On a bottling or canning line, measure OEE at the filler, with an ideal rate for each format. Count blocked and starved time as filler stops, with the cause named at the machine responsible. To add formats, use good minutes (good containers / ideal per minute), since a shift of small bottles otherwise looks far better than a shift of large ones.
Why a beverage line needs its own rules
A filling line is a chain of machines at different speeds, with conveyors holding product between them. The filler can stop for ten minutes because a palletizer jammed, and the stop log at the filler says nothing useful unless you make it.
The wider food picture sits in OEE and downtime in food and beverage. This page stays on the line itself, from depalletizer to pallet.
Measure at the filler
Most lines are designed so the filler is the slowest machine, and the others are rated faster so they can catch up after a stop. That design is called the V-graph.
So the filler's OEE is the line's OEE, and the other machines explain why it moved.
Check that your line really matches this pattern. After a new labeller or a format with a slow pack pattern, the true limit can move downstream.
What counts as what on a filling line
| Event | Where it goes | Beverage note |
|---|---|---|
| CIP, sanitation, flushing | Stop | Often validated, so time the waiting around it. See sanitation downtime. |
| Format or flavor change | Stop | Split by type: flavor only, new bottle, new label. |
| Downstream full, filler blocked | Stop | Name the machine, as in blocked and starved. |
| No empty bottles, filler starved | Stop | Often the depalletizer or an air conveyor. |
| Short jams under your threshold | Performance | Too short to log one by one, but they add up. |
| Slowed to control foaming | Performance | A real loss, unless the rate was accredited for that product. |
| Underfill, cap or label reject | Quality | Count rejects after the filler, or they read as good. |
Set the ideal per format
A filler does not run a 0.5 liter bottle and a 1.5 liter bottle at the same rate. Each format needs its own ideal, proven on your line with good quality.
The nameplate is a start, but use what your line has held. How to find the ideal run rate shows the test.
Worked example: two formats in one shift
Example: a 480-minute shift runs 0.5 liter at an ideal of 600 bottles a minute, then 1.5 liter at 240 a minute. Stops were 65 minutes for CIP and format change, 32 blocked by the labeller and 18 for filler jams, so run time is 365 minutes.
| Format | Filled | Good | Ideal / min | Ideal min made | Good min |
|---|---|---|---|---|---|
| 0.5 L | 98,400 | 96,000 | 600 | 164 | 160 |
| 1.5 L | 29,280 | 28,800 | 240 | 122 | 120 |
| Total | 127,680 | 124,800 | 286 | 280 |
Availability is 365 / 480 = 76.0%, performance is 286 / 365 = 78.4% and quality is 280 / 286 = 97.9%. OEE is 280 / 480 = 58.3%.
The largest loss, 79 minutes, has no entry in the stop log. It only appears because the good-minutes check forces the numbers to add up, and how to reduce micro-stops is the next stop.
Notice the 32 blocked minutes as well. A team that looks only at filler faults would see 18 minutes and fix the wrong machine.
Where the data comes from
Filling lines are often well prepared for this. The Weihenstephan Standards define machine data points for packaging and bottling, including operating states and counters, with an OPC UA interface described in OPC 40600.
Check whether your machines support them before you buy sensors. Otherwise PLC signals for OEE or a retrofit sensor at the filler will give a run signal and a count.
Count containers both at the filler and after the last inspection. The gap between those counts is your quality loss, and a sudden change there usually means a sensor or a reject problem.
Beverage traps
- "Line efficiency" with exclusions. Many sites quote a figure that leaves out CIP or changeovers. Ask what it excludes before comparing it with OEE.
- Benchmarks from automatic systems. Hedman et al. (2016) found a median and average of 74% in 7 Swedish food and beverage companies, but most machines in that study's data reported a default 100% for speed and quality.
- Giving away product. Overfill does not show in OEE, since every overfilled bottle counts as good. Track it by weight or volume.
- Short campaigns. More format changes lower the score even on a well-run line. Compare like with like.
The world-class figure of 85% is usually traced to Nakajima, and it is no line target on its own. How to set OEE targets starts from your own baseline instead.
Short stops are the loss a filling line logs least. Fabrico takes stops, speed and counts live from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and where cameras are fitted it shows video of a stop, streamed on request next to its machine data. See how Fabrico handles short stops with cameras.
What people ask next
Where should OEE be measured on a bottling line?
At the filler, because the line is normally designed so the filler sets the pace. Other machines then explain the losses, as the V-graph guide shows.
Does CIP count against OEE?
Yes, in most setups a CIP inside scheduled production is a stop, even though it is required. Plants differ, so write the rule down, and see what counts in OEE for the same logic on planned work.
How do you calculate OEE with different bottle sizes?
Convert each format's good containers to minutes at its own ideal rate, add the minutes, then divide by scheduled minutes. OEE for mixed products has more examples.
What is a good OEE for a beverage line?
We found no neutral figure that settles this. One Swedish study found 74% in its food and beverage group with automatic data, but it carries a large caveat, so use the benchmark report and your own trend.
Should blocked time count as filler downtime?
Yes. It reduces filler output, so it counts against the filler's availability, but code it with the machine that caused it. See upstream and downstream stops.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.