Decoking-Induced Carburisation in P22, P5 and P9 Heater Tubes
Updated: Sep 1
The cleaning cycle that is meant to extend tube life can just as easily end it.
AT A GLANCE | |
The problem | A steam-air decoke hot spot pushes local tube-wall temperature past 593°C (1100°F) while the coke layer supplies exactly the carbon activity carburisation needs. The cleaning cycle becomes the damage event. |
Why it's missed | Coil outlet temperature is an average. A single-fired heater's uneven burn hides a hot pass at precisely the resolution a routine decoke log cannot see. |
What causes it | One-sided combustion during steam-air decoke, and the "ratchet effect," where each successive spall recovers to a higher clean tube-metal temperature than the last. |
How TCR finds it | Infrared scanning through the burn itself, TOFD inspection at the next shutdown, and destructive metallography only at locations these two flag. |
The decision | API 579-1/ASME FFS-1 Part 3 (brittle fracture), applied pass-by-pass and elevation-by-elevation, because the damage is exactly as patchy as the burn that caused it. |
Why it matters
Our earlier post in this series covered high-temperature carburisation on P22, P5 and P9 tubes as a material-and-environment problem: carbon diffuses in above 593°C (1100°F) wherever a carbon-rich, reducing condition meets a susceptible alloy. This post is about one specific way that condition gets created, deliberately, on a schedule, by the plant itself: steam-air decoking.
The published record on this is blunt. A Stress Engineering Services study presented to the Canada Coking Conference, prepared with Phillips 66's fixed-equipment engineering group, lists carburisation among the recognised damage mechanisms in coker heaters and illustrates it with a photograph captioned plainly: brittle fracture in a carburised 9Cr (P9) coker heater tube. A 2024 published case study of a 9Cr-1Mo coker tube found cracking after only four years in service, the metallurgical finding a synergy of carburisation, oxidation, sulphidic corrosion, softening and creep acting together, not any single mechanism working alone.
Decoking exists to protect the tube, by restoring heat transfer and pulling tube-metal temperature back under control. The literature is equally clear that an uncontrolled decoke can undo that protection in the same event that was meant to deliver it.
The decoke event: how a cleaning cycle becomes a damage event
Steam-air decoking burns coke off the tube ID by raising coil outlet temperature (COT) with a controlled air-steam mixture, with steam moderating the burn rate so combustion temperature does not exceed the tube's maximum allowable metal temperature (TMT). For P9 (9Cr-1Mo), that ceiling is commonly held near 1250°F (677°C). Published operating data from a major coker heater licensor shows a typical online spall starting around 1200–1250°F skin temperature and recovering to a clean TMT of roughly 900–950°F once the burn completes.
Those are heater-average numbers. The burn itself is not uniform, and that is where the risk sits.
One-sided firing during online spalling recovers less of the coked coil each time, so a single-fired heater's peak skin temperature per spall creeps upward, well past the carburisation threshold, cycle after cycle. Double-fired heaters, with more even heat flux, largely avoid this drift.
A refiner running a single-fired coker heater sees exactly this "ratchet effect" in its own spall records: each successive online spall recovers to a somewhat higher clean TMT than the one before, because one-sided firing cannot deliver even heat flux across the coil. Left unmanaged, that drift eventually forces a full steam-air decoke or mechanical pigging outage. Every spall along the way, and the eventual full decoke itself, is a fresh excursion above the carburisation threshold at whichever pass runs hottest, an excursion the plant's own COT instrumentation, reading an average, will not show.
What happens inside the metal
The chemistry is the same carburisation mechanism our earlier post described, but a decoke hot spot delivers both ingredients the mechanism needs in one place at one time: tube-wall temperature above 593°C, and a coke-and-combustion-product layer sitting directly on that hot wall supplying carbon activity.
Carbon entering the P22, P5 or P9 matrix precipitates carbides in a well-documented sequence for 9-12% chromium steels: from the matrix, through transitional epsilon-carbide and iron carbide plus M7C3, settling into M23C6 as exposure continues. M23C6 nucleates preferentially along prior-austenite grain boundaries, is thermodynamically the most stable carbide in the sequence, and carries a low bond strength to the surrounding matrix. That combination is exactly what makes the damage intergranular: once the grain-boundary carbide network is continuous, a crack has a ready-made low-energy path to follow.
Carbon diffusing in from a decoke hot spot precipitates M23C6 along prior-austenite grain boundaries first, in the sequence matrix → epsilon-carbide → Fe3C+M7C3 → M23C6. Once the boundary network is continuous, cracking runs along it rather than through the grain.
