyers who need to decide whether on-machine probing replaces or only reduces the inspection room.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
What a CNC machine tool online detection system actually is
A CNC machine tool online detection system is a measurement loop that runs inside the machine tool, not beside it. A touch probe or a non-contact sensor is loaded into the spindle or a turret station, the control moves the axis to a known position, and the contact event is recorded as a skip signal. Software converts that signal into a coordinate, compares it with the nominal value, and can update the work offset or offset the tool before the next pass.
The word online is the important part. The part stays clamped in the same fixture, at the same temperature, with the same zero point used for cutting. No re-fixturing error is introduced between machining and measurement, and no operator carries the part to a different room. That is why the method catches a drifting tool or a loose clamp before a whole batch is scrapped.
It is not a coordinate measuring machine and it does not replace one. A CMM runs in a temperature-controlled room with granite ways and calibrated stylus geometry. An on-machine probe runs on a machine that has just been cutting, with chips nearby and thermal growth in the ballscrew and spindle. The two systems answer different questions.
The practical question is always the same: what do you need to know, and how fast do you need to know it? If you need to know whether this specific bore will clean up at nominal minus 0.02 mm, on-machine probing answers it in seconds. If you need a certified dimensional report for a customer submission, the CMM still does that job.
One more boundary. On-machine probing measures features the stylus can reach. Deep bores, undercuts, narrow slots and internal threads are often out of reach, and a probe cannot see a surface finish value, a hardness reading or an internal defect. Those still need separate instruments.
Probe types and how the measurement loop closes
The most common setup is a kinematic touch-trigger probe. Three pairs of balls and cylinders sit in contact under spring load. When the stylus touches the workpiece, the contacts break, and the interface sends a skip signal to the control within a few microseconds. The position of the axis at that instant is stored, corrected for the stylus ball radius and the probe trigger direction, and reported as a point.
Transmission matters as much as the mechanism. Infrared probes need line of sight between the probe and a receiver mounted in the work envelope. Radio probes work through the enclosure and are better for large parts or deep cavities where the spindle blocks the beam. Both need a clean, charged battery, and a low battery is a common cause of intermittent readings that look like machine error.
Measuring the tool is a separate loop. A tool setter mounted on the table measures length and diameter by touching the tool tip, and the control writes the value into the tool offset table. This catches a chipped insert or a tool loaded in the wrong holder before the first cut. On a 16-station turret, a single mis-set tool can scrap a whole first article.
Scanning probes and non-contact sensors close the loop differently. A scanning probe deflects continuously and streams thousands of points, which is useful for profile comparison against a CAD model. A laser sensor measures without contact, which suits soft materials, thin walls and parts that cannot take stylus force. Both need more setup and more data handling than a simple touch trigger.
After the data is captured, the control decides what to do with it. Three actions are common: report the deviation and stop for an operator decision, update the work offset automatically, or update a tool offset so the next pass removes the correct amount of stock. The third action is where the real value sits, because it corrects the process, not just the part.
Where the accuracy limits sit and what drives them
A repeatable touch-trigger probe on a good machining center typically resolves position to a few microns. That is not the same as overall measurement uncertainty. The machine itself contributes positioning error, the probe contributes trigger scatter, and the thermal state of the machine contributes a slow drift that can exceed both.
Thermal drift is the largest and most underestimated term. As the spindle and ballscrews warm up, the machine geometry changes. A probe cycle run ten minutes after a cold start can read differently from the same cycle run three hours later. Shops that rely on on-machine data usually run a warm-up cycle and re-check a master artifact at intervals through the shift.
Stylus choice sets a hard limit. A long stylus bends more under contact force, and bending shows up as a false coordinate. A short, stiff stylus with a ruby ball at Ø2 mm or Ø4 mm is the normal choice for bores and faces. If a feature needs a 100 mm stylus to reach, expect the uncertainty to grow and plan for a slower feed into the touch.
Temperature also acts on the part. Aluminium expands roughly twice as fast as steel for the same temperature change. A part measured at 28 °C and then inspected at 20 °C will not match. On-machine measurement is only trustworthy when the part and the machine are near the same temperature, and when the drawing states the reference temperature.
The honest summary: an on-machine probe is excellent for relative decisions and poor for absolute certification. It tells you whether this bore is on size, whether this tool has worn, whether this fixture has shifted. It does not hand you a calibration certificate. Pair it with a CMM or a height gauge when the drawing calls for a certified number.
How the cycle is built into a production run
The measurement cycle is written as a subprogram and called from the main program at fixed points. A typical sequence starts with the probe calibrating against a master sphere or ring gauge, then measuring the datum face and two edges to establish the work offset, then measuring the first part feature by feature. The control then either adjusts or alarms.
First-article measurement is the highest-value use. Instead of cutting a part, unclamping it, carrying it to inspection and waiting, the operator probes the critical features while the part is still held. If a bore is 0.03 mm undersize, the control can offset the tool and re-cut in the same setup. That single change removes most of the scrap risk on a new program.
