OEE Monitoring System: What to Measure on Shop-Floor Machines

An OEE monitoring system needs three groups of signals from each machine: a reliable running or stopped state (for availability), a part or cycle count against an agreed ideal cycle time (for performance), and a good or reject count (for quality).

OEE is availability × performance × quality.

On mixed-vintage shop floors the hard part is not the formula, it is getting a trustworthy run signal and part count from machines that were never built to report them.

A logging layer or edge gateway can collect those signals and send them to a dashboard or control room; it does not replace your MES, SCADA or PLC.

Who this is for: production, maintenance and automation engineers who have been asked to “show OEE” for a line of presses, CNCs, moulding machines or packing machines of different ages and makes.

Why this question keeps coming up

A recent enquiry from an instrumentation firm asked a common thing: their client wanted OEE on a set of older machines and wanted to know what had to be wired before any software discussion. That is the right order. Decide the signals first; the dashboard comes later.

OEE in one formula

OEE is defined as a key performance indicator in ISO 22400-2, the standard for KPIs in manufacturing operations management (ISO 22400-2:2014). In plain terms:

• Availability = actual run time ÷ planned production time.
• Performance = (ideal cycle time × total count) ÷ run time.
• Quality = good count ÷ total count.
• OEE = availability × performance × quality.

Worked example (illustrative numbers only): a shift has 480 min planned time. The machine runs 420 min, so availability = 0.875. Ideal cycle time is 30 s and it makes 760 parts, so performance = (0.5 min × 760) ÷ 420 = 0.905. 740 parts are good, so quality = 0.974. OEE = 0.875 × 0.905 × 0.974 ≈ 0.77.

Each factor depends on a different physical signal. If one signal is wrong, the OEE number is wrong in a way the formula will not show you. If you want the reasoning for measuring this before buying an MES, read our post on an OEE baseline before MES.

Availability: get the run signal right

Availability needs two things: when the machine is running and what counts as planned time.

Run signal options, roughly in order of preference:

1. A PLC or CNC status bit (cycle active, auto mode, spindle on) read over the machine’s own interface, if one exists. Check the controller vendor’s documentation for which bit means “producing” and not just “powered”.
2. A spare contact or output from the machine’s control circuit, such as a motor contactor auxiliary contact or a cycle-start relay, wired to a digital input.
3. Motor current via a current transformer and a threshold. This works on machines with no spare contacts, but you must pick a threshold that separates idle from cutting or pressing. Idling hydraulic pumps can draw a lot of current while nothing is produced.
4. Spindle or shaft speed from a proximity sensor on a rotating part.

Planned time is a business rule, not a sensor. Agree shift times, breaks and planned maintenance with production before you start, or availability will be argued over forever.

Downtime reasons (no material, tool change, breakdown, waiting for QC) usually come from an operator entry: a button panel, an HMI screen or a tablet. Keep the list short; ten reasons that people actually press beat forty that nobody uses.

Performance: count cycles, not guesses

Performance needs a total count and an ideal cycle time for each product or die.

• Count source: a stroke or cycle pulse from the controller, a proximity sensor on a moving part, or a photoelectric sensor at the outfeed. For high speed counts, check that the input’s maximum count frequency covers your fastest cycle. Inductive proximity switches are covered in IEC 60947-5-2:2019.
• Multi-cavity or multi-up tools: one stroke can make several parts. Store parts per cycle with the job, not as a fixed machine setting.
• Ideal cycle time: take it from the process engineer’s standard, not from last month’s average. If you use an average, poor performance becomes the baseline.

Micro-stops (stops of a few seconds) are often where performance loss hides.

A count signal logged with a timestamp lets you see them later without any extra sensor.

Quality: the signal most plants skip

Quality = good ÷ total. Sources:

• Automatic reject: a reject gate, a checkweigher or vision system output, wired as a second count.
• Operator entry: rejects entered per shift or per lot on an HMI or tablet.
• Rework: decide up front whether reworked parts count as good. ISO 22400-2 treats quality ratios in terms of good quantity, so write the rule down.

If you only have end-of-shift reject entries, quality will be daily rather than live. That is fine for a first baseline.

Download the Free “OEE Monitoring Signal Checklist”

Enter your email and the form emails you the copy link to the Google Sheet.

Signal checklist for mixed-vintage machines

Use this per machine. It doubles as the downloadable checklist (lead magnet: oee-monitoring-signal-checklist).

Points to check on site:

• Is there a spare contact, or do you need an interposing relay so you never load the machine’s control circuit?
• What voltage is the signal: 24 V DC, 110 V AC, 230 V AC? Use the right input or an isolating relay.
• Can the machine’s own controller be read without affecting its program? Ask the OEM before connecting anything to a running controller.
• Who owns the clock? Every count and stop should carry one consistent timestamp. NTP is defined in RFC 5905 (IETF).

Where an edge gateway fits, and where it does not

An edge gateway or logger sits between the machine signals and the dashboard. Its jobs are narrow:

• read digital inputs, counters and analog signals, or poll a controller over Modbus or similar;
• timestamp and buffer data if the network drops;
• send data to a server, cloud or control room screen.

It does not run your machine, replace the PLC, or replace an MES that handles work orders, genealogy and scheduling. For retrofit wiring and protocol options on older machines, see retrofitting legacy machines with OPC UA and Modbus. For Modbus itself, the protocol specifications are published by the Modbus Organization.

Embedos Edge is our platform of products, solutions, dashboards, insights and services. On an OEE project it captures machine and sensor data and sends it to the cloud or a control room, with homepage widgets and smart alerts for stoppages. EmView monitoring screens can show the same data in a control room (sign in at emcommandview.embedos.io; )

It sits alongside your PLC, SCADA and MES; it is not a replacement for any of them. Any OEE figure it shows is only as good as the run, count and reject signals you wire into it.

Common mistakes

1. Counting from the wrong point. A sensor before a reject gate counts total; one after it counts good. Know which you have.
2. Mixing ideal and average cycle times across products.
3. No agreed planned time, so availability changes depending on who reads the report.
4. Trusting a single number. Always look at A, P and Q separately. An OEE of 0.77 tells you little; availability of 0.875 with performance of 0.905 tells you where to look.
5. Starting with too many machines. Wire two or three, check the counts against the shift register for a week, then roll out.

Validate before you report

For the first week, compare logged counts with the operator’s shift register and logged stops with maintenance records. Differences point to a bad signal choice, not bad operators. Fix the signal, then publish numbers.