Flame Detection: UV, IR, and Multi-Spectrum Sensors

A flame detector watches for the optical signature of combustion rather than for smoke or heat. It looks at ultraviolet emission, infrared emission, or both, and uses spectral analysis to distinguish a real fire from sunlight, hot machinery, welding arcs, and reflective surfaces. Flame detectors dominate in spaces where smoke detection is impractical and heat detection is too late: open process areas, hazardous-area plant, hangars, fuel storage, turbine halls, and warehouses with rapid-developing flammable inventory. Used correctly, flame detection gives sub-second alarm at flaming-stage fires; used badly, it false-alarms on welding flashes and arc spikes.

This article covers the optical principle, the UV and IR bands, multi-spectrum analysis, hydrocarbon and hydrogen flames, field of view, the typical spaces where flame detection wins, and the false-alarm sources that have to be designed out rather than tuned out.

What a flame emits

A flaming combustion process emits broadband optical radiation across the ultraviolet, visible, and infrared spectra. The exact spectrum depends on the fuel: hydrocarbons emit a strong CO2 emission feature near 4.3 micrometres in the IR, alcohols emit a weaker feature, and hydrogen flames emit almost no IR but a strong UV component. Flame detectors are designed to discriminate these emission features from background optical sources by combining wavelength-selective detection with time-domain analysis of how the emission flickers.

A real flame flickers at characteristic frequencies, typically in the 1 to 25 Hz range, because turbulent combustion produces a pulsating flame envelope. A welding arc, by contrast, has a distinctive higher-frequency signature, and sunlight reflected off moving water has yet another. Modern flame detectors classify the optical signal in both spectrum and time and discriminate between fire and not-fire on that combined signature.

UV detectors

An ultraviolet flame detector responds to UV emission in the 185 to 260 nanometre band. UV is strongly absorbed by ordinary glass and by atmospheric water vapour over a few metres, so UV detectors have a relatively short effective range and are largely confined to enclosed spaces with line of sight to the protected area. Their main advantage is sensitivity to flame types that emit strongly in UV, including hydrogen and certain alcohols, where pure-IR detectors struggle.

UV detectors false-alarm on sources that share the UV signature of flame: welding arcs, lightning, certain X-ray sources, mercury vapour lamps, and corona discharge from high-voltage equipment. Each of those is real-world enough that UV-only detectors are now uncommon outside specific niche applications, replaced by multi-spectrum heads that include UV but cross-check it.

IR detectors

Infrared flame detectors look at one or several IR bands. Single-band IR detectors are simple and cheap but easy to fool. Two-band IR detectors cross-check by comparing emission in two wavelengths, typically a flame-rich band against a flame-poor band, and discriminate fire from sunlight or hot pipework on the ratio. Three-band IR adds a third reference band and pushes the discrimination further.

Three-band IR is the dominant general-purpose flame detection technology for hydrocarbon environments because it gives reliable detection of pool fires, jet fires, and spill fires with very low false-alarm rates against blackbody radiation, sunlight, and reflective surfaces. The CO2 emission feature is the marker; anything emitting at that wavelength while flickering at the right frequency is treated as a flame.

Multi-spectrum and combined detectors

Modern flame detectors increasingly combine UV and IR detection in a single head and run cross-spectrum logic to confirm an alarm. A UV-IR detector that requires both bands to agree before alarming is far less sensitive to false alarms from any single source: a welding arc trips the UV channel but not the IR, and is rejected; a sunbeam reflected off water trips an IR channel but not the UV. The cross-check is what makes the detector usable in real plant environments.

Each major manufacturer has its own algorithm and its own claimed range for each fire type. The cluster article on the principle covers the spectrum logic in more depth. Performance claims should be cross-checked against independent test data because manufacturer ranges typically assume best-case fire types and weather conditions.

Hydrocarbon, hydrogen, and other flame types

Different flame types favour different detector technologies. Hydrocarbon flames are best handled by IR multi-band detectors that lock onto the CO2 emission feature. Hydrogen flames are nearly invisible in IR, with weak CO2 emission because hydrogen combustion produces water rather than CO2, and demand UV detection or specialised hydrogen-specific multi-spectrum heads. Methanol and ethanol flames are intermediate.

For mixed-fuel environments, the choice is often a multi-spectrum detector that handles the dominant fuel and supplements it with a hydrogen-specific detector if hydrogen is present. Hydrogen storage and fuel-cell installations need detectors validated specifically for hydrogen flames; using a generic IR-only detector in a hydrogen environment is a design error.

Field of view and coverage

A flame detector has a defined field of view, typically a cone of around 90 to 120 degrees half-angle, and a defined detection range that varies with the fire type and size. Within the field of view, the detector responds; outside, it does not. The protected space has to be covered with overlapping fields of view to ensure that any fire of the design size in any location triggers at least one detector quickly.

