CNC machine monitoring is a system that automatically collects machine state, spindle load, active program, alarm codes, part counts and cycle times from the controller or external sensors and shows them on one dashboard. It measures the time a machine actually spends cutting, not the time it is switched on, and shows where the rest goes and why.
Key takeaways
- Most CNC monitoring data (state, spindle load, program, alarms, part counts, tool life) is already available in the controller.
- Networked machines are read through MTConnect, OPC UA or vendor interfaces such as FANUC FOCAS; older machines without a port are retrofitted with sensors.
- Utilisation figures only mean something once every machine shares one state model: cutting, waiting, setup, alarm, planned stop.
- According to Siemens' 2024 report, large plants average 25 unplanned downtime incidents and lose 27 production hours a month.
- The soundest start is a pilot on two or three bottleneck machines with a success criterion written down in advance.
Why is lost time on CNC machines invisible?
The machine tool is the most expensive asset in a CNC (computer numerical control) shop, yet in many businesses what it did all day is still learned from a paper sheet filled in at the end of the shift. The sheet says "ran 8 hours, 140 parts". Hidden inside those hours are program loading, work offset setting, first-off inspection, minutes spent waiting for a tool and a feed hold nobody cleared.
The record-keeping problem grows with complications specific to CNC work. The same machine may run five different jobs in a day, each with its own program, tooling and target cycle time. If an operator turns the feed override down to 70% to rescue a difficult part, the machine still looks busy but produces more slowly than planned. That lost speed never shows up in a manual log.
We covered the general case of getting data off older equipment in our guide to retrofitting legacy machines. This article deals with what is specific to CNC: what the controller can tell you, which states you need to separate and how to turn raw signals into numbers a production manager will act on.
What does not monitoring CNC machines cost?
Unplanned downtime is expensive in every sector, and what is not measured cannot be managed.
In its 2024 report, Siemens estimates that the world's 500 largest companies lose almost $1.4 trillion a year to unplanned downtime, equivalent to 11% of their revenues. (Siemens – The True Cost of Downtime 2024)
The same Siemens report found that large plants average 25 downtime incidents a month (down from 42 in 2019) and lose 27 hours a month to unplanned stoppages (down from 39). These figures come from large plants, but the same logic applies in a CNC shop: short, frequent stops usually consume more time over a month than a single big breakdown.
The payback from data-driven maintenance has also been measured. Independent surveys cited in the US Department of Energy's (DOE) Operations and Maintenance (O&M) Best Practices Guide, Release 3.0 from 2010, report that, as an industry average, a functioning predictive maintenance programme cuts downtime by 35–45% and maintenance costs by 25–30%. These are not results every plant will repeat, but the precondition for any such gain is a monitoring layer that captures spindle and axis data consistently.
What data should CNC machine monitoring collect?
The advantage of CNC equipment is that most of the useful monitoring data already lives in the controller. The table below shows which data point answers which question, and where it typically comes from.
| Data point | Question it answers | Typical source |
|---|---|---|
| Mode and execution state (auto, feed hold, alarm, manual) | What is the machine doing right now? | Controller |
| Spindle speed and load | Is it really cutting? Is the tool struggling? | Controller or current sensor |
| Feed and spindle override | Is the program running at its intended speed? | Controller |
| Active program name and part counter | Which job, how many parts, what cycle time? | Controller |
| Alarm number and message | Why did it stop, and which alarms keep recurring? | Controller |
| Tool number and tool life counter | When should the tool be changed? Is breakage likely? | Controller |
| Vibration and temperature | How healthy are the spindle bearings? | External sensor |
| Stop reason (material, program, inspection, waiting) | What context can no sensor see? | Operator terminal |
The stop reason row matters most. The controller knows the machine has stopped, but not whether the operator is waiting for material or for a quality sign-off. A simple screen where the operator picks a reason with one tap closes that gap, the same principle behind an andon system.
How do you get data out of a CNC controller?
Most modern CNC controllers can be networked and expose data through either a standard or a vendor-specific interface. The two most common routes are MTConnect, an open standard designed specifically for machine tools, and OPC UA, a general-purpose industrial standard. According to the OPC Foundation, OPC UA (Open Platform Communications Unified Architecture) was released in 2008; our article OPC UA explained walks through how it works.
On some controllers these interfaces are a paid software option; on others the maker's own communication library is used instead (FANUC's FOCAS, short for FANUC Open CNC API Specifications, is a well-known example). Older machines with no network port are retrofitted with a current clamp, a stack-light signal and a part counter. A machine-by-machine inventory settles which route fits where.
