Monitoring a turning centre starts with counting correctly.
Bar-fed, twin-spindle and multi-part cycles all break naive counting. What delimits a cycle on each, the real loss profile, and OEE worked two ways.
A turning centre is the machine where cycle definition most often goes wrong, because so much turning work does not produce one part per cycle. Bar-fed machines run several components between bar changes; twin-spindle machines complete a part across two operations; parting off can complete a component mid-program.
Get that right and a lathe is one of the easiest machines to monitor accurately. Get it wrong and every number downstream — output, performance, cost per part — is wrong by a consistent multiplier that nobody notices for a month.
Machine signals, an energy meter and sensors, on one timeline.
Every CNC lathe can carry all three. Each part is recorded with the energy it took and the condition of the machine that made it.
What the machine is doing
- State: running, idle, setup, stopped
- Counts and cycle time, per part number
- Stops with reasons from the operator
- Alarms and fault stops
What it costs to run
- kW, kWh, kVA and power factor, live
- Energy per part, per shift and per job
- Idle energy: power drawn while producing nothing
- Maximum demand and load profile
How healthy it is
- Condition sensors connected where they matter
- Per-machine baselines, not generic limits
- Alerts before a trend becomes a breakdown
- Readings stored against every part
Sensors typically connected on a CNC lathe
An energy meter connection is available on every machine type we monitor. Sensors are chosen per machine at the pilot, and connected by the MachineWise team. See energy monitoring and condition monitoring.
| Machine configuration | What one cycle means | What to watch |
|---|---|---|
| Chuck work, single part | One program run, one component | Straightforward. The default assumption is usually correct. |
| Bar-fed, single spindle | One program run, one component, repeated until bar-end | Bar change is a legitimate stop, not a breakdown. Log it as its own reason. |
| Bar-fed, multiple parts per run | One program run, several components | Needs a parts-per-cycle multiplier set per part number, not per machine. |
| Twin spindle / sub-spindle | One component completed across two spindles | Counting both spindle events doubles output. Count completions, not operations. |
| Live tooling, mill-turn | One program run, one component, longer cycle | Longer cycles make micro-stop detection more important, not less. |
Bar loading and bar-end
Predictable, frequent and rarely logged. Measured separately it becomes a scheduling question rather than an invisible drag on availability.
Insert changes and offset corrections
Short, frequent and almost never written down. On high-volume turning these accumulate to more than the breakdowns.
Chip clearing and coolant
Especially on materials that produce stringy swarf. Often the largest single micro-stop category on a lathe.
Waiting for gauging
First-off inspection and in-process checks on a machine that has finished its cycle and is holding for approval.
On a networked CNC turning centre you get machine state, cycle timing, counts as the control reports them, and alarm presence. What you do not get, on any control, is whether a completed cycle produced one component or four — that is a property of the program and the fixture, and it has to be agreed during setup.
You also do not get a reason for a stop. Bar change, insert change and chip clearing all look identical to a control: the machine stopped. Separating them is what turns a lathe's availability figure into something a supervisor can act on, and it is why reason capture at the machine matters more on turning than on almost any other machine type.
OEE on a turning centre, both ways.
The standard definition first, then the variant MachineWise computes and why it differs. Quote the ISO figure externally; use the variant internally.
| ISO 22400-2 — the standard | MachineWise variant — internal | |
|---|---|---|
| Availability | Planned production time 435 min, actual production 350 min → 80.5% | Reporting time 435 min, machine on-time 384 min → 88.3% |
| Performance / Effectiveness | 480 parts × 42 s ideal cycle ÷ 350 min → 96.0% | Productive (in-cut) time 276 min ÷ 384 min on-time → 71.9% |
| Quality | 466 good of 480 → 97.1% | 466 good of 480 → 97.1% |
| Result | 75.0% | 61.6% |
The ISO figure needs a maintained ideal cycle time per part; the variant needs none, which is why it survives a floor running fifty part numbers a month. The two numbers answer different questions and are not interchangeable. The ISO figure is comparable to published benchmarks; the variant is not, and should be treated as your own baseline to improve against. Both are computed side by side in the OEE calculator.
A turning changeover is usually shorter than a machining-centre changeover and happens far more often. Chuck jaws, tool offsets, a program call and a first-off — twenty to forty minutes on a well-organised job shop, repeated several times a week per machine.
That frequency is the point. A changeover measured once looks minor; measured across a month it is frequently the largest single availability loss on a turning floor, and it responds to preparation and sequencing rather than to capital. Measuring it separately from breakdown is the whole value, and it requires the mode selector or a setup reason code to be mapped at commissioning.
What does one part cost in electricity?
With an energy meter on the CNC lathe, this is measured for every shift and job. Until then, estimate it here with your own numbers.
Values are pre-filled with a typical CNC lathe. Replace them with yours; nothing you type is stored or sent.
Idle energy is usually the fastest saving: it needs a switch-off rule, not capital. The meter shows it per machine, per shift.
Two turning centres on a live dashboard this week.
Bring your bar-fed machine and your most-changed machine. Thirty days, no cost, and the parts-per-cycle question settled properly on your own programs.
Book a free 2-machine pilot →WhatsApp us