Flame Detector

What a Flame Detector Is and What Is Being Decided

Start with the split, because it decides every later paragraph. An area flame detector for fire and gas systems is a fixed, weatherproof optical device mounted on structure, aimed at a volume where a flammable release could pool or jet and ignite. It plays no role in firing control. It exists to shorten the time between ignition and executive action from minutes, which is what human discovery takes, to seconds. A burner flame scanner is a different product with shared physics: a sensor in a sighting tube aimed at the root of one burner flame, feeding a flame monitor whose relay logic proves flame at the ignitor and main flame stages and removes fuel when the flame fails, with a flame failure response engineered to the 1 second class that NFPA 85 and NFPA 86 practice expects for boiler and furnace duty.

What is actually being decided in this purchase is three commitments at once. First, a technology commitment: the sensing band must match the fuel, because a triple IR detector tuned to the carbon dioxide band is nearly blind to a hydrogen flame, and a UV detector beside a welding bay will alarm all shift. Second, a coverage commitment: cone of view, mounting position, and detector count must produce a geometry in which the design fire size is detected at the actual distance, through the actual obstructions. Third, an integration commitment: outputs, voting, area certification, and SIL claims must fit the fire and gas logic solver that receives the signal. A quotation that answers only the first question is a price tag, not a protection design.

The boundary against building fire equipment must be drawn in writing, because the words overlap. A smoke detector and a heat detector protect enclosed rooms: they respond to combustion products and rising temperature at a ceiling or wall, per the EN 54 family of component standards, and they belong to the building fire alarm system. An optical flame detector watches open process areas where a plume may never reach a ceiling sensor and where the required response is seconds. The passive neighbors are further away still: a flame arrester is a quench element inside a vent line or tank nozzle that stops a flame front propagating, and it detects nothing at all; a safety relief valve manages overpressure, which is a different layer of protection with different physics. None of these devices proves that a burner is lit, and none substitutes for a flame scanner in a burner management system.

Gas detection is the layer that works before ignition, and the two systems are usually engineered together. A fixed gas detector reads the concentration of a specific gas continuously at a mounting point; a combustible gas detector reports percent of the lower explosive limit for flammable vapors before any ignition source finds them; a toxic gas detector covers the personnel layer for hydrogen sulfide and similar hazards in ppm ranges. Portables extend the same coverage to confined space entry and hot work, which is where the handheld multi gas detector that maintenance crews carry belongs. Flame detection is the event layer after ignition. Specifying all three as one fire and gas scope, with honest coverage maps for each, is what a modern facility layout actually buys.

Four families of standards organize the datasheet, and a buyer should quote them by name. NFPA 85 governs boiler and combustion systems hazards, including the burner management logic that consumes scanner signals; NFPA 86 governs ovens and furnaces; API RP 556 covers instrumentation and controls for fired heaters, including flame scanning and viewing systems. For the area devices, FM 3260 fixes the performance approval testing of optical flame detectors and EN 54-10 is the European component standard for flame detectors; SIL capability for the whole loop is assessed under IEC 61511. A datasheet that cites none of these is not a datasheet; it is a feature list, and it should be returned to the sender.

The RFQ must therefore state the protected hazard in engineering terms, not adjectives: the fuel and its expected fire geometry, the design fire size to be detected, the distance and obstructions between the plausible fire and every proposed mounting point, the required response time, the area classification and ambient envelope, the voting architecture, and the protocol the logic solver speaks. The chapters that follow give the numbers and the failure modes that turn those statements into a specification a vendor can be held to.

How Optical Flame Sensing Works

The physics starts with what a flame radiates. Every hydrocarbon flame emits ultraviolet energy between roughly 185 and 260 nanometers, produced by excited OH and CH radicals in the reaction zone, and the atmosphere happens to be opaque to solar radiation in that band, which is why a sensor there is called solar blind: the detector can stare at open sky and see essentially no sunlight. The same flame radiates strongly in the infrared, dominated by hot carbon dioxide emission between 4.3 and 4.5 micrometers and by water vapor bands near 2.7 to 2.9 micrometers. These two windows, deep UV and the carbon dioxide band, are the workhorses of the whole product class, and every technology choice in chapter 3 is a decision about which window, or which combination, a service can trust.

