Industries / Consumer goods / Candles and home fragrance
Safety testing
Safety testing is burning candles and running scent products under set conditions to check that they do not start fires, break their containers or put too much soot and gas into indoor air, and tracking the fires and recalls that show where products fail.
5
Milestones achieved, from 2004 to 2017.
2
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2004
A fire test for candles
In 1997 staff of the US Consumer Product Safety Commission (CPSC) asked ASTM subcommittee F15.45 to write candle standards. In 2002 they sorted 593 investigated candle incidents from 1998 to 2001 into patterns such as "candle flared up", "container shattered" and "candle reignited", so the committee knew what to test for. ASTM published a provisional standard in 2002 and finalised it in 2004 as F2417. It sets requirements for flame height, secondary ignition (a second flame starting on the wax or a decoration), behaviour at the end of the candle's life and stability, checked over repeated four-hour burns, or eight hours for gel candles. Sister standards cover warning labels (F2058), glass jars (F2179) and holders (F2601, 2007). The current edition, F2417-26, was issued on 15 June 2026.
LessonThe test was built from real incidents sorted by what went wrong.
Credit
ASTM subcommittee F15.45; CPSC staff, among them Susan B. Kyle, Allyson Tenney and Scott Ayers
2007
Europe tests fire and soot
The European standards body CEN published EN 15493 for fire safety and EN 15426 for sooting, with EN 15494 for safety labels (see Packaging and labels), prepared by its committee CEN/TC 369 on candle fire safety. The soot test catches soot on a glass plate held over the flame and measures how much light the dirty plate then blocks, giving a "soot index". Its 2018 edition sets one pass mark: across three candles, the average soot index must be below 1.0 per hour. It covers only single-wick candles up to 100 mm across.
Credit
CEN/TC 369
2014
US keeps standards voluntary
On 17 September 2014 the CPSC denied a 2004 petition from the National Association of State Fire Marshals to make candle standards mandatory. Its staff cited industry estimates that 90% of US candle consumption already met the ASTM standards, and a fall in fires: by CPSC estimates, candle fires attended by fire departments rose from 5,400 in 1990 to 15,900 in 2001, then fell to 6,600 in 2011.
LessonStaff said the drop could not be credited to the standards alone, because US candle consumption had also fallen about 35% since 2004.
Credit
CPSC staff (project manager Scott Ayers); National Association of State Fire Marshals (petitioner)
2015
EU emission rules shelved
A draft European Commission decision, sent to the Council on 12 November 2015, would have set what European candle standards must cover: "no visible release of soot" and a soot index of no more than 1.0 per hour, gases kept within WHO indoor air guidelines, and lead and nickel capped at 0.4 and 0.5 mg/kg. The draft noted that the EN candle standards were not referenced in the EU's Official Journal, so meeting them gives no legal presumption of safety. Norway's EEA law database records that the Commission withdrew the proposal.
LessonEurope has a soot test but no binding emission limit written for candles.
Credit
European Commission (draft); Europalov, Norway (record of withdrawal)
2017
Draughts drive candle soot
A Danish Environmental Protection Agency survey (No. 157, April 2017), by DHI and the Danish Technological Institute, modelled home exposure from candles. In a heavy-use case where two of four candles sooted in a draught, the yearly level of PM2.5, particles under 2.5 micrometres, went over the WHO annual guideline of 0.010 mg/m³ "by up to 28 times". In 2021 a Lund University and Aarhus University team burned five types of pillar candle in a 21.6 m³ chamber with fluctuating air, which they call stressed burning, and found the particle mass "dominated by soot", with the wax and wick strongly changing soot, PM2.5 and PAHs (polycyclic aromatic hydrocarbons, a family of combustion pollutants).
LessonA calm test burn understates what people breathe in a draughty room.