Detection and forensics: the burn itself is the first inspection point
Because the damage is created during a specific, scheduled event, the most effective detection window is the decoke itself, not the next turnaround.
During the burn. Refiners running steam-air decoking on coker heaters have moved to continuous infrared scanning through the entire burn, precisely because coil outlet temperature and thermocouple data alone do not resolve a hot pass. TCR Advanced treats this as the first line of defence: an IR record of every pass, at every elevation, for the full duration of the decoke, not a spot check at the end.
At the next shutdown. TOFD detects a carburised layer through a phase-shift second lateral wave, a distinct high-amplitude signal generated with a high-frequency probe (above 15 MHz), reflecting off the boundary the carburised case forms against sound base metal. It is not a velocity or attenuation trend. Comparing the resulting TOFD map against the IR record from the hot pass or hot tube identified during the burn raises confidence in the damage-probability picture for that location, which is what TCR Advanced uses to decide where sampling and destructive assessment are justified. AMPP's own 2023 review of carburisation assessment methods is candid that this remains a genuinely difficult inspection problem industry-wide, and that destructive metallography is still the most reliable single confirmation available. TCR Advanced treats IR-during-burn and shutdown TOFD data together as the screen that decides where to look, not as a replacement for that destructive confirmation.
Eddy current testing, useful on austenitic HK/HP tubes because carburisation there measurably increases ferromagnetism, does not apply on P22, P5 or P9. All three are already ferromagnetic, so there is no magnetic-state transition for an eddy current coil to detect. This is the same limitation our earlier post flagged, and it holds regardless of whether the carburisation was decoke-induced or process-induced.
Where the TOFD signal or the decoke IR record flags a location, TCR Advanced pulls a boat sample with a sequence of destructive tests including Vickers hardness traverse, Charpy impact testing and SEM/EDS carbon mapping, the same destructive sequence used for process-side carburisation, confirming case depth and toughness loss before any FFS number is calculated.
The decoke record itself is an input to the FFS decision, not a separate exercise from it. A pass that ran hot during the burn earns closer inspection whether or not it shows any wall loss.
FFS assessment and the RLA path
Because the intense heat from a decoke burn is uneven, the resulting carburisation damage occurs only in isolated patches. Evaluating the heater's reliability against a heater-wide average hardness, or a single heater-wide FFS conclusion, is not meaningful here, since averaging hides exactly the localised severity that matters.
The assessment has to be targeted instead. TCR Advanced uses the IR record from the burn and the TOFD inspection at shutdown together to identify the hot passes and elevations, then runs API 579-1/ASME FFS-1 Part 3 (brittle fracture) on those exact locations, using the hardness and toughness data from destructive sampling described above, exactly as set out in our earlier post on process-side carburisation. That combination, IR during the burn plus TOFD at shutdown plus destructive confirmation only where both flag a location, is what gives the inspection programme its precision against a damage mechanism that does not distribute itself evenly, because the burn that caused it did not either.
Where a flagged pass also carries meaningful creep exposure, which is common on coker radiant tubes given the temperatures involved, TCR Advanced cross-checks the brittle-fracture assessment against a Larson-Miller remaining-life estimate for the unaffected base metal, using the same creep-life methodology set out in Part 2 of this series. The two assessments answer different questions on the same tube: whether the carburised layer is at risk of brittle fracture now, and how much creep life the surrounding sound material has left.
Prevention and next steps: managing life through the burn, not just around it
The ratchet effect and the hot-spot risk it creates are operating decisions, which means they respond to operating discipline.
Evaluate every burn, not just the outlet. IR scanning through the full decoke, pass by pass, is the single change that converts a decoke from a blind procedure into a documented one. Without it, the plant genuinely does not know which pass ran hottest until a tube fails or an inspection happens to find it.
Track the ratchet, do not wait for it to force a shutdown. A single-fired heater's clean TMT trend after each successive spall is itself a leading indicator. Logging that trend, rather than treating each spall as an isolated event, tells the plant how many spalls remain before a full steam-air decoke or mechanical pigging outage becomes unavoidable, and gives advance warning of exactly which passes to watch.
Feed the decoke record into life management, not into a filing cabinet. TCR Advanced logs IR-through-burn data, shutdown TOFD results and any destructive findings into AiOM against the same pass-and-elevation grid, so a tube's decoke thermal history and its carburisation risk are read together at every turnaround, not reconstructed from memory when a tube finally shows a problem.