In-process measurement runs between operations on the same part. After roughing, the probe checks stock left on a critical wall; the control adjusts the finishing pass accordingly. This is where adaptive behaviour pays off on castings and forgings, because incoming stock varies from part to part and a fixed finishing pass either cuts air or overloads the insert.
Tool wear monitoring is the routine use that runs every day. The probe measures a reference feature at a set interval, and the control tracks the trend. When the trend crosses a limit, it flags the tool for change. This replaces a fixed tool-life counter, which either changes tools too early or too late.
Post-process verification is the version that supports the shipping decision. After the last operation, the probe checks the drawing dimensions that matter and writes them to a log. For a run of 10,000 parts this gives a per-part record without moving every part to a CMM. Final inspection before shipment still happens, and reports are available on request.
When on-machine detection is the wrong answer
If the drawing requires a certified dimensional report, a probe reading from the machine does not satisfy it. The report has to come from a calibrated instrument with a traceable chain. Using machine data as a substitute is a documentation problem, not a technical one, and it shows up at audit.
If the part is thin-walled and flexible, stylus force can deflect it. A 0.5 mm aluminium wall moves under a few newtons of contact. The probe reads the deflected shape, not the free shape, and the reading is repeatable but wrong. Non-contact sensing or a CMM with low force is the better route.
If the feature is deep, small or undercut, the stylus simply cannot reach it. A Ø1.5 mm stylus in a 60 mm deep bore will bend and may snap. Reaching into a cross-hole or an internal groove needs custom styli, and even then the uncertainty grows with length.
If the machine is busy, every probing cycle is spindle time that is not cutting. A full first-article probe routine can take several minutes. On a high-volume run with a stable process, that time may not pay back, and a periodic CMM check on a sample is cheaper.
If the material is abrasive or the coolant is misty, probe contacts and infrared windows get contaminated. Radio transmission and sealed contacts handle this better, but maintenance intervals shorten. Budget for cleaning and calibration, not just the purchase.
On-machine probing compared with a CMM and manual gauging
Pick the method by the question you need answered, not by habit.
| Method | Best for | Typical limit | Setup cost |
|---|---|---|---|
| On-machine touch probe | Work offset setting, tool wear, first article | Few microns repeatability, thermal drift | Subprogram plus probe calibration |
| On-machine scanning probe | Profile comparison against CAD | Large data sets need filtering | Higher, needs software support |
| CMM in a metrology room | Certified dimensional reports | Part must be moved and soaked | Highest, needs controlled room |
| Manual gauging | Quick spot checks, simple features | Operator-dependent, slow on complex parts | Lowest, hand tools only |
The clear trade-off
Use on-machine probing when you need fast relative decisions on a clamped part, such as work offset, tool wear and first-article correction. Use a CMM when you need a certified, traceable dimensional report. Most shops need both, and the probe pays for itself on first-article scrap alone.
Questions engineers ask before specifying a probe
Does an online detection system remove the need for final inspection?
No. On-machine probing reduces the number of parts that reach final inspection with a defect, but it does not replace it. At GreatLight, every shipment still goes through raw material check, in-process monitoring and final inspection, with reports available on request.
The two activities answer different questions. The probe tells you whether the process is still centred. Final inspection tells the customer what was shipped and gives a record that survives an audit.
How accurate is a spindle-mounted touch probe in real production?
Repeatability of a few microns is realistic on a well-maintained machine with a short stylus. Absolute accuracy is worse because machine positioning error and thermal drift add to it.
For our tightest work we hold ±0.005 mm (±0.0002 in) on the machine, but that figure comes from a controlled process with warm-up and a stable shop temperature, not from the probe alone.
Can the probe measure surface finish?
No. A touch probe returns coordinates, not roughness. Surface finish needs a separate instrument, and the values we quote are Ra 0.2–0.8 μm for fine finishes, Ra 0.8–1.6 μm for high-quality finishes and Ra 1.6–3.2 μm as-machined.
If a callout specifies both a tolerance and a finish, plan two measurement steps on two instruments.
What does it cost in cycle time?
A datum and work-offset routine is usually well under a minute. A full first-article routine that checks several features can run several minutes.
The payback comes from avoided rework. Re-cutting one feature in the same setup costs far less than unclamping, re-fixturing and re-zeroing the part.
Does probing work on five-axis parts?
Yes, and it is often more valuable there. Complex angles make manual setup slow and error-prone. We run 16 simultaneous 5-axis machining centers with a Ø400 mm rotary table, and probing the datum in the rotary frame removes a large part of the setup uncertainty.
Stylus reach is the constraint. Check that the probe can reach each critical feature at the required angle before you commit to the cycle.
How do we handle data and confidentiality?
Probe data is stored with the part program and the inspection record for that run. Uploads are secure and confidential, and an NDA is available on request.
If your drawing needs a formal dimensional report, we generate it from the inspection room instruments, not from the probe log.
Send us the drawing and the critical callouts
Tell us which features carry the tight tolerance and we will say whether on-machine probing, CMM inspection or both makes sense for your run. Quotation and free DFM analysis within 12 hours.
12-hour quote100% inspectionNo minimum order quantityNDA on request
More from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
