Industries / Food, farming and forestry / Food and beverage processing / Flour milling
Grinding
Grinding breaks the cleaned grain open and reduces it to flour, between stones, metal rolls or, in small mills, spinning hammers or plates.
5
Milestones achieved, from c. 125 CE to 2015.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
c. 125 CE
Sixteen water wheels at Barbegal
Near Arles in southern France, Roman builders set 16 water mills in two rows of eight down a hillside, fed by an aqueduct diverted to them around 120–130 CE. The complex is often called one of the first industrial sites, and one estimate puts its capacity at 25 tonnes of flour a day (J. R. Spain, 2008). In 2018 a study of the lime scale left on the wooden wheels showed that the mills often stood idle for months at a time. The authors suggest the mills made hardtack, a long-keeping biscuit, for nearby ports rather than daily flour for a city.
LessonCapacity alone does not show how a mill was used; the deposits it left can.
Credit
Sürmelihindi, Leveau and colleagues (Science Advances, 2018)
1790
A mill that runs itself
Oliver Evans of Delaware designed a water-powered flour mill in which machines moved grain and flour through every step of milling, with no need to carry it by hand. A raking machine called the hopper boy spread and cooled the warm meal before sifting, a job done before by boys with rakes. George Washington reviewed and signed Evans's patent application in 1790 and installed the system in his own mill at Mount Vernon in 1791. Evans's 1795 book, The Young Mill-Wright and Miller's Guide, spread the design.
Credit
Oliver Evans
1870s
Iron rolls replace millstones
A roller mill crushes grain between pairs of turning cylinders instead of between two stones. With rolls, the miller can open the grain gradually, in many light passes with sifting in between, which is called gradual reduction. In Budapest, Abraham Ganz pioneered rolls of chilled cast iron, which have a very hard surface, and András Mechwart began making them around 1870 for sale worldwide. Friedrich Wegmann of Naples sold rolls of porcelain, and said he had sold 6,000 machines in Europe by 1881. In 1878 Henry Simon installed Britain's first all-roller, gradual-reduction mill, for McDougall Brothers in Manchester.
Credit
Abraham Ganz, András Mechwart, Friedrich Wegmann, Henry Simon
1881
World's largest flour mill
The Pillsbury A Mill in Minneapolis began work on 5 July 1881, and its US National Historic Landmark nomination calls it the largest and most advanced flour mill in the world at the time. It started at 4,000 barrels of flour a day and later settled at about 17,500. Mills like it made Minneapolis the US capital of flour milling from 1880 to 1930.
Credit
C. A. Pillsbury & Company; architect L. S. Buffington
2015
Why atta stays stone-ground
In India, wheat is eaten mainly as chapati, a flatbread made from atta, a whole wheat flour, and families have long had it ground fresh in a chakki, a small mill with a pair of stones. Researchers ground the same wheat in a chakki and in hammer, disc, pin and roller mills. Chakki atta had the most damaged starch (16.0% against 9.1% from the roller mill), absorbed the most water (85% against 71.5%), and made the best-rated chapatis.
LessonThe right mill depends on the food; roller mills built for white bread flour do not make the best chapati flour.
Credit
Inamdar, Sakhare and Prabhasankar (Journal of Food Processing and Preservation, 2015)
Watch for
AI sets the roll gap
Bühler says its Grinding Optimizer software reads roller temperatures, vibration, power use and yield, and recommends changes to the gap between the rolls. It says this cut energy use in a mill's grinding section by 8–10% and, in one case, the power drawn by one pair of rolls by up to 38%, with quality kept in range. About half a dozen mills ran it as a pilot in 2026, in advisory mode, with millers deciding whether to act.
Achieved whena mill that runs it publishes a season of energy and flour-yield figures, checked by an outside auditor against the same mill before the change.
Claimed by
Bühler, in a white paper dated 2026
Village mill on less power
Researchers from the University of Liège and the Polytechnic School of Abomey-Calavi measured five village maize mills on solar mini-grids in Benin during normal work, including one stone mill they had fitted to inject water while grinding. For wet maize, they report that their mill ground 108 kg an hour at 0.03 kWh per kg, against 0.17 kWh per kg for the fuel-powered mill.
Achieved whenthe water-injection mill is tested by someone other than its developers, on dry and wet maize, over a full season.
Claimed by
Maxime Sehou and colleagues, ECOS 2025 conference, Paris, June–July 2025
Open
01
Grind size from a photo
A phone method that measures how fine a sample of maize or sorghum meal is from a single photo, so that village hammer and plate mills, and the people who use them, can compare meal fineness without a laboratory sieve set.
The briefCredit · unclaimed02
Village mill energy per kilo
An open dataset of the energy that working village maize mills use per kilogram: litres of diesel per kg for diesel mills and kWh per kg for electric ones, with screen size, grain moisture and mill type recorded for each batch.
The briefCredit · unclaimed
Open 01 · the brief
Grind size from a photo
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A phone method that measures how fine a sample of maize or sorghum meal is from a single photo, so that village hammer and plate mills, and the people who use them, can compare meal fineness without a laboratory sieve set.
Why it is still open
Photo measurement of particle size has been checked against sieves for soil (arXiv 2311.06613, which reports results "comparable" to sieving) and is sold for gravel (the Granumetrics app), and coffee drinkers use phone apps for grind size without agreed validation. No method checked against sieves for maize or sorghum meal was found.
Built with what exists
A phone; a black card with a printed scale target; open image tools such as Fiji (ImageJ) or OpenCV; a stack of test sieves for the reference.
First step
Sieve and photograph five samples of meal from one mill, each ground through a different screen.
Done looks like
30 meal samples from at least five mills, where the photo's median particle size is within 10% of the sieve result and its share of particles above 1 mm within 5 percentage points, published with the photos and sieve data.
Open 02 · the brief
Village mill energy per kilo
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open dataset of the energy that working village maize mills use per kilogram: litres of diesel per kg for diesel mills and kWh per kg for electric ones, with screen size, grain moisture and mill type recorded for each batch.
Why it is still open
Figures for Africa's village mills rest on assumptions and small samples. CrossBoundary assumes 50 kg of maize milled per litre of diesel across an estimated 500,000–750,000 mills in Sub-Saharan Africa. A January 2026 white paper for Partners in Food Solutions, SNV and GAIN drew on 48 survey answers and six site audits, calls its findings indicative, and gives about 2.0 MJ/kg for small hammer mills in its text but 0.3–0.7 MJ/kg for micro and small processors in a table. Wu, Wanume and Modi (Energy Policy, 2026) analysed a year of 15-minute meter readings from 21 grain mills on NOA Uganda's mini-grids, but recorded electricity only, not the grain milled.
Built with what exists
A mini-grid operator's smart-meter readings for electric mills, so nobody works on live wiring; fuel logs and a measured jerrycan for diesel mills; a hanging scale for each customer's grain; KoboToolbox forms on a phone.
First step
Ask a mini-grid operator that already logs mill meters, such as NOA Uganda, to match a week of one mill's readings against weighed customer batches.
Done looks like
At least 30 mills in one country, each logged for five working days, with kilograms per kWh or per litre and the spread between mills, published as open data with the method.