Industries / Construction and property / Fire protection and life safety
Alarms and detection
Alarms and detection is the work of sensing a fire early, by its smoke, heat or gases, then warning the people inside and calling the fire service.
7
Milestones achieved, from 1852 to 2024.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1852
Street boxes signal a fire
On 28 April 1852 Boston put into service the first municipal electric fire alarm system that used call boxes signaling automatically to show where a fire was, according to the plaque the IEEE dedicated in 2004. The original system had 40 street boxes, three box circuits, three bell circuits and 16 alarm bells, wired to a central office, and the first alarm came in at 8:25 p.m. on 29 April 1852. It was later adopted across the United States and Canada, and Boston still runs about 1,500 street boxes plus more than 1,000 "master boxes" that connect buildings' own alarm systems to the city.
Credit
William F. Channing, who invented it, and Moses G. Farmer, who helped build it and became the first superintendent of fire alarm telegraph
1941
Ionization chambers sense smoke
An ionization detector uses a small radioactive source, typically americium, to let the air in a chamber carry a tiny electric current, and smoke entering the chamber reduces that current. The Swiss physicist Walter Jaeger had been trying since the mid-1930s to build such chambers to detect poison gas; by the account of Siemens, which now owns his company, the devices barely reacted to the gases but reacted reliably to smoke. Jaeger published "The Ionization Chamber as a Fire Alarm" in a Swiss trade journal in 1940 and founded Cerberus with the physicist Ernst Meili at Bad Ragaz in 1941. Siemens says the first detector, the F1, was introduced in 1941 and sold from 1943, that only a few hundred were made because of teething problems, and that the F5 of 1959 was the first ionization detector made in the millions.
Credit
Walter Jaeger and Ernst Meili of Cerberus, by the account of Siemens, which bought Cerberus in 1998
1975
Smoke alarms run on batteries
In 1975 Statitrol, a company founded by Duane D. Pearsall, began mass production of the SmokeGard 700, which IEEE Spectrum calls "widely regarded as the first practical battery-operated ionization-type smoke detector." A battery alarm needs no wiring, so it can go into any existing home. The US Atomic Energy Commission had approved smoke detectors for homes in 1969, and in 1974 the National Fire Protection Association (NFPA) revised NFPA 74, its standard for household alarms, to support battery models. BRK Electronics says a model it sold in 1969 was the first such product approved by Underwriters Laboratories, and it launched its First Alert battery alarm in 1976. IEEE Spectrum reports that just under 10% of US homes had a smoke detector at the end of 1975 and 20% a year later.
Credit
Duane D. Pearsall and Statitrol; BRK Electronics
1976
Test fires in real houses
In 1975 and 1976 the National Bureau of Standards, now NIST, sponsored Underwriters Laboratories and the IIT Research Institute to set test fires, with real furniture, in houses due for demolition in what would become Indiana Dunes National Park. NIST later called these "Indiana Dunes tests" a comprehensive, independent set of tests that "clearly demonstrated the potential of smoke alarms to save lives," and says they led to NFPA 74, the standard for installing home alarms that is now part of NFPA 72. By NFPA and US Fire Administration estimates, the share of US homes with a smoke alarm rose from under 10% in 1975 to at least 95% in 2000, while home fire deaths were cut nearly in half.
Credit
National Bureau of Standards; Underwriters Laboratories; IIT Research Institute
2008
Fires now grow faster
NIST repeated the Dunes method in a manufactured home and a two-story house, with alarms bought from stores, and published the results as Technical Note 1455. Both common alarm types gave occupants some escape time in most fires: ionization alarms reacted sooner to flaming fires, and photoelectric alarms, which sense smoke by the light it scatters, reacted sooner to smoldering ones. But the average time until conditions became unsurvivable in flaming tests was 3 minutes, against 17 minutes in 1975. NIST says part of the gap comes from different survival criteria, and that fires in the new tests clearly grew faster. It also found ionization alarms prone to sound in the early stages of some cooking, before smoke was noticeable.
LessonThe time an alarm buys depends on how fast fires grow, and in these tests that time had fallen sharply since the 1970s.
Credit
NIST, with funding from four federal agencies and Underwriters Laboratories
2023
Scotland stops answering automatic alarms
From 1 July 2023 the Scottish Fire and Rescue Service no longer sends crews to automatic fire alarm signals from workplace premises unless someone confirms a fire or signs of one; hospitals, care homes, hotels and homes still get an immediate response. It had been sending crews to about 28,000 false alarms a year, almost a third of its activity, and it reported 52 fewer needless call-outs a day over the first nine months. False alarms are a large share of the work elsewhere too: in England in the year to March 2025, fire services attended 250,226 fire false alarms, 41% of all incidents, and 176,262 of them came from fire alarms or firefighting equipment going off in error, including when set off by accident, by the government's count.
LessonWhen nearly all automatic signals turn out to be false, fire services stop treating them as emergencies, and checking each alarm becomes the building's own job.
Credit
Scottish Fire and Rescue Service; England's Ministry of Housing, Communities and Local Government (statistics)
2024
Alarms must ignore hamburgers
On 30 June 2024 the eighth edition of UL 217, the US standard for smoke alarms, took effect, NIST says. It requires alarms to respond better to smoldering and flaming fires, including new tests with burning polyurethane foam, and for the first time includes a test against nuisance alarms: two standard hamburger patties are broiled 3 m (10 ft) away, and the alarm must stay silent. In 2016 NIST tested 45 alarms of the previous generation and none could pass the new standard. In a kitchen laboratory in 2024, NIST researchers Amy Mensch, Emma Veley and Thomas Cleary cooked a range of foods under 12 alarms and found the new alarms performed about the same as older ones at staying quiet during cooking.
