LEDS-Level track. The same single-constraint topology, but with 36 named failure modes instead of five rolled-up Weibulls. The throughput answer barely moves. Everything you can do with it changes.
At a glance
| Experiment | Single Run |
| Duration | 90 days |
| System efficiency | 53.2% |
| Production | 6.89M bottles · ~68,900 pallets |
| Output unit | bottles |
What this model is
Identical in shape to One Constraint, Five Interrupts — one Constraint, the whole line, 100 bottles/min — and built from the same plant's data. The difference is that the events are kept drilled down to 36 named failure modes from a Line Event Data System rather than rolled up to five per-machine Weibulls.
These are fault names an operator would recognise: plow areas, glue stations, picker trees, magazines, micro stops.
The line
| Node | Name |
|---|---|
c0 |
Bottling Line |
The interrupts
36 named failure modes, distributed across the line:
| Machine | Failure modes |
|---|---|
| Capper | 12 — Micro Stop, Upper Plow Area, Capper, Compression, and 8 more |
| Case Packer | 2 — Photo Eye, Case Packer |
| Labeler | 3 — Labeler, Label Placement, Misalignment |
| Filler | 7 — Filler, Silo to Weigh Hopper, Mix to Bulk Bin, Bulk to Filler Bin, and 3 more |
| Palletizer | 12 — Hoist Area, Palletizer, Slat Sorter, Film_Break, and 8 more |
Each carries its own time-to-failure and time-to-repair distribution, fitted per mode. Click a node in the sandbox to see them.
Open the model file and you will count 37 entries, not 36. The extra is an unnamed Interrupt 0 — an unused placeholder that contributes nothing to the results. The same placeholder sits in the 1-Constraint model.
The same throughput, a completely different answer
The efficiency prediction lands within a fifth of a point of the rolled-up model. If the question is how much does this line make, the simpler model was already right.
But the rolled-up model's five interrupts were statistically identical, so its ranking was noise — a 1.2-point spread that reshuffles on a different seed. Here one failure mode carries about 20% of all downtime on its own, and the top five account for roughly 56%.
That is the difference between a model that can tell you the number and a model that can tell you what to do. The Weibulls on the Constraint-Level track are not invented: they are fits to this same data, aggregated to the machine level. Aggregation is what destroyed the ranking.
What the run produced
- Bottleneck operating availability: 53.2%
- System OEE: 53.2%
- Bottleneck total downtime: 1,010.9 hours across 27,636 failure events
- Mean time between failures: 2.5 min · Mean time to recover: 2.2 min
- Throughput: 6,894,812 bottles
Top contributors
| Rank | Interrupt | Downtime | Share |
|---|---|---|---|
| 1 | Capper_Upper Plow Area | 201.9 hrs | 20.0% |
| 2 | Labeler_Misalignment | 150.5 hrs | 14.9% |
| 3 | Filler_Micro Stop | 116.7 hrs | 11.5% |
| 4 | Capper_Picker Tree Assembly | 58.9 hrs | 5.8% |
| 5 | Filler | 42.7 hrs | 4.2% |
This is what a clear bottleneck looks like. The Constraint-Level track (same plant, same events, but fit at the machine level) showed a 1 pp spread across five rolled-up interrupts — by design, because aggregating into one Weibull per machine hides the drivers. The LEDS-Level drill-down puts Capper_Upper Plow Area at 20% front-and-centre; top five account for ~56% of total downtime. Same topology, same throughput envelope — but the drill-down tells you exactly where to invest.
Where to go next
Split the same LEDS data across five machines: Five Constraints — LEDS.
Open it
This model ships with ReliaSim® and runs in the browser sandbox — no install required. Open 1 Constraint in the sandbox.
Every recorded figure on this page is the published result for the model as it ships.