Packaging Line OEE Monitoring in South Africa

Define the output the line is responsible for

A filler, label applicator, case packer and palletiser can all count the same product at different stages. Adding their counts together creates an impressive number with little operational meaning. The project first identifies the unit the line delivers and where it becomes a completed unit. That may be a filled bottle, an accepted pouch, a sealed carton or a finished roll. It must match the production team's reporting responsibility.

The line layout then determines which additional observations help explain losses. A counter at the constraint measures its work; a downstream counter can reveal accumulation, rejection or a transfer problem. Those counts remain separate observations with named locations. We agree how work in progress at the start and end of a shift is treated, so a temporary build-up between machines does not appear as missing finished stock.

Select a first line that can produce a trustworthy reference

A useful starting line has a clear output unit, a recurring operating problem and people able to witness representative production. Include the production supervisor, maintenance or automation contact and the person who owns the shift report. If the project concerns changeovers, select a line with an upcoming product change. If it concerns short stops, include the operating conditions under which those stops normally occur.

Send a simple layout, the machine and controller details, a short SKU list, current reporting sheets and available count or status signals. We use those inputs to propose a measurement boundary and identify remaining survey questions. A film converter reporting metres and rolls needs a different counting arrangement from a cartoner reporting individual cartons. The equipment selection follows that distinction rather than imposing one standard counter kit on every packaging process.

Capture counts without changing their meaning

Existing PLC or controller outputs are assessed for approved read access. Where the required signal is unavailable, a suitable external counter or sensor can be included in the scope. The survey considers mounting, product spacing, reflective or transparent materials, lanes, conveyor behaviour and maintenance access. In washdown areas, equipment and installation details must match the actual cleaning conditions reviewed with the site.

Count acceptance should cover the difficult examples as well as a clean run: closely spaced products, a stopped product under the sensor, reverse movement, empty conveyor movement and rejects diverted before the final count point. Sensor filtering is tuned against those observations. A filter that removes electrical noise but also merges two legitimate products would distort the record. Retain the original configuration and agreed count test so later adjustments can be checked.

Give each SKU an approved speed reference

OEE performance compares production against an ideal cycle time or rate. That reference must use the same unit as the count. A standard in bottles per minute cannot be applied directly to a carton count without a defined conversion. Different pack sizes, formats and machine configurations may need different approved references. Store the applicable version with the job record so an older shift can still be reproduced after a standard is changed.

Discuss why the chosen reference is appropriate with production and process engineering. A historical average can absorb recurring losses into the baseline, while a catalogue maximum may not describe the configured line and product. The reporting scope should distinguish the operating target used by the supervisor from the reference used in OEE. Reaching a planning target and reaching the ideal production rate answer different management questions.

Make short stops visible without counting the same loss twice

A stopped conveyor, a machine fault and the absence of a count pulse are different signals. The configuration defines which evidence establishes a stop and how a brief interruption is classified. Agree the duration boundary between a separately recorded downtime event and a small stop represented within performance loss. Keep that boundary consistent in the report so the same interruption is not deducted from both availability and performance.

Edge collection may need to record events much faster than the dashboard refreshes. Where LoRaWAN or 4G carries the monitoring data, the design specifies what is counted locally, what is transmitted and how a sequence is retained during a communications interruption. The acceptance test follows actual events through this path. A frequent screen refresh cannot compensate for a counter or edge device that failed to capture the event at source.

Separate the symptom from the reason for a stop

The constraint may stop because it is starved, blocked, being cleaned or waiting for an operator. An automatic state record can establish timing; the explanation may need additional signals or an operator entry. We agree a short, practical reason list and who can correct it. Excessive detail is difficult to maintain during a busy shift, while a single miscellaneous category gives the improvement team little to work with.

Keep the observed event and the later explanation distinguishable. If a supervisor changes a reason after investigating a jam, the event duration should remain traceable and the correction should carry a time and user reference. The shift review can then show unresolved reasons as a work queue. It should not hide missing explanations by automatically assigning every stop to the nearest available fault code.

Measure changeover as an operating sequence

Define the beginning and end of a changeover in terms the team can witness. A useful sequence may include the last accepted unit of the previous product, cleaning or adjustment, the first trial output and approval of the first acceptable new product. These timestamps distinguish mechanical work from material waiting and quality release. The relevant milestones depend on the line and the customer's own production procedures.

Record the old and new SKU, crew context and agreed reason codes. When comparing changeovers, compare relevant product transitions rather than treating every format change as the same task. A quick size adjustment and a full material or cleaning change create different work. The report should help the production team choose an improvement action, such as preparing materials earlier or reviewing a recurring adjustment, using the recorded sequence as evidence.

Connect quality records to the counting period

A reject counter, operator tally and quality-release system may produce information at different times. Define the source of total count and good count, then specify how a delayed quality decision affects the report. An unresolved batch should remain visibly unresolved until the agreed decision is available. If rework returns through the line, record the pass and the treatment of its output so it is not silently counted twice as new production.

For an illustrative six-bottle carton, a downstream count of 1,950 full cartons represents 11,700 bottles at that boundary. An upstream count of 12,000 bottles does not by itself prove that 300 were scrapped: some may still be between the counters, diverted for review or counted in a different interval. This is a hypothetical reconciliation example. The actual installation defines the points, time window and exceptions needed to explain a difference.

Accept the line record before using it to rank shifts

Use the OEE pilot acceptance checklist to record expected results, actual results, defects and sign-off. Tests are planned with the site and use safe simulation where a live interruption would affect production. The acceptance decision should cover the user's report and the event history, not just whether a value appears on the screen. Initial monitoring establishes a usable baseline; subsequent production changes determine the improvement achieved.

Understand the installation and subscription scope

The quotation identifies count and status points, retained equipment, additional sensors, mounting and electrical work, edge collection, connectivity, platform configuration and commissioning. It also identifies who supplies SKU information, maintains reason codes and approves report definitions. Interfaces to an ERP, quality or scheduling system are priced against the actual available export or API and the required mapping.

Separate initial engineering and installation from recurring platform, mobile data, support and maintenance costs. Expansion to another line may reuse reporting conventions while still requiring different sensors and count logic. Include access windows, cleaning arrangements, spares and the process for changing a counter or machine. The broader OEE management service provides the plant-level framework; this packaging scope makes one line's evidence usable within it.

Frequently asked questions

Method reference

Vorne: calculating OEE explains the availability, performance and quality relationships used here. Reviewed 5 November 2026. Product conversions, stop rules, reference speeds and acceptance tolerances are specified for the actual line; the numerical example above is illustrative.