Optical flame detection covers six mainstream technologies: UV (185–260 nm), single-IR (typically 4.4 µm CO₂ band), dual-IR, combined UV/IR, triple-IR (IR3), and visual flame imaging; each maps to a specific hydrocarbon, hydrogen, or metal-fire duty [S3].
Selection is governed by three orthogonal gates: spectral match to the expected fire signature, the hazardous-area classification per ATEX 2014/34/EU or IECEx, and a third-party performance class (FM 3260, EN 54-10) that defines field of view and false-alarm rejection [S3]. A detector that passes the spectral gate but fails the area-classification gate will not be approved for the plot, regardless of detector range.
Spectral technologies and the fuels they actually see
UV tubes (185–260 nm) respond to OH* radical emission from hydrogen, methanol, ammonia, and metal fires such as magnesium, but are blinded by oil films, welding flash, and most hydrocarbons above 0.5 m flame path [S3].
Single-IR (4.4 µm) targets the CO₂ resonance band and works for hydrocarbon fires, but a flickering sunlight or rotating heater background can push it across the alarm threshold if no flicker discrimination is applied [S3]. Dual-IR adds a reference wavelength (commonly 4.0 µm or 4.8 µm) to subtract ambient IR, and is the baseline spec for offshore drilling modules and gas-turbine enclosures.
Triple-IR (IR3) sensors take ratios across three IR bands (typically 4.4 µm plus two reference bands near 2.9 µm and 4.8 µm) and tolerate arc-welding, sunlight, and hot CO₂ plumes at the cost of slower response on small (~0.1 m²) hydrogen flames [S3]. UV/IR combines a UV phototube with one IR band; it is the standard pick for aircraft hangars, where JP-8 and hydraulic-fluid fires coexist with halogen-lamp interference.
Visual flame imaging (CCD/CMOS + IR video) analyses flame shape, motion, and flicker frequency in real time, and is specified where the plant needs a video record of the event for incident review, not only an alarm contact [S3].
Performance classes, false-alarm rejection, and detector range
EN 54-10 classes flame detectors into Class 1 (most sensitive, optical density ≥ 0.4 dB/m, fire size 0.1 m² n-heptane at 25 m), Class 2 (≥ 0.45 dB/m at 17 m), and Class 3 (≥ 0.5 dB/m at 12 m) for the reference n-heptane fuel; FM 3260 carries a similar sensitivity grading and adds arc-welding and sunlight rejection tests that most IR3 and visual-imaging units quote explicitly [S3].
Detector range is fuel-dependent: a single UV/IR unit rated 23 m on n-heptane typically falls to 9–12 m on ethanol and 4–6 m on hydrogen, so the line-of-sight layout must be drawn against the actual fuel, not the headline hydrocarbon number [S3]. For IR3 detectors, false-alarm rejection is most often quoted against a rotating 1 kW halogen lamp at 1 m and 100 Hz fluorescent flicker, both of which the unit must ignore while still alarming on a 0.1 m² n-heptane pan fire at the rated range.
Response time is another numeric gate: hydrocarbon fires in process modules are specified at 5–10 s to alarm at full range, while refinery EN 54-10 Class 1 bids often cap the time-to-alarm at 30 s including signal verification [S3].
Comparison: UV, single-IR, dual-IR, IR3, UV/IR, visual imaging
Selection logic lines up roughly as follows. UV detectors win for hydrogen, ammonia, and metal fires, but lose on oily hydrocarbon and sunlight exposure. Single-IR is the cheapest, smallest, and is fine indoors with no welding or rotating heaters. Dual-IR is the workhorse for offshore gas turbines, pipeline compressor sheds, and FPSO modules, with FM 3260 sunlight rejection at a moderate cost step. IR3 trades a slower hydrogen response for the best welding/sunlight immunity and is the go-to for tank-farm perimeter detection and LNG loading arms. UV/IR is the standard for aircraft hangars, paint spray booths, and offshore drilling floors where halogen lamps and hot exhaust coexist. Visual flame imaging is the most expensive, the slowest on a per-pixel basis, but the only option when a video record of the alarm event is a contractual requirement. [S2]
A buyer can shortlist by answering three numeric questions: what is the worst-case fuel and the smallest credible fire size in m², what is the max line-of-sight distance to that fire, and what is the false-alarm source population (welding, halogen, sunlight, rotating heaters) at the same site. If the answer to the third question is empty, a single-IR or dual-IR unit is usually acceptable; if it contains welding plus sunlight, jump to IR3 or visual imaging [S3].
