BRENNEN

Industries / Food, farming and forestry / Food and beverage processing / Flour milling

Testing and flour safety

Mills test wheat and flour for how they will behave in a bakery, and for germs and mould poisons that grinding does not remove.

6

Milestones achieved, from 1928 to 2024.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1928

    A kneading machine that measures

    Carl Wilhelm Brabender, founder of Brabender in Duisburg, Germany, developed the first Farinograph in 1928. It mixes flour and water and records the dough's resistance to kneading, which gives the flour's water absorption, the amount of water it takes to make a dough of set stiffness, and how the dough behaves as it is mixed. Mills and bakeries still use it to check flour against specifications.

    millermagazine.com

    Credit

    Carl Wilhelm Brabender

  2. 1965

    Reading grain with light

    Near-infrared (NIR) analysis measures how much light just beyond the visible range a sample absorbs at each wavelength, which tracks its water, protein and other contents. Karl Norris of the US Department of Agriculture published a method with Hart in 1965 to measure the moisture of grain and seeds in this way. In 1982 Hruschka and Norris predicted the protein of ground wheat with a correlation of 0.998 and a standard error of 0.15 percentage points. A 2026 article sponsored by the software firm Equinom says most modern flour mills now run NIR tests as standard practice for protein, moisture and ash.

    spectroscopyonline.comspectroscopyonline.comworld-grain.com

    Credit

    Karl Norris and colleagues, US Department of Agriculture

  3. 1968

    A standard test for sprouting

    When wheat sprouts in a wet harvest, it makes more alpha-amylase, an enzyme that breaks down starch. The Falling Number test, developed in Sweden by Sven Hagberg and Harald Perten, times in seconds how long a stirrer takes to fall through a hot flour and water gel; more enzyme thins the gel faster and gives a lower number. The International Association for Cereal Science and Technology (ICC) approved it as Standard 107/1 in 1968, last revised in 1995. The American Society of Baking says the AACC International method 56-81.03 is identical.

    icc.or.atasbe.org

    Credit

    Sven Hagberg and Harald Perten; ICC

  4. 2011

    Flour with a kill step

    A kill step is a process, such as heating, that kills germs. ConAgra Mills introduced SafeGuard, a heat-treated flour sold as ready to eat in foods that are not cooked, in 2011; Ardent Mills now sells it with "up to a 5-log validated pathogen reduction", meaning up to 100,000 times fewer live germs, by its own account. Honeyville, Siemer Milling and Bay State Milling sell heat-treated flours too; Bay State's SimplySafe, launched in October 2017, is validated by the company's account to cut Salmonella up to 5-log.

    world-grain.comfoodbusinessnews.net

    Credit

    ConAgra Mills; Ardent Mills

  5. 2016

    Raw flour made people ill

    The US Centers for Disease Control and Prevention (CDC) counted 63 people in 24 states infected with Shiga toxin-producing E. coli O121 or O26, of whom 17 were hospitalized, with illnesses from 21 December 2015 to 5 September 2016. General Mills recalled Gold Medal and other flours from 31 May 2016 and widened the recall twice. The CDC told the public that it is not safe to eat raw dough or batter made with any flour.

    LessonGrinding does not kill germs that come in on the grain, so flour must be cooked or treated before it is eaten.

    archive.cdc.gov

    Credit

    CDC and state health departments; General Mills

  6. 2024

    Tighter EU limits on DON

    Deoxynivalenol (DON) is a poison made by Fusarium moulds that grow on grain in wet seasons. Commission Regulation (EU) 2024/1022, applying from 1 July 2024, lowered certain existing limits. It sets 600 micrograms per kilogram (µg/kg) for cereal milling products such as flour, 1,000 µg/kg for most unprocessed cereal grains, 1,500 µg/kg for unprocessed durum wheat and maize, and 400 µg/kg for bakery wares. It cites the European Food Safety Authority's finding that average long-term intakes in infants, toddlers and other children were above the tolerable daily intake.

    eur-lex.europa.eu

    Credit

    European Commission

Watch for

  1. Screening DON by light

    Researchers at BOKU University in Tulln, Austria, report a near-infrared method that sorts wheat samples into those above and below the EU limit of 1,000 µg/kg DON without grinding them. Their best version found 84% of the samples over the limit, with an F1 score, a combined measure of hits and false alarms, of 0.74. It used Austrian samples only, and the spectra are published.

