OEE for Paper and Pulp: Breaks, Grades, and Broke

OEE for a paper machine measures salable tonnes against the tonnes the machine could have made at full design speed during planned time. Availability is sheet-on-reel time, performance is running speed against rated speed, and quality is on-spec tonnes after web breaks, grade-change broke, and reel rejects are removed.

A paper machine is not a discrete-parts line, and forcing it into a piece-count OEE model gives a number nobody trusts. The web runs continuously at high speed; the machine rarely comes to a clean stop; and most of its losses show up as speed held back, sheet torn mid-run, or off-spec paper quietly repulped rather than as tidy downtime events. This post maps the three OEE factors onto how a continuous web actually behaves, so paper and pulp operations get a number that reflects the losses that really cost them tonnes.

How is OEE defined for a paper machine?

OEE on a paper machine is best read in tonnes, not minutes, because the asset makes a continuous product measured by weight. The three factors keep their names but take on paper-specific meanings.

The subtlety is that a paper machine can be running, the reel turning, the number counting up, while it is losing badly. During a grade change it may never stop, yet make a stretch of off-spec paper that goes straight to the repulper as broke. That is a quality loss with zero downtime. Miss it and OEE looks far healthier than the salable output justifies. The discipline that keeps the OEE calculation honest on discrete lines matters even more here, because the losses hide better.

Why do web breaks dominate paper-machine availability?

Web breaks dominate because one break can cost far more than the tear itself. When the sheet snaps, the machine loses the tail, and re-threading through the press and dryer sections at speed takes minutes to tens of minutes, during which the reel is not turning and the stock is still flowing. A machine averaging a couple of breaks a day can lose more availability to threading than to any scheduled stop.

Breaks are also the classic frequency-versus-duration trap. A single long breakdown is easy to remember and log; a run of short breaks blends into the shift and gets undercounted, exactly the failure mode described in the six big losses. Runnability programs exist because break reduction is the highest-leverage availability lever on most machines: published mill audits have reported break-frequency reductions on the order of 40% from systematic runnability work, which flows straight to sheet-on-reel time. Treat every break as a stop with a cause code, wet-end, press, dryer, edge, the same way a discrete line treats machine downtime and the Pareto tells you where the tonnes are going.

Performance loss on a continuous machine is subtler than on a discrete line and often larger than the crew believes. A paper machine is frequently run below its rated speed on purpose, to hold formation, to stay inside the dryers' evaporation limit, or to keep a marginal felt from causing breaks. None of that shows as downtime; it shows as a performance factor quietly stuck in the eighties. Because the loss is continuous rather than event-driven, it is easy to normalize: "that grade always runs at this speed" becomes an unquestioned ceiling. Tracking actual speed against the grade's true design rate every reel is what surfaces the gap, and closing even part of it compounds across every tonne the machine makes.

How do grade changes and broke show up in the number?

Grade changes and broke are where paper OEE diverges most from a discrete-parts line, because the loss is tonnes of off-spec paper, not stopped time. Switching basis weight, brightness, or coating forces a transition during which moisture, caliper, and weight profiles drift outside spec. That transition paper is real production the machine made at full speed, it counts against quality, not availability, because the reel never stopped.

The trap is that broke is repulped and reused, so it feels free. It is not. Every tonne of broke is a tonne of design capacity, steam, and fiber-preparation energy spent making paper you cannot sell, plus the recirculation load it adds back to the wet end. An honest paper OEE counts salable tonnes over total tonnes produced, so grade-change broke, reel rejects, and off-spec runs all land in the quality factor where they belong. If they instead disappear into "recycled fiber," quality reads high and the machine looks more effective than the shipping scales say. This is the same logic as first-pass yield: what matters is what came right the first time, not what was recovered.

Scheduling compounds this. A grade sequence that jumps between light and heavy basis weights, or between coated and uncoated, forces long transitions and more broke than a sequence that steps gently through the grade ladder. That makes the production schedule an OEE lever, not just a service-level decision: sequencing grades to minimize transition tonnage can recover salable output without touching the machine at all. It is one of the few OEE improvements that costs nothing but planning discipline.

How do you build an honest paper-machine OEE?

Build it in tonnes, capture the continuous losses the discrete model misses, and keep grade broke visible. Six steps take a machine from turning reels to a number the mill can act on.

1. Set the design rate correctly. Ideal rate is design speed × running trim width × basis weight, expressed in tonnes per hour. Because it moves with grade, compute performance against the design rate for the grade actually running, not a single nameplate.
2. Log every web break as a stop with a cause code. Section, likely cause, threading time. Short breaks especially, those are the undercounted ones.
3. Measure availability as sheet-on-reel time. Planned time minus breaks, threading, and unplanned stops, divided by planned time. Scheduled shutdowns sit above the OEE line.
4. Compute performance against the grade's design rate. Actual tonnes per hour over design tonnes per hour captures speed held back for runnability or drying.
5. Count salable tonnes for quality. Subtract grade-change broke, reel rejects, off-spec moisture or caliper runs, and edge trim from total tonnes. Salable over total is the quality factor.
6. Reconcile to the scale. Salable tonnes from OEE should match the tonnes shipped and inventoried. When they diverge, broke is being undercounted somewhere.

What does the data say about paper-industry capacity?

Paper is a capital-heavy, capacity-sensitive industry, which is why squeezing tonnes out of an existing machine usually beats adding one. The Federal Reserve's industrial data, published in the G.17 Industrial Production and Capacity Utilization release and tracked as the paper manufacturing capacity utilization series (NAICS 322) on the St. Louis Fed's FRED database, shows the sector running well below full capacity, in line with total manufacturing near 75.8% in April 2026. Those are economic measures, not machine OEE, so use them as context: on a continuous asset, the gap between rated speed and running speed, plus break time and broke, is usually worth more recoverable tonnage than any capital project. Set targets against what a good OEE score means for your grade mix rather than a universal benchmark, and price recovered points in throughput at contribution margin.

What makes paper OEE trustworthy?

Trustworthy paper OEE comes from capturing the losses that never fully stop the machine. Two failures cause most inflated numbers. The first is undercounted short breaks: the sheet snaps, threads in three minutes, and the shift moves on without a log entry, so availability reads high. The second is invisible broke: grade-change and off-spec tonnage repulped without ever hitting the quality factor. Both are solved by measuring continuously from the machine rather than reconstructing the shift from memory. Harmony derives availability, speed, and salable-yield losses from machine signals and operator-tagged reasons (see the platform), which is what keeps breaks and broke from hiding in a continuous process, and it avoids the estimation errors catalogued in common OEE mistakes. For proof the approach holds up on a real floor, see the CLS field story then run your grade's numbers with the OEE calculator. The payoff is specific: on a continuous web, a single point of recovered performance or a single fewer break per day compounds across every reel, every shift, all year.