The Right Data Sources for OEE Calculation and Collecting Downtime Reasons

Why does OEE exceed 100 percent or fail to match the shift report? Check planned time, the product’s ideal cycle time, total count and good count on the same data set.
Technical summary
Technology
OEE, ISO 22400-2, PLC counter and status signals, MES, historian
Topic
Signal definitions and downtime reason collection for a reliable OEE metric
Level
Intermediate
Prerequisites
  • Basic knowledge of PLC signals (counters, status bits)
  • Production planning terms (shift, work order)
In this article8 sections

After an OEE screen has been set up, a line showing performance above 100 percent, or a figure that does not match the shift report, shows there is a problem in how the inputs are defined. OEE (Overall Equipment Effectiveness) shows how much of what a machine could produce in the planned time, at ideal speed and without defects, it actually produces; for every value in the formula, it must be clear which signal and which written definition it comes from. Below, we describe which signal is taken from where, how the ideal cycle time is determined and how downtime reasons are collected without wearing out the operator.

OEE components and formula

OEE is the product of three ratios. ISO 22400-2 gives these ratios in its own terms (availability, effectiveness, quality ratio) and time definitions; their equivalents in common OEE practice are as follows:

Component Formula What it measures
Availability Run time / Planned production time Downtime losses
Performance (Ideal cycle time × Total count) / Run time Speed losses and short stops
Quality Good count / Total count Scrap and rework
OEE Availability × Performance × Quality Overall effectiveness

When the three ratios are multiplied and simplified, OEE = (Good count × Ideal cycle time) / Planned production time. This identity is the most useful cross-check in the field: even if the components are calculated wrongly individually, you check the total by an independent route.

The definition of every term in the formula should be written down across the plant. Are breaks deducted from planned production time? Which class do planned maintenance, trial production and waiting for material fall into? Which rule the calculation follows when a counter is reset, when the product or shift changes and when data is missing is also written in the same document. If these decisions are made differently per line, comparing OEE between lines becomes meaningless.

Signals required for each component and their sources

Reliable OEE is built with the signal set least dependent on operator input; these signals should be taken directly from the machine with IIoT-based production data collection. The minimum list:

Signal Typical source Where it’s used
Machine running state PLC status bit or status code Availability
Total count produced PLC counter (incrementing integer) Performance, quality
Scrap / reject count PLC counter, quality station, operator Quality
Active product or work order MES, PLC recipe number Selection of ideal cycle time
Operating mode PLC (automatic, manual, setup) Downtime classification
Alarm / fault code PLC alarm word Automatic downtime reason
Shift and plan information MES, ERP or shift calendar Planned production time

The most debate arises over the “machine running” signal. The motor running bit is also 1 while the machine is idling; availability calculated with this bit comes out higher than reality. A more robust way is to tie the condition “in automatic mode and in cycle” to a status code in the PLC program, or to use the time elapsed since the last counter increment. A single status code defined in the PLC (for example 1 = producing, 2 = waiting, 3 = fault, 4 = setup) simplifies both the calculation and downtime classification.

For counters there is one rule: take the signal as an incrementing counter, not as a pulse. A pulse signal is missed within the scan or sampling interval; taking the difference from an incrementing counter preserves the total even after a connection outage. Counter overflow (for example a 16-bit unsigned counter wrapping to 0 after 65535) and resetting when the PLC restarts must be handled on the calculation side. Some PLC counter blocks stop at the upper limit instead of overflowing (for example the IEC counter CTU of type INT staying at 32767). A team that doesn’t notice this looks at a counter that has frozen during long shifts and thinks production has stopped. In that case, manage the counter yourself in the PLC program or use a counter of type DINT/UDINT:

# Production difference from two consecutive counter readings
def counter_delta(previous, current, max_value=65535):
    if current >= previous:
        return current - previous
    # The value dropped: overflow or reset?
    wrap_delta = (max_value - previous) + current + 1
    if wrap_delta < 1000:       # reasonable upper limit: set per line
        return wrap_delta
    return current              # reset assumed; production equals the new value

The choice of platform on which these signals will be collected and stored is covered in the article choosing a historian.

Determining the ideal cycle time

The whole performance rate depends on the ideal cycle time. The value is per product; different products on the same line have different ideal times. Three sources are used to determine it:

  1. Machine design speed: The nominal capacity given by the manufacturer. Usually optimistic, but correct as a reference.
  2. Best sustained speed observed: The cycle time of the fastest uninterrupted periods without quality problems in historical data.
  3. Bottleneck station time: On multi-station lines, the slowest station determines the line speed.

If OEE data is to be combined with company-wide indicators, define data owners and refresh frequency for corporate dashboards and management reporting.

Don’t use the “standard time” set as a corporate target as the ideal cycle time. The standard time carries losses within it and makes performance look higher than it is. The ideal cycle time should be kept in one place in the product master data and selected automatically when the active product changes. If product changeover depends on manual entry by the operator, shifts calculated with the wrong product are a common error in the data.