    Achieved whenthe work passes peer review and another lab tests the method on wheat from other harvests or countries and reports similar rates.

    api.crossref.orgzenodo.org

    Claimed by

    Dedy Nadeak, Rudolf Krska, Stephan Freitag and colleagues, BOKU University, in a preprint of July 2026

Open

  1. 01

    Loaf volume from mill flour

    A test of whether models that predict loaf volume from near-infrared spectra, trained on wheat breeding lines, work on the blended flours that mills sell: measure a set of commercial flours, predict their loaf volume with a model trained on published breeding data, bake them, and publish the errors.

    The brief
    Credit · unclaimed
  2. 02

    Heat-treated flour baking test

    A side-by-side test of ready-to-eat, heat-treated flours against ordinary flours of the same type from the same mills, on standard tests: falling number, farinograph water absorption and stability, gluten content, and a standard cookie or small-loaf bake.

    The brief
    Credit · unclaimed

Open 01 · the brief

Loaf volume from mill flour

Credit · unclaimed

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

What to build

A test of whether models that predict loaf volume from near-infrared spectra, trained on wheat breeding lines, work on the blended flours that mills sell: measure a set of commercial flours, predict their loaf volume with a model trained on published breeding data, bake them, and publish the errors.

Why it is still open

CIMMYT, the International Maize and Wheat Improvement Center, has published near-infrared spectra with loaf volume, protein, dough tests and flour yield for the wheat lines behind a 2025 study, with two files open and the rest released on request. On 1,001 held-out samples the study's best model reached an R² of 0.74 and an error of 22.92 mL per 100 g (Li, He, Ibba and colleagues, Food Chemistry, 2025). Those samples are single lines from one breeding program. Ziegler and colleagues (2025) write that predictions of loaf volume from protein content or flour spectra "are poor", and Equinom's models for mill blends (June 2026) are closed. No test of a breeding-line model on commercial mill flour was found.

doi.orgbuscador.una.edu.niapi.semanticscholar.org

Built with what exists

CIMMYT's data, requested through its Dataverse; time on a Fourier-transform near-infrared instrument like the Thermo Antaris named in CIMMYT's file names, at a university cereal lab; flours bought from shops and mill outlets; a standard small-loaf test bake; Python with scikit-learn.

  1. First step

    Download CIMMYT's two open files, fit a model on one and test it on the other, and request access to the rest.

  2. Done looks like

    60 commercial flours from at least 10 mills, each with its spectrum, protein and measured loaf volume, published with the prediction error of the breeding-line model next to that of an estimate from protein alone.

Open 02 · the brief

Heat-treated flour baking test

Credit · unclaimed

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

What to build

A side-by-side test of ready-to-eat, heat-treated flours against ordinary flours of the same type from the same mills, on standard tests: falling number, farinograph water absorption and stability, gluten content, and a standard cookie or small-loaf bake.

Why it is still open

Mills make heat-treated flour by their own processes, and Ardent Mills says its SafeGuard process "doesn't alter gluten functionality" (World Grain, 2018). Laboratory studies heat their own flour samples; a search of journal records on 18 November 2026 found no independent side-by-side test of commercial heat-treated flours.

world-grain.com

Built with what exists

Heat-treated flours from several mills and their ordinary counterparts, bought in small lots; tests at a contract cereal lab; a home-scale oven for the bake test.

  1. First step

    List which heat-treated flours can be bought in small amounts, and order two pairs for a first round of falling number and farinograph tests.

  2. Done looks like

    At least 8 pairs tested, with all results published and each difference shown against the lab's own repeat-test spread.