LessonA single test food does not stand for all cooking, and people who suffer repeated nuisance alarms tend to disable their alarms.
Credit
UL (the standard); NIST (the tests)
Watch for
Detectors that ignore steam
Siemens launched its Sinteso Nova and Cerberus Nova detectors on 27 March 2026 and says their ASAplus method, which combines optical sensing at several wavelengths of light with two heat sensors, reduces false alarms and unnecessary evacuations. On 10 November 2026 Honeywell introduced System Sensor Distinguish, an AI-enabled detector it says classifies its sensor readings to tell real fire from steam, dust, mist and vapor while keeping early smoke detection. Neither company has published a false-alarm figure.
Achieved whenan independent laboratory test, or a published field record from buildings that installed the detectors, shows fewer nuisance alarms with no slower response to test fires.
Claimed by
Siemens, March 2026; Honeywell, November 2026
Open
01
Nuisance alarms in real kitchens
A field study, and an open dataset, of how often smoke alarms built to the eighth edition of UL 217 sound during ordinary cooking in occupied homes. In each volunteer home, mount extra alarms of each sensing type (ionization, photoelectric, dual-sensor) at 1.5 m, 3 m and 6 m (5, 10 and 20 ft) from the stove for 30 days, alongside the home's own alarms, which stay in place. A small logger records the time each test alarm sounds, a particle sensor records the smoke and grease aerosol next to it, and residents log what they cooked and how.
The briefCredit · unclaimed02
Fires after skipped alarm calls
A public analysis of whether fires in workplaces and other non-domestic buildings were found larger, or reported later, after fire services stopped sending crews to every automatic alarm. First a sourced table of which fire services in England changed their response to automatic alarms, and when; then a before-and-after comparison against services that did not change, using England's incident-level fire records.
The briefCredit · unclaimed
Open 01 · the brief
Nuisance alarms in real kitchens
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A field study, and an open dataset, of how often smoke alarms built to the eighth edition of UL 217 sound during ordinary cooking in occupied homes. In each volunteer home, mount extra alarms of each sensing type (ionization, photoelectric, dual-sensor) at 1.5 m, 3 m and 6 m (5, 10 and 20 ft) from the stove for 30 days, alongside the home's own alarms, which stay in place. A small logger records the time each test alarm sounds, a particle sensor records the smoke and grease aerosol next to it, and residents log what they cooked and how.
Why it is still open
The only field study found is a 2008 pilot by US Consumer Product Safety Commission (CPSC) staff in nine Washington, DC homes of staff and their acquaintances, using alarms of that time; at 3 m from the stove nuisance alarms came with 4.9% of cooking events for ionization alarms, 8.7% for dual-sensor and 6.0% for photoelectric. NIST's 2024 tests of alarms built to the new standard were done in a laboratory kitchen, and found them no better than older alarms at staying quiet during cooking. Searches found no field test of alarms built to the eighth edition.
Built with what exists
Retail battery smoke alarms listed to UL 217 eighth edition or later; a microphone on an ESP32 or Raspberry Pi board, or a commercial sound-level logger, to timestamp each alarm's beeping; a Sensirion SPS30 particle sensor; a paper or phone cooking log; Python or R for rates and confidence intervals.
First step
Mount one alarm of each type 3 m from your own stove next to an SPS30 sensor and a sound logger, and log a month of your own cooking to test the setup.
Done looks like
At least 30 homes and 1,000 logged cooking events, nuisance-alarm rates by sensing type and distance with confidence intervals, set beside the CPSC's 2008 rates, and the event table and sensor traces published under an open license.
Open 02 · the brief
Fires after skipped alarm calls
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public analysis of whether fires in workplaces and other non-domestic buildings were found larger, or reported later, after fire services stopped sending crews to every automatic alarm. First a sourced table of which fire services in England changed their response to automatic alarms, and when; then a before-and-after comparison against services that did not change, using England's incident-level fire records.
Why it is still open
The Scottish Fire and Rescue Service reported how many call-outs its 2023 change saved, but gave no figures on fires; a search found no evaluation of fire size or damage since the change. London's fire commissioner, deciding in March 2024 to stop attending most alarms from commercial buildings by day, noted that at least 42 other UK fire services already dealt with such alarms partly by not attending. Its report looked back over five years of London's own alarm-raised fires and judged the added risk low, while warning that the policy "may have resulted in more developed fires in some cases"; it was a forecast, not a measurement afterwards. England already publishes a record for each primary fire in a building other than a home, with fields for the fire service, year, whether an alarm was present and raised the alarm, the time from discovery to the call, a late-call flag, fire size on arrival and the extent of damage.
Built with what exists
England's "Other building fires" and "False alarms" incident-level datasets (ODS files under the Open Government Licence); each fire service's published policy on automatic alarms, gathered by an AI research agent that keeps a link behind every date it reports (see the Note "Findings that cite notes"); Python with pandas, or R, for a difference-in-differences comparison, which measures the change in services that switched against the change in those that did not.
First step
Download the two datasets and list the ten fire services whose false alarms from automatic systems in non-domestic buildings fell most between 2015–16 and 2024–25; those are the likeliest to have changed policy.
Done looks like
A table of at least 20 English fire services with a sourced policy and start date for each; estimates, with confidence intervals, of the change in fire size on arrival and in the share of late calls for non-domestic fires; code and table published, and the method checked by one fire statistician or fire service analyst.