Range follows the inverse square law, which is why range numbers must always be tied to a fire size. The irradiance reaching the sensor falls with the square of the distance: halve the distance and the signal quadruples, so a detector that reliably proves a 0.1 square meter heptane pan fire at 30 meters will prove a fire of roughly four times that radiant intensity, about 0.4 square meters, at around 60 meters. Vendors express this in type test tables, fire size against distance, and the honest procurement move is to demand that table rather than a single maximum range figure. The same law explains why a small change in mounting position often buys more detection capability than a change of model: moving a detector 10 meters closer to the hazard can double the received signal from a given fire.

Flames also flicker, and flicker carries identity. Open pool and jet flames modulate their radiation at a few hertz to a few tens of hertz, which area detectors use to reject steady hot sources. Burner practice pushes further and discriminates between flames on the same burner: the main flame of an oil or gas burner shows dominant modulation in the 10 to 40 Hz class, while a gas pilot shows a faster 60 to 120 Hz class signature, figures quoted as engineering classes rather than law. The flame monitor amplifier applies adjustable gain and a bandpass filter so the scanner can follow the pilot through light off and then track the main flame once it is established, ignoring the neighboring burner through the same sighting line because its flicker spectrum arrives differently attenuated.

The scanner chain then converts photons into proof. The sensor, a UV phototube or solid state device or an infrared cell depending on the model, sits at the back of a sighting tube aimed at the first third of the flame, because the root carries the strongest and most characteristic emission. The flame signal strength is read out as 0 to 10 V or 4 to 20 mA at the flame monitor, with an adjustable threshold that separates flame on from flame off. On loss of signal the monitor drops its flame relay inside the 1 second class response time, and the safety shutoff valves close. Because a device that silently fails would be worse than no device, the chain proves itself alive with a self check: a mechanical or electronic shutter blanks the sensor view on a schedule, at least every 24 hours during firing and on every shutdown per NFPA 85 practice, and the electronics must observe darkness; if the signal does not collapse when the shutter closes, the monitor declares a fault and locks out the next light off attempt.

Area detectors face a noisier world and answer with spectral logic. A single channel IR detector at 4.3 to 4.5 micrometers can be confused by any hot surface radiating in the band, so it modulates and thresholds the signal, and even then hot refractory and direct sun remain classic false sources. Dual IR adds a reference band outside the carbon dioxide emission and ratios the two, rejecting steady blackbody sources. Triple IR carries three infrared channels, typically two reference bands and one emission band, and votes among them, which is what gives the technology its long range and its false alarm immunity for hydrocarbon fires. Combined UV IR units demand both signatures inside a coincidence window of a few seconds, so an arc welding flash (strong UV, no carbon dioxide flicker) or a hot stack (infrared, no UV) fails the test and stays silent. Video based flame detection adds a camera and image algorithms to the infrared sensing, giving operators a picture to verify, at the cost of needing light and processing time.

Response time on the area side is deliberately slower than the scanner side, and the difference is a design decision, not a deficiency. A fire and gas detector runs an adjustable alarm delay, typically 3 to 10 seconds, long enough to ride through a flare pass, a lightning flash, or a passing hot exhaust, before declaring fire. Voting then trades speed against spurious trips: two out of three logic across overlapping fields of view will not trip on a single blinded or lying detector, while a one out of one arrangement trips fastest but forwards every false alarm to deluge and shutdown. The specification should state the voting, the delay, and the required detection time for the design fire as one coherent set, because a vendor quoting a 3 second laboratory response has said nothing about the installed behavior of the loop.

Sensor Technologies, Construction, and Duty Boundaries

UV detectors come in two sensor generations. The UV tube is a gas filled Geiger style tube that counts photon events; it is fast, cheap, and sensitive in the 185 to 260 nanometer band, with a finite tube life counted in tens of thousands of operating hours. Solid state UV sensors, typically silicon carbide photodiodes, remove the tube life limitation and tolerate vibration better. Both share the same behaviors a buyer must plan around: they are the fastest technology available, responding in milliseconds, and they are blind to sunlight, hot refractory, and most artificial lamps. They are equally famous for their enemies. Electric arcs, welding within line of sight, lightning, and grinding sparks all radiate UV and will alarm an unprotected channel. Ordinary glass blocks the deep UV band, so every window and lens in front of the sensor must be quartz, and an oil film, a water film, or steam in the sight path attenuates the signal severely.