Who should NOT pick the cheapest single-IR option
Single-IR (4.4 µm only) is a poor choice for any open-frame process module with arc-welding nearby, any outdoor tank-farm detector exposed to direct sun at low elevation angles, and any LNG or hydrogen service where the fire signature is faint or non-hydrocarbon; the false-alarm and miss-rate trade-off is not acceptable on a firewater deluge interlock [S3].
UV-only units similarly fail inside engine test cells, paint booths with hydrocarbon aerosols, and any location where the lens will be exposed to oil mist; UV tubes also have a 2- to 5-year service life depending on duty cycle, and the maintenance window is shorter than for solid-state IR or visual imaging [S3].
Area classification, ingress, and the audit paper trail
Hazardous-area approval is non-negotiable: ATEX 2014/34/EU with II 2 G Ex db eb IIC T4 Gb is the typical Zone 1 label for refinery and offshore service, and IECEx equipment certificates are increasingly demanded by EPCs operating in multiple jurisdictions [S3].
Ingress protection at IP66/IP67 is the de-facto minimum for outdoor flame heads; for offshore modules in green-water zones, IP66 plus a stainless-steel 316L housing and a sunshade is the standard. Cable entries must be Ex d or Ex e certified glands; a wrong gland is a common audit finding during IEC 60079-14 inspections and invalidates the Ex rating regardless of detector performance.
Functional safety must also be addressed: a flame detector that feeds a deluge valve or ESD-1 shutdown should be SIL 2 capable per IEC 61508, with a documented failure rate (SFF, PFD) on the certificate, not just a vendor declaration [S3].
Use cases drawn from the warehouse and ASRS fire-safety literature
High-bay automated storage and retrieval system (AS/RS) warehouses are a special case: ceilings commonly exceed 24 m, and a ceiling-mounted point detector is the wrong tool, because smoke and flame plumes are dragged horizontally by the stacker-crane air movement before reaching a beam detector [S3].
The standard remediation is a network of UV/IR or IR3 flame detectors mounted on the stacker crane itself, giving line-of-sight to the pallet face the crane is currently servicing, paired with aspirating smoke detection at the rack-top level. A flame photometric detector, dual-channel, measures the light emission intensity at two wavelengths simultaneously to discriminate flame from background; the dual-channel flame photometric detector principle is the same one that underpins laboratory gas-chromatography FPD modules, scaled to flame-detection wavelengths [S1].
For paint-shop robots, IR3 with a narrow 30° cone and a 6 s to alarm limit on a 0.1 m² solvent pan is the conventional spec; for offshore drilling floors, UV/IR with a 90° cone and FM 3260 welding immunity is more common. The selection is rarely about the most expensive sensor; it is about the one whose failure mode the plant can tolerate [S3].
Limitations, failure modes, and maintenance windows
Every optical flame detector is a line-of-sight device; obstructions, lens fouling, and condensation will mute the signal and the detector will fail silently if there is no periodic optical-integrity self-test. A unit with a built-in through-the-lens self-test (often called an OPTI-CHECK or OIS) issues a fault contact when the received signal drops below the calibration baseline, and is the only acceptable choice for un-manned sites [S3].
Window cleaning intervals depend on atmosphere: an offshore deck detector may need a weekly wipe, while an indoor control-room detector can run 12 months. Spare-part holding should cover at minimum one UV tube (typical 2–5 year life) and one IR sensor module per 20 units in service.
False alarms remain the dominant operational cost; the bid evaluation should weight the documented false-alarm rate per 1000 detector-hours, not the headline range figure, because a 25 m range with one false alarm per week will be disarmed by operations within a quarter [S3].
Shortlist logic and next procurement step
A defensible 2026 procurement path is: (1) lock the fuel list and smallest credible fire size, (2) compute line-of-sight coverage with a 1.5× overlap factor, (3) eliminate any technology whose spectral response misses the fuel, (4) apply the area-classification and SIL 2 gate, (5) compare the surviving two or three technologies on false-alarm rate and total cost of ownership over a 10-year window, and (6) require an EN 54-10 or FM 3260 test certificate dated within 36 months of bid opening, because sensor drift on the test bench is a real issue on long-storage stock [S3].
Trackable signals for the next review: any change in IEC 60079-0 equipment-classification guidance for optical heads, any new FM 3260 welding-immunity test threshold, and any site incident data the operator will share post-commissioning. A flame detector specification that survives this gate is the one a process engineer can sign without a redline. A wider industrial selection guide approach, applied to the fire-protection discipline, also drives more rigorous embedded part selection for commercial buildings when the same project team is responsible for both the structural and the life-safety packages. The closest process analogue, where spectral band choice and false-alarm rejection are weighed against range and cost, is detailed in our dust detector, gas detector, and heat detector reference pages.