BRENNEN

Industries / Food, farming and forestry / Agriculture and farming / Animal feed

Pelleting and extrusion

Pelleting presses the mixed meal through holes in a steel die, usually after softening it with steam, into pellets that hold together through handling; fish feed is instead cooked in an extruder into porous pellets that float, or sink at a set speed.

5

Milestones achieved, from 1931 to 2022.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1931

    A pellet mill that sold

    CPM, founded in San Francisco in 1883 as Toulouse & Delorieux, a maker of wine presses, and later renamed California Pellet Mill, says it introduced "the first commercially successful pellet mill" in 1931. By the company's later figures, about 55% of the 40 million tons of feed made by US mills in 1958 was pelleted, and about 60% of 78.2 million tons in 1978, from 9,977 mills registered with the Food and Drug Administration. A 2016 review of the feed industry's history gives an outside account: Purina introduced pelleted feed in the 1920s, other makers took it up quickly, and by the 1930s many feed mills specialised in pellets.

    onecpm.comimages.engormix.comcambridge.org

    Credit

    California Pellet Mill Co. (CPM), by its own account; spread of pelleting from D. Coffey and colleagues (Journal of Applied Animal Nutrition, 2016)

  2. 1962

    A test for pellet strength

    H. B. Pfost and colleagues described a standard way to measure pellet quality at the 1962 Feed Production School: 500 g of pellets, sieved free of dust, are tumbled for 10 minutes at 50 turns a minute in a box with a baffle, then sieved again, and the share left whole is the pellet durability index (PDI). It became an American Society of Agricultural Engineers standard, S269, in 1969, last revised in 2012.

    LessonThe index depends on how the test is run: in 2009 Pacheco and Stark got 87 to 90% from the same pellets by starting with 300 to 600 g, and 47 to 76% when they added two hex nuts, so results from mills that change the test cannot be compared.

    bookstore.ksre.ksu.edu

    Credit

    H. B. Pfost and colleagues, Kansas State University; standardized by the American Society of Agricultural Engineers

  3. Late 1990s

    Extruders make high-fat fish feed

    Salmon and trout grow on feed far richer in oil than a pellet press can hold. An engineer at the extruder maker Wenger wrote in 2001 that pellet presses are limited to diets under 12% fat, while twin-screw extruders, which he said were "not common" in feed making five years before, handle up to 27%. An extruder cooks the moistened meal at 80 to 130 °C under pressure and forces it through a die. The porous pellets that come out are dried and then soaked with oil under vacuum, to as much as 40% oil, and drying takes about 65% of a fish feed plant's energy, by the account of a 2013 Wageningen University thesis.

    globalseafood.orgedepot.wur.nl

    Credit

    extruder makers including Wenger Manufacturing, by the account of its engineer Galen Rokey; figures from Vukasin Draganovic, Wageningen University

  4. 2010

    Tumbling test misses fish feed

    Mette Sørensen and colleagues made fishmeal feeds with different starches and extruder screw layouts and tested the pellets two ways. A Holmen tester, which blows pellets round a chamber with air, gave durability from 37% to 79%, while the tumbling box gave 98.4 to 98.8% for every feed.

    LessonA test built for livestock pellets could not tell these fish feeds apart, while the air-driven test could, so the test has to suit the feed.

    nofima.com

    Credit

    Mette Sørensen, Giao Nguyen, Trond Storebakken and Margareth Øverland (Aquaculture Research, 2010)

  5. 2022

    Mill records predict durability

    A University of Guelph team trained 12 kinds of machine learning model on 2,471 records from two lines of a commercial feed mill over 8 months, covering recipe, pellet mill settings and weather, to predict each batch's durability index. The best model missed by 3.3 index points on average; outdoor temperature and the amounts of bakery by-product and wheat in the recipe mattered most. A 2025 follow-up on 2,691 records from the same mill, with simpler models, missed by about 2 points.

    ebi.ac.ukpmc.ncbi.nlm.nih.gov

    Credit

    J. You, D. Tulpan, M. C. Malpass and J. L. Ellis, University of Guelph, with a co-author from Trouw Nutrition Canada on the 2025 paper

Watch for

  1. Roller gap set automatically

    Inside a pellet mill, rollers press the meal through the holes of a ring-shaped die, and the gap between rollers and die affects output and pellet quality. Bühler says its Kubex 5 pellet mill, for 15 to 60 tonnes an hour, adjusts that gap during production from its speed and motor current, which prevents rollers slipping and the die blocking, and that the mill has "the lowest energy consumption in its class"; it gives no energy figure.

