In a chemical plant, OEE = (on-spec tons / best proven rate) / scheduled hours, measured on the unit that limits output. Trips and grade changes are stops, a turnaround planned well ahead is excluded, and off-spec is a quality loss even if you can recycle it. Never set an ideal rate outside the unit's safe operating limits, and track yield and energy separately.
Chemical plants run on rules a packaging line never meets
A chemical unit may run for months, then come down for weeks. Its rate can be limited by heat transfer, feed quality or a permit, and the value of a ton depends on how little feed and energy it took to make.
OEE can handle the first part well and says nothing about the second. This page covers what to count on a chemical unit, and where to stop relying on it.
For the method on one stage, read continuous processes or batch processes. To choose where to measure across a site, read OEE for process manufacturing.
Trips, turnarounds and transitions
| Event | How to count it | Note |
|---|---|---|
| Interlock trip, emergency shutdown | Stop | Restart time until on-spec can far exceed the trip itself, so write one rule for it. |
| Fouling cleanout, exchanger bypass work | Stop | See heat exchanger troubleshooting. |
| Running slow as fouling builds | Performance | A loss that creeps up over weeks, so trend it. |
| Grade transition | Stop, plus off-spec as quality | Sequence matters. See counting changeovers. |
| Steam, cooling water or feed lost | Stop | Caused outside the unit, so name the source. |
| Catalyst change as a routine step | Stop, or excluded if planned well ahead | Pick one rule per unit. |
| Turnaround planned well ahead | Excluded from scheduled time | It still counts in TEEP. |
| Unit idled: no demand | Excluded if planned | See OEE with no orders. |
Be strict about the word "planned". A shutdown that was moved up two days because a pump failed is a stop, whatever the calendar now says.
Setting the rate
Use the best rate the unit has proven with on-spec product, per product family, as in how to find the ideal run rate. Nameplate is often out of date after debottlenecking, and it can be too high or too low.
The ideal has a safety boundary
A metric must never push a unit past its limits. If raising the rate means changing technology, equipment or procedures, treat it as an engineering change and not an OEE target.
In the US, OSHA's process safety rule, 29 CFR 1910.119, requires written management of change for such changes on covered processes, except replacements in kind. Its process safety information must also cover safe upper and lower limits for items such as flows, so check your own site's rules.
So when a rate limit is a safety or permit limit, cap the ideal there. Time at that limit is then full speed, not a speed loss.
On-spec tons and the grey areas
Good means on-spec at the first attempt, confirmed by the lab. Chemical plants have grey areas that make this harder, so settle them in writing.
- Recycled off-spec. It goes back as feed and may lose little value, but it wasn't good the first time. Count it against quality, and don't count it again as new output in a way that hides the loss.
- Downgraded product. Sold as a lower grade, it is not good output for the grade you planned. Report the tons and the price gap.
- Transition material. Product made during a grade change is off-spec unless the lab says otherwise.
- Lab delay. Waiting for a result is a stop if the unit holds for it. If the unit keeps running, it only delays the count.
Worked example: a solvent column in June
Example: a distillation train with an ideal rate of 12 tons an hour. June has 720 hours, and a 120-hour planned turnaround is excluded, so scheduled time is 600 hours.
Stops were 14 hours of trips, 18 of exchanger cleaning and 16 of grade transitions, 48 hours in all, so run time is 552 hours. The train averaged 10.8 tons an hour, which is 5,961.6 tons, and 180 tons were off-spec.
A shift log would list the 48 hours of stops, and the team would work on trips. Yet running 10% under the proven rate cost 55.2 hours, which is more.
The next question is why the rate was low, and causes of reduced speed gives a checklist. If fouling or feed caused it, it is a loss to fix, and if a tank farm limit caused it, the constraint sits elsewhere.
What the same June can hide
Suppose yield on feed slipped from 96% to 94% and steam per ton rose by a tenth, while OEE stayed at 80.3%.
- Yield and selectivity. Good tons are counted the same however much feed they took. Track yield next to OEE, using production yield.
- Energy. Pushing a column harder can lift OEE and lift energy per ton as well. Report both, with cost per unit as the referee.
- Demand-set rates. Where the plant follows what the market will take, low performance may be a decision. Report a planned rate cut as its own no-demand loss, or the number loses its meaning.
- Safety and environment. A unit should stop when it must. Never let a high score make someone hesitate over a trip, and keep process safety metrics apart from it.
We found no neutral published OEE benchmark for chemical units. Bars by process type exist, and you can read them in the process manufacturing guide, but your unit's own trend is a better guide.
How to get the data
The control system and historian usually have all of it already. You need flow or weight totals, a run-state rule, lab results and a log that says why.
Define "running" with a real tag, such as feed flow above a set value. Then reconcile flow-meter tons with shipped and tank-level tons, because a drifting meter moves performance without anything changing on the unit.
Ways a chemical OEE goes wrong
- Counting a restart as run time. Time to get back on-spec is part of the stop's cost. Decide whether it is a stop or off-spec production, and keep it the same.
- Moving unplanned work into the turnaround. It flatters availability this month and hides a failure.
- Using one ideal across grades. A hard grade then looks like poor performance.
- Averaging units. Weight by scheduled hours, and look at the constraint first.
Chemical units often have the signals but no shared view of the losses. Fabrico takes stops, speed and counts live from PLCs or IoT sensors, and gives OEE per line and shift, with operators tagging the cause at the line. See how Fabrico connects to process signals.
What people ask next
How is OEE calculated in a chemical plant?
Divide on-spec output by the best proven rate to get ideal hours, then divide by scheduled hours. The worked steps for one unit are in OEE for continuous processes.
Does a turnaround count against OEE?
A turnaround planned well ahead is normally excluded from scheduled time, but it still lowers TEEP. Unplanned work that gets folded into it should still count as a stop.
Is recycled off-spec a quality loss?
Yes, because it was not good the first time and it used unit time. Report its value separately if recycling recovers most of the material. See how to reduce scrap and rework for reducing it.
Should reaction time count as a speed loss in batch chemistry?
No. Time the recipe requires is part of the ideal cycle, and only time beyond it is a loss. See OEE for batch processes.
What should I track alongside OEE in a chemical plant?
Yield, energy per ton and process safety events at the very least. OEE sees none of them, so read it with production yield and your cost per ton.
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.