Credit
Danish EPA, DHI and Danish Technological Institute; Lund University (Joakim Pagels, Aneta Wierzbicka) and Aarhus University (Merete Bilde, Marianne Glasius)
Watch for
Camera records of burn tests
A US patent application for a "Candle Analysis System" describes a camera that stores video of a burning candle, with flame height, melt pool size, burn time and room and jar temperatures logged to a database and tests sorted into passed and failed; it says burn testing is now done by hand on record sheets. The record shows a final rejection in June 2026 and the application still pending.
Achieved whena system like it is in use at a maker or test lab and its readings match an accredited lab's manual readings on the same candles.
Claimed by
Fontana Candle Co LLC, Lancaster, Pennsylvania (inventors Eric and Katie Roering), May 2025
New container flammability tests
EN 15493:2026, valid from 1 October 2026 in Estonia's national edition, adds requirements and test methods for whether a container candle's container can catch fire, and for candles with several wicks. ASTM issued F2417-26 in June 2026; its changes are not public. CPSC staff's notes of the January 2026 F15.45 meeting record that the group is drafting a standard for candle-making kits, and will put indoor air quality on its September agenda after a December 2025 report of a 15-month-old child in hospital with raised carbon monoxide blamed on a candle.
Achieved whenaccredited labs report results against the new clauses, and recalls or EU Safety Gate alerts start citing them.
Claimed by
CEN/TC 369 and ASTM F15.45, January to October 2026
Open
01
Open burn-test rig, lab-checked
A bench rig that burns a consumer candle in a draught-free box, inside a ventilated room or fume hood and always attended, and records what F2417 and EN 15493 testers now note by hand: flame height several times a second from a camera with a millimetre scale in view, flame flicker, burn rate in grams per hour from a balance under the candle, jar temperature and elapsed time. It publishes the video and data for every run.
The briefCredit · unclaimed02
Public ledger of candle failures
An open, machine-readable table of every candle and candle-accessory recall and alert from the US CPSC, the EU Safety Gate, Health Canada, Australia's ACCC and the UK Office for Product Safety and Standards, each coded by failure mode (jar cracked, flame too high, flare-up, wick relights, tips over, holder caught fire and others) and matched to the standard clause that should have caught it, refreshed monthly.
The briefCredit · unclaimed03
Low-cost soot index reader
A cheap reader for the EN 15426 soot test: put each sooted glass plate on an even light source, measure how much light it blocks with a phone camera or a light sensor, and compute the soot index. Check it against an accredited lab's readings of the same plates, and publish the protocol and plate images so small makers can screen candles before paying for lab tests.
The briefCredit · unclaimed
Open 01 · the brief
Open burn-test rig, lab-checked
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A bench rig that burns a consumer candle in a draught-free box, inside a ventilated room or fume hood and always attended, and records what F2417 and EN 15493 testers now note by hand: flame height several times a second from a camera with a millimetre scale in view, flame flicker, burn rate in grams per hour from a balance under the candle, jar temperature and elapsed time. It publishes the video and data for every run.
Why it is still open
A US patent granted in February 2024 states that "candle-burn testing is performed by human visual observation". The DEKRA tester who checks candles for Germany's RAL quality mark notes his observations each hour when an alarm clock rings, and judges plate distance "with a quick look". EN 15426 lists a stop watch, a rule and a 0.1 g balance as its equipment, and Intertek's candle testing page mentions no cameras. Automated testers exist only as patents (Bruce Campbell, 2024; Liaoning Kaimai Petrochemical, deemed withdrawn in 2025; Fontana Candle, pending), none with published data or a check against an accredited lab.
Built with what exists
A Raspberry Pi 5 with a Raspberry Pi Camera Module 3 or HQ Camera; OpenCV to find the flame and measure its height against a printed scale; a lab balance with 0.1 g resolution and a USB or RS-232 output, such as an Ohaus Scout, or a 5 kg load cell with an HX711 amplifier; K-type thermocouples on MAX31855 boards for jar temperature; a hot-wire anemometer to confirm still air; a non-combustible tray and a fire extinguisher.
