Industries / Industrial equipment and services / Industrial machinery / Conveyors and material handling
Safety and maintenance
This work keeps people out of the moving parts of conveyors and keeps the machines from failing, through guards and stop switches, written standards, inspection and monitoring.
5
Milestones achieved, from 1957 to 2025.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1957
A safety code for conveyors
B20.1-1957, the Safety Code for Conveyors, Cableways, and Related Equipment, is the edition that US construction safety rules still require conveyors to meet. The same rule (29 CFR 1926.555) asks for a stop control at the operator's station, a warning sound before start-up, emergency stops that must be reset before the conveyor can restart, and guards where a conveyor passes over people. ASME now publishes B20.1; its 2024 edition covers bulk, package and unit-handling conveyors from design to operation.
Credit
ANSI, as OSHA cites the code; the American Society of Mechanical Engineers (ASME), which publishes it now; OSHA
2015
Where conveyor injuries happen
Britain's Health and Safety Executive (HSE) reports that conveyors are involved in 30% of machinery accidents in the food and drink industries, more than any other class of machine. By HSE's figures, 90% of conveyor injuries happen on flat-belt conveyors, 90% involve well-known hazards such as the in-running nip between belt and end roller, and 90% happen in normal work: production, clearing blockages and cleaning. Its guidance asks for safety by design, such as lift-out rollers that cannot trap fingers, and interlocked guards where conveyors are cleaned every day.
LessonMost injuries come during routine cleaning and clearing, so guards have to suit those jobs as well as production.
Credit
HSE, "A recipe for safety" (HSG252), second edition, 2015
2019
Monitors that missed the heat
Australia's New South Wales Resources Regulator reported pulley failures that started fires in underground coal mines. One pulley's temperature monitor showed a rise only from 33 °C to 45 °C before the pulley failed, below trip points set at 65 °C or 85 °C. In a second incident a faulty probe had been refitted during a pulley change three months earlier, with the new probe left tied up nearby. The regulator told mines to test monitoring devices against the real temperatures of components. Queensland's inspectorate had reported in 2016 a defective idler bearing left running until a planned maintenance day, which caught fire first.
LessonA reading that does not rise when the belt runs or fall when it stops is itself a warning, and nobody had noticed it.
Credit
NSW Resources Regulator (bulletin SB19-14); Resources Safety and Health Queensland
2024
Robot sites set the pace
A US Senate committee staff report (Health, Education, Labor and Pensions Committee majority staff, chaired by Bernie Sanders) found after an 18-month investigation that Amazon warehouses recorded more than 30% more injuries than the industry average in 2023. Workers at Amazon's robotics sites, where robots carry tall storage containers to fixed stations, told the committee they were expected to stow up to 350 items an hour and pick up to 450. The report says an internal Amazon project, Project Soteria, studied a proposal to slow the pace in robotics facilities. Amazon rejected the findings and says its own data showed lower injury rates at robotics sites than at others in 2022.
LessonA station that brings goods to the worker removes walking but can leave a person turning and reaching hundreds of times an hour, so its layout and the rate set for it decide the strain.
Credit
HELP Committee majority staff; Amazon (response)
2025
Fewer severe, more minor injuries
Gordon Burtch of Boston University, Brad Greenwood of George Mason University and Kiron Ravindran of IE University compared Amazon's robotic fulfillment centers, where machines take over heavy lifting and long walks, with its older ones. In ILR Review in 2025 they reported that robotics were associated with a 40% decrease in severe injuries, such as broken bones and falls that cost work days, and a 77% increase in non-severe ones, such as sprains, strains and repetitive motion injuries, which rose most around Prime Day and the winter holidays. They attribute the rise at least partly to the faster pace of work; in online posts they analyzed, workers at robotic sites described pick rates two to three times those at older sites in some cases.
LessonA safety record that counts only injuries that cost work days can miss the repeated strain that robot stations add, as George Mason University's summary of the study points out.
