A step change in three exhaust thermocouples at full speed no load, with everything steady before and after, is a different animal from a slow spread drift. Step changes come from one of two places: something in the gas path stopped delivering heat to that sector, or something in the measurement chain stopped reporting it. The order in which you check those two decides whether you spend a shift with a laptop or a week with the covers off.
Read the Spread Numbers Before You Open Anything
The combustion monitor does not alarm on a single temperature. It computes the average exhaust temperature, then the highest spread (hottest minus coldest thermocouple), the second spread (hottest minus second-coldest) and the third spread, and compares them against an allowable spread that scales with average exhaust temperature and firing conditions. At full speed no load the allowable spread is at its tightest relative to actual firing temperature, which is exactly why a marginal can shows up at FSNL and hides at base load.
Pull these off the control display or the historian before touching hardware:
- Every exhaust thermocouple, in their physical circumferential order - not the order they appear on the screen.
- Highest, second and third spread, plus the allowable spread value at the moment of the alarm.
- Average exhaust temperature and compressor discharge conditions across the same window.
- Which thermocouple was the minimum during the event, and whether the minimum jumped between adjacent channels.
If the second and third spreads climbed with the highest, several neighbours went cold together and the problem is in the gas path. If only the highest spread moved and the second stayed put, one channel wandered off alone and the odds shift toward instrumentation - even when the trend shows three lines dipping, because two of them may just be normal scatter riding on a lower average.
Check 1: Map the Cold Thermocouples Against the Exhaust Twist
Exhaust gas leaves the combustion cans and turns through the turbine stages, so the sector that a given can heats does not sit directly downstream of it. The offset - the exhaust twist - is machine specific and comes from the OEM combustion troubleshooting curves for the unit. Get that mapping from the turbine documentation rather than assuming a one-to-one relationship; on a fourteen-can machine the thermocouple count does not match the can count, so one weak can typically pulls two or three adjacent thermocouples down and leaves a partial dip on the ones either side.
Branch on adjacency:
- Three cold thermocouples physically adjacent in the exhaust annulus - a combustion or hot-gas-path problem in the can that the twist curve points to. Go to Check 3.
- Cold thermocouples scattered around the annulus - no single can can produce that pattern. Go to Check 2.
- Cold thermocouples adjacent on the terminal board or the same input card, but not adjacent in the annulus - stop and treat it as a control system fault. Go to Check 2 and stay there.
That last case is the one crews miss. Screen order usually follows wiring order, and wiring order usually follows physical order, so a plot of three neighbouring traces dropping looks like combustion when it is actually one card, one shared cold junction, or one loose common.
Check 2: Clear the Instrument Path Before Pulling Hardware
Wrong fix first: the reflex is to acknowledge the alarm, run it again, then pull the cans on the sector the trend points to. If the fault is in the measurement chain you will find nothing, reassemble, and get the same alarm on the next start.
Work the chain end to end:
- Read all three suspect thermocouples cold, with the unit down and equalised. Anything reading materially off the others at ambient is a sensor or wiring problem, not combustion.
- Check whether the three inputs share a card, a terminal board, or a common return. A single card or a loose common lifting under vibration drops several channels in one step - which is precisely the signature on the trend.
- Verify cold junction / ambient compensation. A loose or mis-terminated compensation sensor, or the wrong reference applied to one terminal board, biases a whole group of thermocouple signals at once and produces a clean step.
- Loop-check with a millivolt source at the field end of the suspect channels and confirm the indicated temperature on the control display. Then repeat at the panel end to split field wiring from card.
- Inspect the thermocouple penetrations in the exhaust plenum for burned insulation, chafed leads, and moisture in junction boxes. Wet junction boxes shunt the loop and read low.
If all three channels prove out electrically and the cold readings track each other, the measurement chain is clean. Move to the gas path.
Check 3: Replay the Startup Trend for the Early Anomaly
The irregularity on one of those same thermocouples about six minutes into the start is the most useful piece of data on the whole 34.5-minute run. Six minutes in puts the machine somewhere in the acceleration/warm-up region, where crossfire and flame stability in every can are being proven. A channel that misbehaves there and then participates in the FSNL spread event was not healthy at any point in the run - the can it watches was struggling from light-off onward, and FSNL simply pushed it past the alarm limit.
Pull the previous three or four starts from the historian and plot the same channels. Two questions decide the branch:
- Did the same thermocouple show the same early-start signature on previous runs? A repeating signature on one channel across multiple starts, with normal behaviour at load, points to a specific can or its fuel supply.
