This article is for production managers, schedulers, production planners and methods and continuous improvement leads who want to make changeovers not only faster but also better planned. It explains why changeover time depends on the order of jobs, how to build a real changeover matrix from measured changeovers, and what planning learns from it.
- Changeover time depends on which product you change from and to.
- A changeover matrix holds the duration for each pair, ramp-up included.
- It should rest on measured changeovers, with median and spread.
- Sequencing then works with the reality of the line.
What Future Factory writes about changeover times
Fanny Dobrenova wrote Rüstzeiten minimieren oder optimieren? Finden Sie die goldene Mitte (“Minimise or optimise changeover times? Find the happy medium”) for Future Factory GmbH, an ERP software vendor for manufacturers based in Graz, Austria. She sets technical and organisational measures side by side, from SMED to planning the job sequence, and argues for a middle way between fast changeovers and clean results. On the role of data, she writes: “Die Analyse der im ERP-System verfügbaren Daten ermöglicht die effektive Planung von Rüstzeiten.” (Analysing the data available in the ERP system enables effective planning of changeover times.)
The article brings the planning view: group products with similar properties and run them one after another, move changeovers into periods of low load, sequence jobs sensibly. This article adds the line view: which changeover times does planning work with, and do they match what actually happens during the changeover? A sequence is only as good as the changeover times behind it.
Why one changeover time per machine is not enough
Many master data records hold one changeover time per machine or per item. In practice, the duration of a changeover depends on which product you change from and which product you change to.
- Changeover within a family. Often only parameters or stops are adjusted. The changeover is short and easy to plan.
- Change of material, colour or recipe. Cleaning, purging or quality release are added. The duration depends heavily on the direction, for example light to dark or the reverse.
- Format change with tooling. Tools, moulds or guides are swapped. This is where most of the changeover time sits.
- Ramp-up after the changeover. It takes more time before the machine runs stably at nominal speed and makes good parts. This time is missing from many planned values.
For example, on a packaging line, switching between two formats of the same family is quick because only settings change. Switching to a different film, on the other hand, means cleaning, adjusting and a trial run. If planning works with an average value, every sequence looks equally good on paper, and exactly the information that would make one sequence better is lost.
The real changeover matrix: from which product to which
A changeover matrix records, for each pair of preceding and following product, how long the changeover takes. To really help planning, it should rest on measured changeovers and hold more than an average.
| Field | Content | What it is for |
|---|---|---|
| From (preceding product) | Product or family before the changeover | Row of the matrix |
| To (following product) | Product or family after the changeover | Column of the matrix |
| Number of measured changeovers | How often this pair was observed | Shows how reliable the value is |
| Typical time (median) | Central measured value, robust to outliers | Planning value for sequencing |
| Spread | Gap between short and long changeovers | Planning buffer, sign of a missing standard |
| Ramp-up share | Time from restart to stable nominal speed | Shows whether the changeover or the ramp-up costs the time |
| Most frequent delay reasons | Missing tool, release, material, adjustment | Starting points for improvement |
The median says more than the average here: a single changeover where a tool had to be found should not distort the planning value. The spread still tells planning how much buffer it needs.
Measure changeover time so the matrix is right
We described how to measure a single changeover properly in Measure changeovers before you start SMED. For the matrix, these steps matter as well:
- Use the same start and end for every changeover. From the last good part of the old run to the first good part at nominal speed of the new run, ramp-up included. This is what makes pairs comparable.
- Record each changeover with its preceding and following product. Without that link you get an average per machine, not a matrix.
- Collect over several weeks. Frequent pairs quickly have enough measurements; rare pairs stay uncertain for longer and are flagged as such.
- Feed the matrix into planning. The measured values replace flat master data in the planning system and are updated regularly.
- Check the effect. After a sequence change, measurement shows whether total changeover time per week really goes down.
What planning learns from measured changeovers
Once the matrix is made of measured values, patterns appear that flat values hide.
- Some pairs are expensive, others almost neutral. That gives sensible families and campaigns, just as Future Factory recommends grouping similar products, but now with a measured rationale.
- Ramp-up is often bigger than expected. If much of the changeover time comes after the restart, the planned value is too short and the improvement lies in setting standards rather than tooling.
- High spread is a standard work issue. If the same pair varies widely between shifts, no sequence will help; a clear changeover standard will.
- After SMED, the matrix changes. Improved changeovers must flow back into planning, or it keeps working with old values. We show how to hold SMED results in Sustaining SMED gains with changeover tracking.
Future Factory points out that it is sometimes better to take a little longer over a changeover than to risk quality. Measured changeovers show where that happy medium lies: where a shorter changeover is eaten up again by a longer ramp-up or more scrap. Measuring OEE after each changeover makes this visible.
How TEEPTRAK provides measured changeover times
TEEPTRAK measures stops, short stops and changeovers automatically and in real time. A module is wired into the electrical cabinet and a sensor is mounted at the line, with no connection to the PLC. Installation takes about one hour per machine, and OEE is available within 48 hours. Every changeover is recorded with its start, end and ramp-up, and operators enter the product and reason at the line. That produces the measurements planners and methods engineers can use to check and maintain their changeover matrix. TEEPTRAK is used in more than 700 plants in over 30 countries, with an average gain of 10 to 15 OEE points.
Further reading
- Future Factory: Rüstzeiten minimieren oder optimieren? Finden Sie die goldene Mitte
- Measure changeovers before you start SMED
- Sustaining SMED gains with changeover tracking
FAQ
What is a changeover matrix? A table that gives the changeover duration for each pair of preceding and following product. It lets planning compare sequences by their real total changeover time.
Why should a changeover matrix be based on measured values? Because estimated or flat values rarely reflect the differences between pairs or the ramp-up. All sequences then look similar on paper, even though they are not on the line.
How often should the changeover matrix be updated? Regularly, and always after improvements such as SMED, new products or tooling changes. Otherwise planning works with values that no longer apply.
0 Comments