The infrared families trade that fragility for a different set of compromises. A single channel IR detector on the 4.3 to 4.5 micrometer carbon dioxide band sees hydrocarbon flames through smoke and light lens fouling that would blind UV, but any hot surface in view is a candidate false source until modulation discrimination proves otherwise. Dual IR rejects most steady sources by ratioing the emission band against a reference band. Triple IR is the premium hydrocarbon technology: three channels voted together give the best immunity to sunlight, hot bodies, and modulated interference, plus the longest ranges on pool fires. Its documented weakness is fuel chemistry: hydrogen, methanol, ammonia, and sulfur flames emit almost nothing in the carbon dioxide band, so a triple IR detector tuned for hydrocarbons can stare through a hydrogen fire. Those fuels need UV sensing or a hydrogen specific infrared variant.

The combined UV IR detector is the refinery workhorse for open flame hazards because it demands coincidence: ultraviolet energy and modulated infrared carbon dioxide emission arriving inside the same window of a few seconds. It keeps most of the UV speed, discards the welding arc false alarm, and ignores the hot stack. For hydrogen rich and methanol services it reverts to UV sensing for the fuel relevant channel, which is exactly why the fuel must be written into the RFQ. Video flame detection, a camera plus infrared or visible imaging and recognition algorithms, earns its place where operations wants eyes on the event anyway; it verifies alarms, records them, and tolerates moderately obstructed views, but it needs illumination, it is slower than photon sensors, and its detection envelope is vendor algorithm specific rather than standardized.

Burner scanner construction is its own discipline. The scanner head carries the sensor behind a quartz or sapphire lens, screws onto a sighting tube that fixes the field of view onto the flame root, and must survive the local radiant and convective heat: standard heads serve to roughly 200 deg C ambient behind the tube, and beyond that the design moves to air or water cooled jackets and shortened, angled, or fiber optic sight paths. On small gas fired pilots, the electrical alternative is the flame rod, which measures flame ionization conductance and costs a fraction of an optical scanner, but it only works on fully conductive gas flames and short sight distances. The scanner cable is part of the instrument: UV tube scanners in particular output small pulse signals, and vendors publish maximum cable lengths and capacitance limits per cable type, which the installation chapter returns to because exceeding them silently corrupts the flame signal.

The duty boundaries deserve their own sentence each, because crossing them is the most common purchasing error. An area flame detector cannot prove that a single burner is lit: it sees a fire anywhere in a 100 degree cone tens of meters deep, which is precisely wrong for burner management. A burner scanner cannot guard a process area: it watches one flame root through one tube. And neither device detects anything before ignition or after the flame is enclosed, which is why the gas detection layer and the passive protection layers from chapter 1 are specified alongside, not instead. The acceptance machinery differs too: FM 3260 and EN 54-10 type testing governs the area devices, while scanner acceptance lives inside NFPA 85, NFPA 86, and API RP 556 practice for the burner they serve.

Key Specification Parameters

The table below lists the datasheet lines that decide whether a quoted detector actually protects the hazard. Each row is purchase defining; the paragraphs after the table explain what a strong answer looks like and what a weak one hides.

The cone of view row is where paper coverage most often diverges from real coverage. A 120 degree cone sounds generous until the sensitivity roll off is read: most models lose a substantial fraction of sensitivity in the outer 10 degrees, and rectangular cones with different horizontal and vertical angles do not map onto an oblique hazard the way a circular sketch suggests. Draw the coverage in plan and elevation against the release and pool locations, not as a single triangle. Then check the range row against fire size: the class numbers in the table below are honest vendor type test classes, and the procurement action is to demand the type test report for the exact model, because the class is only transferable between models of the same sensor architecture.

