Industries / Food, farming and forestry / Food and beverage processing / Flour milling
Dust safety
Dust safety is keeping flour and grain dust from catching fire or exploding inside a mill, and from damaging the lungs of the people who work in it.
7
Milestones achieved, from 1878 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1878
Washburn A Mill explodes
On the evening of 2 May 1878 the Washburn A Mill, a flour mill in Minneapolis, exploded, and the neighbouring Diamond and Humboldt mills exploded with it. All 14 workers in the A Mill died and four more died in the fire that followed. University of Minnesota professors S.F. Peckham and Louis W. Peck showed by experiment that burning flour dust caused it, and concluded that two millstones running dry had struck a spark. The rebuilt A Mill, running by 1880, was built to be safer.
LessonFlour dust in the air can explode like a fuel, so sparks and hot surfaces near it must be controlled.
Credit
Cadwallader Washburn (owner); S.F. Peckham and Louis W. Peck (investigation)
1979
Bremen mill rebuilt safer
A flour dust explosion at the Roland mill (Rolandmühle) in Bremen, Germany, killed 14 people and seriously injured 17, by Martin Glor's count. The mill, almost completely destroyed, was rebuilt with explosion protection throughout: redesigned silos and bridges between buildings, dust extraction, and pressure relief, weak panels that let an explosion's pressure escape before it bursts the building. The mill's H.P. Erling wrote in 1985 that pressure relief is expensive but should be part of every new or rebuilt mill.
Credit
Bremer Rolandmühle; H.P. Erling
1987
US grain handling rule
The US Occupational Safety and Health Administration (OSHA) issued its grain handling standard on 31 December 1987. It covers flour mills, feed mills, rice mills and dry corn mills as well as grain elevators, and requires permits for hot work such as welding and a written housekeeping program to limit dust. Its rule that dust deeper than 1/8 inch (3.2 mm) be removed at once applies only to set areas of grain elevators.
Credit
US Occupational Safety and Health Administration
1999
EU rules for explosive dust
EU Directive 1999/92/EC of 16 December 1999 requires employers to divide places where explosive dust clouds can form into zones 20, 21 and 22, from a cloud present continuously or for long periods to one unlikely and short-lived, and to write and keep up an explosion protection document. Member states had to make it law by 30 June 2003, and workplaces already in use had three more years to comply.
Credit
European Parliament and Council of the European Union
2018
Canada resets flour dust limit
In 2001 the American Conference of Governmental Industrial Hygienists (ACGIH), which publishes recommended exposure limits, set 0.5 milligrams per cubic metre (mg/m3) for flour dust, which causes asthma. Canada's Labour Program had taken 157 samples in 14 Canadian flour mills in 2000: 97.1% were above 0.5 mg/m3, and 42.3% of the workers sampled were above 10 mg/m3. In June 2018 Canada set its federal limit at 3 mg/m3, calling 0.5 impracticable for mills without respirators worn all shift. Quebec samples from 2008 to 2015 showed 45% at or below 3 mg/m3, and the government put the high readings down to cleaning with compressed air.
LessonA limit no mill could meet was replaced by one that dust controls can reach, and the worst exposures came from a cleaning habit, not from milling.
Credit
Employment and Social Development Canada, Labour Program
2023
Faked cleaning logs convicted
Dust explosions and fire at Didion Milling's corn mill in Cambria, Wisconsin, on 31 May 2017 killed 5 of the 19 workers on shift and injured 14. Didion pleaded guilty to falsifying the mill's dust-cleaning and dust-collector logs and agreed to a US$1 million fine and US$10.25 million in restitution; in October 2023 a federal jury convicted two managers, including over cleaning entries filled in for cleanings never done. The US Chemical Safety Board's final report, in December 2023, asked OSHA to write a combustible dust standard for all industries; that recommendation is still open.
LessonA housekeeping rule protects workers only if someone checks that the cleaning in the log was done.