First step
Buy two identical jar candles, set a camera and a balance in a draught-free box in a ventilated room, and burn one attended four-hour cycle while a person reads flame height with a steel rule each hour, then compare the person's numbers with the video's.
Done looks like
10 candle models, 3 samples each, burned through full four-hour cycles in the F2417 style; the rig's flame heights within 3 mm of an accredited lab's readings on samples from the same lot, burn rate within 5% of the lab's weighed value, and pass or fail calls matching the lab on all 30 samples; every video and data file posted on Zenodo with a DOI.
Open 02 · the brief
Public ledger of candle failures
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open, machine-readable table of every candle and candle-accessory recall and alert from the US CPSC, the EU Safety Gate, Health Canada, Australia's ACCC and the UK Office for Product Safety and Standards, each coded by failure mode (jar cracked, flame too high, flare-up, wick relights, tips over, holder caught fire and others) and matched to the standard clause that should have caught it, refreshed monthly.
Why it is still open
ASTM F15.45 reviews recent recalls and incident reports each quarter (24 in-scope incidents in January 2026), but files only a short log that says "Contact ASTM for more details". A 2021 law firm review counted "at least 115 recalls" since 2001 and published no data. The CPSC's last published sorting of incidents by pattern is from 2002. The raw data are public: the CPSC recall database returns 187 records for "candle", and Safety Gate publishes weekly reports back to 2005.
Built with what exists
The CPSC recalls web service (saferproducts.gov/RestWebServices/Recall); Safety Gate weekly report XML; Health Canada and ACCC recall listings; a codebook built from the 2002 CPSC patterns and the F15.45 meeting logs; a small low-cost AI model such as Meta's Muse Ember to tag the free-text hazard descriptions (see the Note "The jobs agents repeat"); GitHub Actions for the monthly refresh; CSV files with a Zenodo DOI.
First step
Pull the CPSC records for "candle", drop fireworks and LED items, and hand-code the candle recalls from 2023 to 2026 to draft the codebook.
Done looks like
Every candle recall in the CPSC database and every Safety Gate candle alert since 2005 coded; two people coding the same random 50 records agree on at least 90%; the model matches their agreed codes on at least 90% of a further 100; six monthly refreshes published in a row.
Open 03 · the brief
Low-cost soot index reader
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A cheap reader for the EN 15426 soot test: put each sooted glass plate on an even light source, measure how much light it blocks with a phone camera or a light sensor, and compute the soot index. Check it against an accredited lab's readings of the same plates, and publish the protocol and plate images so small makers can screen candles before paying for lab tests.
Why it is still open
EN 15426 names one commercial measuring unit, the RMG 2.1 from Heil Metalle GmbH, and allows "equivalent products" if they "lead to the same results", but searches found no low-cost or open reader shown to match it. Low-cost particle sensors have already been compared with reference instruments on candle smoke (Singer and Delp, 2018; Kloppenborg and colleagues, 2026), so plain sensor calibration is not open, and the 2018 study found consumer monitors "substantially under-reported or missed" events dominated by particles smaller than 0.3 µm.
Built with what exists
Heat-resistant glass plates of 100 by 100 mm and 3.5 to 4.5 mm thick, and the wire-mesh cylinder EN 15426 specifies; an LED light panel; a smartphone shooting RAW images, a Raspberry Pi HQ Camera or an ams TSL2591 light sensor; neutral-density filters of known optical density, such as Thorlabs NE-series, for calibration; Python with rawpy and NumPy.
First step
Photograph a set of neutral-density filters on an LED panel with a phone in RAW and check that the computed light blocking tracks each filter's stated optical density.
Done looks like
20 plates from at least 8 candle models read by both the low-cost reader and an accredited EN 15426 lab, with the soot index within 0.1 per hour of the lab value on every plate and the same pass or fail call against the 1.0 per hour limit; protocol and images published with a DOI.