    Achieved whena customer mill or a university publishes kilowatt-hours per tonne and durability index for a Kubex 5 against a mill with a fixed gap, on the same recipe.

    millingmea.com

    Claimed by

    Bühler, October 2025

Open

  1. 01

    Durability at an unseen mill

    A model that predicts the durability index of each batch from a pellet mill's own logged settings, recipe and weather, trained on records from some mills and tested on a mill it has never seen, published with an anonymised dataset from several mills.

    The brief
    Credit · unclaimed
  2. 02

    Sinking speed by video

    A test that films 100 or more fish feed pellets falling through a clear water column with a phone and measures each one's sinking speed and the share that float, in place of timing a few pellets by hand, checked against careful stopwatch timing.

    The brief
    Credit · unclaimed

Open 01 · the brief

Durability at an unseen mill

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A model that predicts the durability index of each batch from a pellet mill's own logged settings, recipe and weather, trained on records from some mills and tested on a mill it has never seen, published with an anonymised dataset from several mills.

Why it is still open

Both Guelph models were built and tested on one commercial mill. The authors write that "the data was from the same mill and further studies could examine the generalization ability of the models to predict outcomes from other feed mills". A GitHub search in November 2026 found no public durability dataset or model.

pmc.ncbi.nlm.nih.gov

Built with what exists

Each mill's process-log exports as spreadsheet files; the mill's own durability tests, run by its staff with a tumbling box to ASAE S269.5 or a Holmen tester; Python with pandas, scikit-learn and LightGBM; recipe names replaced by codes so no mill's formulas are exposed.

  1. First step

    Ask one mill's quality manager for one month of pellet mill logs with the matching durability tests, and refit the four variables Guelph's 2025 model found most influential (expander temperature, fat, acid detergent fibre, a laboratory measure of tough plant fibre, and humidity at the pellet mill) to see how far its coefficients carry over.

  2. Done looks like

    Records from at least 3 mills with at least 2,000 durability tests each; a model trained on the others predicts the left-out mill's batches with an average error at or below 3.3 index points, the error of Guelph's 2022 model inside its own mill; data and code public.

Open 02 · the brief

Sinking speed by video

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A test that films 100 or more fish feed pellets falling through a clear water column with a phone and measures each one's sinking speed and the share that float, in place of timing a few pellets by hand, checked against careful stopwatch timing.

Why it is still open

How fast pellets sink decides whether fish eat them: a 2013 Wageningen thesis notes that "pellets that stay afloat will not be eaten by the fish; pellets that sink too fast may escape being eaten." Published studies still time pellets by hand: a 2026 Universiti Putra Malaysia study, for example, timed three pellets per sample with a digital stopwatch over a 0.34 m water column. Searches of Europe PMC and GitHub in November 2026 found no video method for feed pellets.

edepot.wur.nlpmc.ncbi.nlm.nih.gov

Built with what exists

A clear acrylic tube about 1 m tall filled with fresh or salt water at a recorded temperature; a phone filming at 60 to 240 frames a second against a printed scale; OpenCV background subtraction and a simple tracker in Python to follow each pellet.

  1. First step

    Build the water column and film 50 pellets each of one floating and one sinking commercial feed, then time the same pellets by stopwatch.

  2. Done looks like

    On 30 feed samples, from floating catfish and tilapia feed to sinking salmon feed, the video's mean sinking speed is within 5% of stopwatch timing of 20 pellets per sample, and the method reports a speed for at least 100 pellets per sample and the floating share; videos and code published.