The downtime reason tree and operator input

OEE is useful to the extent that it shows which loss comes from where. For that, downtime must be given reasons. A working reason tree has two or three levels and is kept short:

  • Level 1 – Loss type: Planned downtime, breakdown, setup and changeover, waiting for material, staffing, waiting for quality.
  • Level 2 – Area: Mechanical, electrical, infeed, packaging, external supply.
  • Level 3 – Specific cause: Only for recurring causes that can be acted on.

Rules that work in the field:

  • If there is a PLC alarm code, the downtime reason is assigned automatically; the operator only confirms or corrects it.
  • Stops below a defined threshold (for example 2 minutes) are automatically put into the “short stop” class; no reason is requested.
  • The shift cannot be closed until stops above the threshold have been given a reason; the proportion of the “other” option is monitored separately in reports, and if it is high, the tree is reviewed.
  • The input screen should be next to the machine, touch-operated and usable with one hand. Bulk entry from memory at the end of the shift produces faulty data.

Example OEE calculation

The values below are for an example calculation:

  • Shift length 480 minutes, planned breaks 30 minutes → planned production time 450 minutes
  • Total unplanned downtime 45 minutes → run time 405 minutes
  • Ideal cycle time 6 seconds/unit, total production 3,600 units
  • Scrap 90 units → good count 3,510
Component Calculation Result
Availability 405 / 450 0,900
Performance (6 s × 3.600) / (405 × 60 s) = 21.600 / 24.300 0,889
Quality 3.510 / 3.600 0,975
OEE 0,900 × 0,889 × 0,975 0,780

Cross-check: 3,510 units × 6 s = 21,060 s = 351 minutes; 351 / 450 = 0.780. If both routes give the same result, the signals and times are consistent. In the example, the performance loss (equivalent to about 45 minutes) is the same size as the unplanned downtime. A team that only looks at long stops misses half the loss.

Data errors that mislead OEE

  • Clock mismatch: If the PLC, collector and server clocks differ, production at the shift boundary is written to the wrong shift. The source of timestamps and clock synchronisation are covered in detail in the article timestamps and data quality in production data.
  • Carrying the last value over during a connection outage: When no data arrives, the last state remaining “producing” hides downtime. Intervals with bad quality information must be flagged separately.
  • Scrap counted twice: If both the quality station and the operator enter the same scrap, the quality rate comes out low. Define a single source.
  • Reworked parts: Whether parts that were not good at first pass but were later recovered count as good must be defined; first-pass quality can be monitored as a separate metric.
  • Manually correcting planned production time: Retrospectively changing the unplanned downtime of bad shifts to “planned” is the practice that destroys trust in the metric fastest. Every change must be written to an audit trail.

During commissioning, the first thing to look at is the comparison of a shift’s raw counter difference with the total count on screen. If these two numbers do not match, there is no point discussing the rest of OEE.

Implementation checklist

  • An OEE definition document has been written: the rules for planned production time, planned maintenance, trial production and rework are clear.
  • A single machine status code has been defined in the PLC and carries the same meaning on all lines.
  • Total and scrap counts are read as incrementing counters; overflow and reset are handled in the calculation.
  • The ideal cycle time is kept in the product master data and selected automatically according to the active product.
  • The reason tree has at most three levels; a short stop threshold has been set.
  • The clocks of PLCs, gateways and servers in the OT infrastructure are synchronised to the same time source; data is stored in UTC.
  • OEE is calculated both from its components and from the simplified formula, and the two are compared.

On the side of collecting signals and getting work order information from the MES, the article MES–ERP integration with ISA-95 provides the complementary framework. For designing and commissioning OEE screens, see our production monitoring and dashboards service.

Frequently asked questions

Does planned maintenance count as downtime in OEE?

This is a definition decision and must be applied the same way across the plant. If you deduct planned maintenance from planned production time, availability is not affected; if you count it as unplanned downtime, availability falls. Both are defensible. What matters is that the choice is written down and the same rule is used when comparing lines.

Why does the performance rate exceed 100 percent?

A common cause is that the ideal cycle time is defined as unrealistically long, or that the cycle time of the wrong product is used. Another possible cause is the counter signal counting one part twice. Verify the ideal cycle time from the machine’s design speed and the best sustained speed observed, and check that the correct value is selected at product changeover. Do not clip a value above 100 percent to 100 and report it; show the value as it is and flag it until the cause is found.

Should short stops be given reasons individually?

No. Asking the operator for a reason for every stop lasting a few seconds lowers data quality. Stops below a certain threshold are automatically put into the short stop class and monitored as performance loss. For stops above the threshold, entering a reason is mandatory; the threshold is set according to the line’s cycle time.

Is an MES required for OEE?

It is not required. OEE can be calculated by collecting status, counter and scrap signals together with shift and product information in a historian or a simple database. When work orders, products and operator input need to be managed together, however, an MES or OEE application reduces the maintenance burden.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Industrial

PLC, PROFINET/PROFIBUS and SCADA communication, IT/OT network segmentation and industrial field connectivity. We install and monitor without stopping the production line.

Knowledge Center

Other articles