How do you interpret CNC data? State model and KPIs
Raw machine data says little on its own. The first job is to define one state model for every machine: cutting, in cycle but waiting, setup, in alarm, planned stop and unscheduled. Without this, any utilisation figure blurs "the machine was on" with "the machine made parts".
Once the state model is agreed, these indicators start to mean something:
- Spindle utilisation: the share of planned time during which the spindle was turning under load.
- OEE (overall equipment effectiveness) and its components: availability, performance and quality; the calculation is covered in how to calculate OEE.
- Setup time: the gap between the last good part of one job and the first approved part of the next. In high-mix, low-volume shops, such as made-to-order furniture manufacturing, this is often the single biggest loss.
- Downtime Pareto: stop time ranked by reason, showing which problem to fix first.
- Override deviation: which jobs routinely run below 100% feed; usually a sign of tooling, workholding or programming issues.
The principles for getting CNC indicators onto a dashboard managers actually open are collected in our piece on BI dashboards and KPIs.
How do you set up CNC machine monitoring? Six steps
- Build a machine inventory. List each machine's make, controller model, software version, network port and available interface options. This list decides which method you use on which machine.
- Pick a question and a pilot. Start with something concrete, such as "why is setup on the bottleneck machining centre so long?", and run a pilot on two or three machines. Write the success criterion down before you start. To weigh the pilot against plant-wide priorities, see our Industry 4.0 roadmap for SMEs.
- Define states and stop reasons. State definitions must be identical across machines; keep the reason list short and in words operators actually use.
- Plan the network and its security. Machines belong on a segment separated from the office network, with read-only access to the controller. See our OT security guide for the wider framework.
- Link the data to work orders. When the active program is matched to a work order, you can see how long each order took on each machine. The logic is explained in work order tracking.
- Build it into routines, then scale. If the data is not used in the morning meeting and at shift handover, the project fades within months. Once the pilot hits its target, roll the same model out to the other machines.
A common mistake when setting up CNC monitoring is trying to collect everything from every machine. Spindle load sampled every second is valuable for predictive maintenance, but in the first month management needs state, counts and stop reasons. Where high-frequency data should be processed is the subject of our article on edge computing in manufacturing.
How does CNC data feed predictive maintenance?
Once the CNC monitoring layer is in place, the same data starts serving maintenance. Spindle load creeping upwards on the same program over weeks is an early sign of tool wear or a workholding problem. The frequency of recurring alarm numbers points to the axis or lubrication circuit that needs attention.
For bearing and spindle health, controller data is often not enough; that is where an external vibration sensor comes in. The method is covered in vibration analysis for fault detection, and the wider strategy in our predictive maintenance guide.
To trace recurring stoppages back to their causes, our article on root cause analysis with production data is a natural next read. There, monitoring data is lined up on the same timeline as quality and maintenance records.
How does Digital Bridge deliver CNC monitoring projects?
A machine shop rarely runs a single brand. A new machining centre, a fifteen-year-old lathe and a mill with no network port often sit side by side. Because our hardware and software come from the same team, we bring that mixed fleet onto one dashboard.
- Free site assessment. As the first step in Industry 4.0 projects, we walk the shop floor with you and map the machine inventory, controller interfaces and priority questions.
- Connection method per machine. Machines with an interface are read directly; those without are retrofitted with external sensors. Bringing different protocols into one data model is the core of our system integration work.
- One dashboard, live status. Data flows into our remote monitoring platform, where machine status, threshold alarms, historical data and trends sit in one place. Companies with several sites see every machine on the same screen.
- Work orders and ERP. State, counts and stop data are linked to work orders through MES production management (manufacturing execution system). Our MES integrates with SAP, with Logo and Mikro (two ERP packages widely used in Türkiye) and with custom ERP systems.
- Maintenance pilot. For predictive maintenance projects we start with one to three critical machines and demonstrate value with metrics within the first 30–60 days.
Every project starts with a needs analysis, followed by a written proposal setting out scope, phases and cost. We do not sell off-the-shelf packages, because no two shops have the same machine fleet or job mix.
Where should you start with CNC machine monitoring?
You do not need to connect the whole shop to start CNC machine monitoring. One bottleneck machine and one clear question are enough. Share your machine list and the point where you think you lose the most time, and we will use a free site assessment to decide which machine to start with and how. You can reach us through our contact page.
To see how machine data fits with the other building blocks of a connected factory, browse all our Industry 4.0 articles.