The detection range depends strongly on fuel and fire size: a small alcohol fire detected at 10 metres might require half that range, while a large pool fire might be detected at 30 or 40 metres. Design uses the smallest fire size of operational concern as the basis for coverage, not the largest.

Mounting position is a discipline. Flame detectors are typically mounted high enough to see across the protected area but not so high that the field of view skims hot machinery or reflective surfaces. Tilt is set to favour the fire risk areas: the fuel storage area, the conveyor with the highest ignition risk, the loading bay door. Detection coverage logic applies as much to flame as to smoke.

Hazardous areas and explosion-protection

Many flame detection installations are in hazardous areas under ATEX or IECEx classification, where the detector itself must not be a source of ignition. Hazardous-area certified flame detectors carry the relevant certification, are typically housed in flameproof or intrinsically safe enclosures, and have specific cabling requirements. Substituting a non-certified detector for a certified one in a hazardous area is a serious compliance breach as well as a safety risk.

The relevant certification scheme depends on the jurisdiction and the gas group of the protected area. Hazardous-area design is a specialism in itself; for the purposes of this article it is enough to note that flame detection often coincides with hazardous areas and that the certification has to be checked at design stage rather than discovered at handover.

Where flame detection wins

Flame detection is the right answer in spaces where smoke detection is impractical and heat detection is too late. Open process plant in petrochemical and chemical sectors uses flame detection because smoke disperses too quickly outdoors for smoke detection to alarm reliably. Hangars and large warehouses with rapid-developing flammable inventory use flame detection because the time from ignition to large fire is too short for ceiling smoke detection. Turbine halls, fuel transfer areas, paint stores, and printing presses are common cases.

Flame detection is also used in conjunction with multi-sensor smoke detection in indoor industrial spaces where the early-warning smoke phase is short and the cost of a slow alarm is high.

Common pitfalls

The first pitfall is sun-facing mounting. A flame detector pointing at the rising or setting sun, even with multi-spectrum logic, is operating at the limit of its discrimination ability. Mounting orientation should be checked against the actual sun path on the site, not against a default pointing direction.

The second is welding interference. Hot work in the protected area produces transient UV signatures that can trigger a UV-component detector. The detection scheme should include a hot-work disable that integrates with the permit-to-work system, rather than being permanently desensitised to handle the worst case.

The third is contamination of the optical window. Dust, oil mist, and moisture on the detector window reduce sensitivity over time. Regular inspection and cleaning, on a schedule matched to the environment, is essential. Many detectors include automatic optical-integrity testing that checks the window transparency and signals fault if it drops below threshold; this feature should be enabled rather than disabled to silence nuisance faults.

The fourth is treating flame detection as a substitute for smoke detection. The two technologies cover different fire phases. A smouldering fire that never flames will not trigger a flame detector at all; a flaming fire that produces little smoke may not trigger a smoke detector for many seconds. The right answer in many spaces is both, with cause-and-effect logic that uses each appropriately.

What this article does not cover

This article does not give specific detection ranges, mounting heights, sensitivity classes, or hazardous-area zoning rules. Those depend on the standard, the certification scheme, and the specific detector chosen. The applicable standards include EN 54-10 in Europe, FM 3260 in North America, and the relevant IECEx certifications for hazardous areas.

Flame detection is a precision tool for spaces where flaming fires must be detected fast and where smoke and heat detection are not viable. The cluster article steps through the spectral analysis principle in more detail, and multi-sensor detection covers complementary technologies for the early stages.

Testing, certification, and end-of-life

Flame detector functional testing presents a specific challenge: most installations cannot accommodate a real flame test, and the detector responds to specific spectral signatures that are not produced by routine test sources. Manufacturers supply spectral test sources or test cards that emit the specific UV and IR wavelengths the detector responds to, allowing routine functional testing without lighting an actual fire.

The test source is presented to the detector at the rated test distance, and the detector should alarm within the specified time. The test verifies the optics, the electronics, and the alarm logic. It does not verify the absolute sensitivity threshold; that requires periodic factory recalibration or a more elaborate spectral test rig.

Certification for hazardous areas requires that the detector continues to meet its certification over its service life. Optical drift, electronic ageing, and case corrosion can each affect the detector's certified status. Service intervals for hazardous-area flame detectors include certification-relevant inspections that go beyond routine functional testing, with the inspection documented for the operator's records.

End-of-life is set by the manufacturer based on optical component ageing, particularly for UV detection where photomultiplier or sensor sensitivity degrades over time. Operating life of around ten to fifteen years is typical for current detector designs, with replacement on a planned cycle rather than on failure. Detector failure during a real fire is the worst possible end-of-life event; planned replacement avoids it.

Applied design rules, calculations, and worked examples for flame detection are covered in the courses on this site.