Credit
Gordon Burtch, Brad Greenwood and Kiron Ravindran
Watch for
Fiber that listens for idlers
Huawei says its AI fiber-sensing system, which uses an optical fiber along the conveyor as a continuous microphone, tells failing idlers from healthy ones by their sound, and that it answers the high false-alarm rates that have held such systems back. The article gives no detection or false-alarm figures and names no mine.
Achieved whenA mine or researcher outside Huawei publishes months of detection and false-alarm counts checked against manual inspection.
Claimed by
Huawei, May 2026 (Mining Weekly)
Amazon's falling injury rates
Amazon says its global recordable incident rate improved 43% and its lost-time incident rate 70% over the six years to 2025, that strains and sprains make up more than half of its recordable injuries, and that it has added assistive robots to cut them.
Achieved whenThe site-by-site summaries Amazon files with OSHA's Injury Tracking Application show the same fall, and an outside analysis compares robotic and non-robotic sites.
Claimed by
Amazon, March 2026
Open
01
Probe faults on a bench
A bench rig with a heated bearing housing that reproduces the probe mistakes in regulators' reports (a probe not touching the housing, refitted in the wrong place, the real probe left tied up nearby) and measures how far and how late each one reads against a reference thermocouple.
The briefCredit · unclaimed02
Injury rates by equipment type
A public table of warehouse injury rates that matches each site's yearly OSHA filing to the equipment it runs (goods-to-person robots, automated storage, sorters or none), taken from operators' own announcements.
The briefCredit · unclaimed
Open 01 · the brief
Probe faults on a bench
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A bench rig with a heated bearing housing that reproduces the probe mistakes in regulators' reports (a probe not touching the housing, refitted in the wrong place, the real probe left tied up nearby) and measures how far and how late each one reads against a reference thermocouple.
Why it is still open
After one pulley failed while its monitor read 45 °C against a 65 °C trip, the NSW regulator asked mines in 2019 to test temperature monitors against real component temperatures, but gave no figures for how far each common mounting fault reads low. A Crossref search in November 2026 found a 2023 paper on an optical-fiber bearing temperature monitor for conveyors and none that measured mounting faults.
Built with what exists
A cast-iron pillow-block bearing housing with a cartridge heater, the RTD and thermocouple probes sold for conveyor bearings, an independent thermocouple logger, a thermal camera, and a microcontroller logging every channel.
First step
Buy a pillow block, a cartridge heater and two probes, and log one fault, a probe hanging loose, against a reference thermocouple.
Done looks like
A published table for at least six mounting faults, each giving the offset in °C and the lag in minutes at three heating rates, with the rig drawings and raw logs.
Open 02 · the brief
Injury rates by equipment type
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of warehouse injury rates that matches each site's yearly OSHA filing to the equipment it runs (goods-to-person robots, automated storage, sorters or none), taken from operators' own announcements.
Why it is still open
Studies so far cover a whole sector or one company. Burtch, Greenwood and Ravindran compared robotic and older sites inside Amazon only (ILR Review, 2025). A 2024 study of California warehouses in OSHA's Injury Tracking Application (Beckman, Andamel and Harrison) found a peak of 5.87 injuries per 100 workers in 2021, and its abstract reports no split by equipment. The Senate report used Amazon's own records, and Amazon's comparison of robotic and other sites is its own. No public table across operators and types of equipment turned up.
Built with what exists
OSHA's Injury Tracking Application downloads (yearly summaries since 2016, case detail with coded injury types for 2023 onward), operators' press releases naming sites and systems, and an AI agent to match addresses and pull system types from announcements, with a person checking every match.
First step
Download the 2025 summary file and list every general warehousing establishment (NAICS 493110) run by five operators that have announced automated sites.
Done looks like
At least 300 establishments, each with a sourced equipment type and injury rates per 100 full-time workers for each year filed, and a 20% sample rechecked by a second person with at least 95% agreement.