The outputs and certification rows close the integration picture. Relays remain the fail silent resistant bottom layer; a 4 to 20 mA proportional fire signal gives the logic solver magnitude and trend, HART carries diagnostics, and Modbus RTU or TCP multiplexes fire, fault, and self test status into the fire and gas system over one pair. Intrinsically safe Ex ia variants terminate through a safety barrier in the safe area, which the wiring chapter of the selection section returns to, while Ex d flameproof enclosures demand certified glands and maintained integrity. The SIL row deserves skepticism as routine: a SIL 3 capability certificate describes the device under stated proof tests, typically annual to triennial, and says nothing until the whole loop, wiring, logic solver, and final elements included, passes its IEC 61511 calculation. Ask for the safety manual with the certificate; a capability claim without a safety manual is not usable in a loop design.

Selection Decision Factors and RFQ Language

The decision sequence below runs as a checklist in a specification meeting, and each step eliminates a class of quotations before any price comparison starts.

1. Fix the duty first: area fire and gas detection or burner management proof. The technologies overlap; the acceptance criteria, response times, and approvals do not.
2. State the fuel honestly, including hydrogen, methanol, ammonia, or sulfur fractions, because fuel chemistry eliminates triple IR or locks the choice to UV.
3. Define the design fire: size, geometry, pool or jet, and the standoff distance from every candidate mounting point to the plausible fire location.
4. Map the false alarm environment: welding zones, flare stacks, hot exhausts, reflective water surfaces, and moving machinery in the field of view.
5. Choose the technology per the chapter 3 table, then demand the type test report and the cone of view polar plot for that exact model.
6. Set the voting and delay as one set: 1oo1, 1oo2, or 2oo3 across overlapping views, with the alarm delay chosen against the required detection time.
7. Fix the integration: relays, 4 to 20 mA, HART, Modbus, and the logic solver that terminates them, plus Ex d or Ex ia certification with barriers where intrinsic safety is used.
8. Close the evidence: FM 3260 or EN 54-10 test reports, SIL certificate with safety manual, cooling and purge requirements, and spares for lenses and test equipment.

Neighbor devices get chosen, not confused, and the selection chapter is where that happens. A combustible gas detector reading percent lower explosive limit is the early layer that catches the leak before any flame exists; a fixed gas detector network wired to the same fire and gas system does this continuously by mounting point; a toxic gas detector adds ppm class personnel protection for hydrogen sulfide. At ignition the flame detector takes over as the event layer, and after the event the passive layers, from the flame arrester quenching a vent line flame front to the deluge system the detector triggers, limit consequences. For building spaces, the smoke detector and the heat detector remain the correct ceiling mounted layers of the building alarm system; they are not wrong devices, they are different devices, and specifying them against an open process area is a category error. Temperature instruments cannot substitute either: a thermocouple or an infrared thermometer measures temperature at a point or on a surface and cannot distinguish a flame signature from a hot casing, so they corroborate but never prove.

Integration language belongs in the RFQ in full sentences. State where the signal lands: a safety rated PLC acting as the fire and gas logic solver, a relay based hardwired panel, or both, with the voting executed in the solver and the trip path hardwired where the SIL calculation demands it. State the protocol map, which registers carry fire, fault, and self test status over Modbus, and which relay contacts are normally energized so that a broken wire fails loud. For intrinsically safe installations, name the safety barrier model and its entity parameters in the specification rather than leaving the choice to the installer, and keep the barrier in the safe area with the calculated cable parameters recorded. On the burner side, name the flame monitor chassis and the scanner model as one line, because the pair is type tested together and the self check interval, cable limits, and response time are properties of the combination. The surrounding interlocks, from purge timing to the air and fuel proving pressure switch signals that gate every trial for ignition, sit in the same panel and belong in the same specification package.

The RFQ block should be written once and reused. A complete block reads: quantity and duty, area or burner; the fuel and design fire size with geometry; standoff distances and obstructions per mounting point; required detection time and allowable alarm delay; voting architecture and executive actions; technology with sensing bands stated; cone of view and range class with the type test report referenced; area certification and temperature class; ambient envelope and cooling or purge requirements; outputs and protocol register map; SIL capability with safety manual and proof test interval; installation accessories, mounting brackets, sun hoods, air purge sets, and lens heaters; and spares, lenses, test lamp, and shutter assemblies for the maintenance interval. A quotation that returns without the type test report or the safety manual is incomplete at any price, and the specification should say so in advance.