“A decoke is not maintenance until it is proven to be. Every pass that goes past eleven hundred Fahrenheit during that burn is either brought back under control within minutes, or it is a carburisation event you will find later, on somebody else's schedule.” — Paresh Haribhakti, Managing Director, TCR Advanced |
Frequently asked questions
What is decoking-induced carburisation?
It is carburisation caused specifically by the steam-air decoking event, rather than by ordinary process-side coke contact. A decoke hot spot pushes local tube-wall temperature above 593°C (1100°F) while the coke and combustion by-products sitting on that wall supply the carbon activity the carburisation mechanism needs, so the cleaning cycle itself becomes the damage event.
Why does a controlled decoke still cause carburisation?
Coil outlet temperature is a heater-average reading. Steam-air decoking is rarely uniform across every pass, so even a decoke that looks well controlled on the outlet instrumentation can carry one or more passes well above the average, and above the carburisation threshold, without that excursion being visible on the plant's normal controls.
What is the "ratchet effect" in coker heater decoking?
In single-fired coker heaters, one-sided combustion delivers uneven heat flux, so each successive online spall recovers to a somewhat higher clean tube-metal temperature than the spall before it. Left untracked, this drift eventually forces a full steam-air decoke or mechanical pigging outage, and every spall along the way is a fresh excursion risk. Double-fired heaters, with more even heat flux, largely avoid this drift.
How is decoking-induced carburisation different from process-side carburisation?
The underlying metallurgy is the same carbon-diffusion and carbide-precipitation mechanism described in our earlier post on P22, P5 and P9 carburisation. The difference is timing and location: decoking-induced carburisation is created in a short, specific, scheduled event, at whichever pass and elevation the burn ran hottest, rather than accumulating gradually from continuous process-side exposure.
How is decoking-induced carburisation detected?
TCR Advanced treats infrared scanning through the full decoke burn as the first inspection point, since it is the only way to resolve a hot pass that outlet-temperature control cannot see. This is cross-checked with Time-of-Flight Diffraction (TOFD) ultrasonic testing at the next shutdown, a phase-shift signal from a high-frequency (above 15 MHz) probe that reads the boundary the carburised case forms, not a velocity or attenuation trend. Destructive boat sampling, hardness testing, Charpy testing and SEM/EDS carbon mapping are reserved for locations either method flags. Eddy current testing does not apply, since P22, P5 and P9 are already ferromagnetic.
What does the FFS and life-management path look like?
The assessment runs under API 579-1/ASME FFS-1 Part 3 (brittle fracture), applied at the specific pass and elevation the decoke IR record and the shutdown TOFD result flag, since the damage distributes as unevenly as the burn that caused it. A heater-wide average is not meaningful. Where meaningful creep exposure is also present, this is cross-checked against a Larson-Miller remaining-life estimate for the surrounding sound material. Results are logged against the same pass-and-elevation grid in AiOM so decoke thermal history and carburisation risk are read together at every turnaround.
When are single-fired furnace designs preferred?
Single-fired designs suit heaters where the process fluid is stable and the risk of severe coking is low: crude atmospheric distillation, vacuum charge heaters (often), catalytic reforming and general reboiler duty. A single-fired heater needs lower capital expenditure (CAPEX), a smaller footprint, and simpler piping and burner management than a double-fired design, which is why it remains the default choice outside coking and other high-fouling services.
Can a refinery switch a heater between single-fired and double-fired?
No. The classification describes the furnace's physical architecture: the brickwork, tube placement, fuel-gas headers, and the holes cut into the steel casing for the burners. Once a heater is built, the firing geometry is locked. A refinery cannot convert a single-fired heater to double-fired firing for a decoke, or for any other reason, without rebuilding the furnace.
What happens if a single-fired heater is switched to heavier feedstock?
If a refinery processes heavier, "opportunity" crude oils to save on feedstock cost, a single-fired heater fouls markedly faster. The uneven heat flux drives the ratchet effect described above: successive online spalls cannot fully clean the coil, forcing more frequent full shutdowns. Mechanical pigging is often the better response in this situation, since it removes the thermal-excursion risk that comes with repeated steam-air decoking, and with it, the ratcheting.
Why is mechanical pigging not always the answer?
Unlike online spalling, which is carried out with the heater still at temperature, pigging requires the heater to be taken fully offline, cooled and depressurised. The technique drives a studded or abrasive foam plug through the coil under high-pressure water. It will not cause carburisation, but repeated passes can slowly erode healthy base metal at the ID, contributing to wall thinning over the life of the heater. Pigging trades a carburisation risk for a mechanical erosion risk; it does not remove risk altogether.
Comments