Credit
US Chemical Safety and Hazard Investigation Board (CSB); US Department of Justice, EPA and OSHA
2025
One US dust standard
The National Fire Protection Association's NFPA 660 combines and replaces six earlier dust standards, among them NFPA 61 for agricultural and food processing plants. It walks users through a dust hazard analysis, a written review of where dust could burn or explode and what controls each hazard. After Didion, the CSB had asked NFPA to unify NFPA 61's dust hazard analysis rules, and it closed that request as exceeded.
Credit
National Fire Protection Association (NFPA)
Watch for
AI watch over mill dust
iFactory says its software keeps a live map of dust hazard analysis status across a grain or flour mill and flags equipment whose hazard has changed. It claims 35% fewer dust events and 95% predictive accuracy, and names no customer.
Achieved whena named mill publishes before-and-after records of dust fires and alarms, checked by an insurer, regulator or other outside party.
Claimed by
iFactory, May 2026
Open
01
Cheap sensors for mill dust
A test of whether low-cost particle sensors, the kind sold for home air quality, can track flour dust at the levels set in workplace limits, by comparing them with standard sampling pumps and filters in a closed test box.
The briefCredit · unclaimed02
Test AI dust hazard reviews
A test of whether AI tools, given only a description of a mill as it stood before an explosion, find the hazards that investigators later found. Each investigated case becomes a "before" description and an answer key.
The briefCredit · unclaimed
Open 01 · the brief
Cheap sensors for mill dust
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test of whether low-cost particle sensors, the kind sold for home air quality, can track flour dust at the levels set in workplace limits, by comparing them with standard sampling pumps and filters in a closed test box.
Why it is still open
Ruiter and colleagues tested four low-cost sensor types in two industrial bakeries in 2023 and compared them with respirable samples, the fine fraction that reaches deep into the lungs; placement and workplace changed how well they did. Canada's mill survey measured inhalable dust, the coarser fraction breathed in through nose and mouth, and found 97.1% of mill samples above 0.5 mg/m3; Canada's government considers that level achievable in bakeries but not in mills. A search of Europe PMC on 18 November 2026 found no test of low-cost sensors in a flour mill.
Built with what exists
Plantower PMS5003 and Sensirion SPS30 sensors logged by a Raspberry Pi or ESP32 board; IOM inhalable samplers with personal pumps and an analytical balance, borrowed from a university occupational hygiene laboratory; a sealed acrylic box with a small fan; plain flour. Keep dust at workplace exposure levels, keep sparks and flames away, and wear a dust mask.
First step
Build the box, sift a weighed amount of flour into it with two sensors running, and record the concentrations it reaches.
Done looks like
20 paired filter samples between 0.5 and 20 mg/m3, a correction curve for each sensor type, the share of sensor readings within 25% of the filter result, and the code and data published.
Open 02 · the brief
Test AI dust hazard reviews
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test of whether AI tools, given only a description of a mill as it stood before an explosion, find the hazards that investigators later found. Each investigated case becomes a "before" description and an answer key.
Why it is still open
At Didion the CSB found that not every dust-handling unit had a dust hazard analysis. AI products sold to mills monitor and track hazard analyses rather than write them. Published work on AI-drafted hazard studies covers chemical-process studies called HAZOPs (for example arXiv 2611.10721, November 2026), and the safety engineer Ian Sutton had ChatGPT draft lessons from a finished CSB report on an ammonia release (September 2025); neither tests an AI on a dust hazard it has not been told about.
Built with what exists
Current large language models such as Claude Fable 5.2 (Note: "Three working devices") and the open-weight Mistral Large 4 (Note: "A trillion open parameters"); public investigation reports from the CSB and others; the NFPA 660 dust hazard analysis method as the template.
First step
Write the "before" description of the Didion corn mill from the CSB report, leaving out anything that reveals what happened, and run two models on it.
Done looks like
Five investigated dust explosions at mills or grain plants turned into before-and-answer pairs, at least two AI tools run on each, and a published tally of the investigators' findings each tool caught and of its own entries that two process safety engineers judged sound.