Installation, Standards, Failure Modes, and Maintenance

Installation quality decides whether the purchased range exists in the field. Mount detectors high on structure with the optical axis depressed toward the likely pool or jet location, and keep the last 10 degrees of the cone out of the critical coverage: sensitivity rolls off toward the edge of the field of view. Verify the actual sight lines during layout walkdowns, because grating, cable tray, and small bore piping attenuate and fragment coverage in ways no plan drawing shows. On offshore and coastal sites, specify sun hoods to keep direct radiation off the lens and marine coatings on the housing; in washdown and dusty areas, specify instrument air purge with a flow indicator so the lens stays clean between inspections. Every detector needs access: a detector that requires scaffolding to clean will be cleaned once and then never.

Commissioning proves the geometry and the chain, not just the wiring. For area detectors, perform a hot functional test with a calibrated test lamp or a controlled pan fire at the design distance, record the response time and the alarm arrival at the logic solver, and file the polar plot of the as built field of view against the hazard map. For scanners, align the sighting tube on the first third of the flame, verify flame signal strength during light off at minimum and maximum fire, and record the 0 to 10 V or 4 to 20 mA readings as the baseline for future drift checks. Respect the cable limits published for pulse type scanners: exceed the capacitance or length limit and the signal degrades in ways that look exactly like a dirty flame. Verify self check operation by forcing one cycle and confirming the fault annunciation at the fire and gas PLC or the flame monitor, then verify the fail state by disconnecting power and confirming every relay settles where the trip philosophy says it must.

The standards frame every acceptance number in this chapter. NFPA 85 organizes burner management for boilers: purge logic, trial for ignition limits, flame failure response, and the self checking expectations for scanners. NFPA 86 does the same for ovens and furnaces. API RP 556 extends the discipline to fired heaters, including viewing and scanning arrangements. For area devices, FM 3260 defines the performance approval testing behind every range class, and EN 54-10 is the European component standard for flame detectors. Loop integrity is assessed under IEC 61511, with hazardous area installation under the IEC 60079 series. Adjacent protection layers have their own standards and are never substitutes: the burner interlock world also proves combustion air and fuel pressure with a pressure switch per the applicable NFPA code before ignition is even attempted, and pressure containment belongs to the safety relief valve, not to any detection device.

Failure modes split into the loud and the silent, and the silent ones are bought with maintenance. False alarms are loud: welding arcs into a UV channel, sunlight flickering off water, hot refractory after a shutdown reheating into an IR field of view, lightning, and grinding sparks all generate trips that erode operator trust and cause masking, which is the practice of ignoring or bypassing a detector that cries wolf. The fix is architectural, correct technology per chapter 3, voting, and management of hot work in detector fields. Blind failures are silent: an oil or salt film on the lens, a dust layer, ice, steam quench between the detector and the hazard, a jammed self check shutter, a failed cooling jacket that cooks the sensor, and misalignment after nearby maintenance all reduce or destroy detection without any visible symptom. The self check shutter catches electronic and shutter faults on its 24 hour class cycle; it cannot catch a contaminated lens, which is why scheduled cleaning and functional testing exist.

Maintenance is a schedule, not an instinct. In dirty service, inspect lenses weekly and clean them on a monthly schedule with the solvent and cloth the manufacturer specifies, never dry wiped on quartz, because a scratched window permanently attenuates UV. Run a functional test with a test lamp or controlled fire quarterly in process areas and after any detector is remounted, recording response time against the commissioning baseline. Execute the proof test at the interval stated in the SIL claim, annually to triennially as certified, and file the results in the loop record. Use every opportunity the plant gives: a portable multi gas detector accompanies hot work permits near detector fields, and its readings plus the alarm history tell you whether welding exposures were managed or just survived. Corroborating instruments help diagnose drift, an infrared thermometer scan of the housing and cooling circuit finds hot installations before the electronics do, and a stack thermocouple trend that rises while the scanner signal falls points at the burner, not the detector.

Close the lifecycle with spares and records. Lenses, window assemblies, shutter parts, test lamps, and cooling coils are consumables with plant lead times, and the initial purchase order should carry them. Keep the as built alignment records, the commissioning baselines, the type test reports, and the proof test history in one loop file, because the next RFQ for the next unit is written from that file. A specification that restates measured response times, actual false alarm history, and lens consumption rates converts an opinion into a requirement, which is exactly how a purchasing file gets smarter instead of just older.