- What did those channels do at higher power levels on earlier runs? If the spread closed as load came on, suspect lean operation at FSNL. If the spread widened with load, suspect a leak path - transition piece, seal, or liner.
Match the Pattern to the Cause
| Observed pattern | Mechanism | Next check |
|---|---|---|
| One thermocouple steps down, neighbours steady | Sensor, extension wire, or single input channel | Cold reading and millivolt injection on that channel |
| Three or more step down together, physically adjacent | Loss of heat release in one can - partial flameout, misdirected or blocked nozzle | Fuel side, then borescope that can |
| Three step down together, not adjacent but on one card/terminal board | Card failure, loose common, cold junction compensation error | Control panel wiring and compensation reference |
| Spread grows gradually with load, cold sector fixed | Transition piece crack or seal leak dumping compressor discharge air into the hot gas path | Borescope of transition pieces and aft liner seals |
| Spread present at FSNL, closes at load | Lean operation near blowout in one can at low fuel flow | Fuel nozzle flow, gas or liquid distribution, mode staging |
| Cold sector plus rising CO in the stack | Incomplete combustion in one or more cans | Emissions log against spread; fuel nozzle condition |
Check 4: Work the Fuel Side
On a DLN-1 machine the presence of a secondary nozzle tells you the combustion system stages between primary, lean-lean, secondary and premix as combustion reference temperature climbs. At FSNL the machine sits in a low-fuel-flow mode with the flame at its leanest, which is where a can with the wrong nozzle pressure ratio drops out first. A can that flames out or burns poorly does not produce smoke you would see at the stack, so a clean stack rules out very little.
- Gas fuel: compare nozzle supply pressure against the expected pressure ratio for the mode the unit is in. Look for a fouled or eroded metering orifice, a partially closed manifold valve, or an incorrectly seated nozzle changing the split between primary and secondary.
- Distillate or naphtha: blocked liquid nozzles are common, especially on naphtha. Flow-test the nozzles for that sector on a bench and compare flow numbers set to set, not just look for a pattern of spray.
- Dual fuel: check the atomizing air and the purge/check valves on the idle fuel system. A leaking check valve on the unused fuel drips into one can and skews it.
- Emissions: log CO across the FSNL soak. A CO rise coincident with the spread event confirms poor combustion rather than a measurement fault.
The loose secondary nozzle with the badly installed gasket you already found is a genuine defect and worth having corrected, but one loose gasket does not explain three thermocouples stepping down simultaneously in an otherwise steady condition. Treat it as fixed and keep going.
Check 5: Inspect Cans, Liners and Transition Pieces
Once the instrument path is clean and the fuel side checks out, the cold sector is a hardware problem. A cracked transition piece is the classic producer of a localised cold spot: compressor discharge air short-circuits into the transition and dilutes the gas reaching the first-stage nozzle in that sector. The thermocouples downstream go cold in a step when the crack opens under thermal load, and stay cold, which matches a sudden drop with steady behaviour either side of it.
- Borescope the cans identified by the twist map, plus one on each side. Look at the liner cap, the effusion and cooling holes, the crossfire tubes and retainers.
- Inspect transition pieces at the aft frame and side seals for cracking, wear, and missing seal material.
- Check the crossfire tubes specifically - a burned or displaced tube prevents reliable light-off and cross-lighting in that can, which is what an early-start anomaly on the same thermocouple points to.
- Inspect the first-stage nozzle segments visible through the transition for burning or blockage in the cold sector.
- Pull and bench-check the nozzles from the suspect cans and at least one known-good can for comparison.
Run the Repair, Then Verify Before You Load
Get it running, then fix it properly - but do not clear the alarm and load the machine on the assumption that it will close up. Repeated operation with a cold sector means one first-stage nozzle sector runs cold and the rest run hot to hold the same average, which is how you turn a combustion problem into a hot gas path bill.
- Correct the defect found, reassemble with new gaskets and seals on every joint disturbed, and torque nozzle and cover hardware to the OEM values.
- Confirm the exhaust thermocouple wiring landed back on the correct terminals if any of it was disturbed - a swapped pair produces an artificial spread that reads exactly like a bad can.
- Start the unit and hold at FSNL for a full soak, at least as long as the run that produced the alarm.
- Trend continuously and confirm the highest spread stays well under the allowable value, the second and third spreads track normally, and no channel steps.
- Load in stages, holding at each step, and confirm the spread closes rather than opens.
Log this set at one-second resolution or faster through the whole run:
each exhaust thermocouple (physical order)
average exhaust temperature
highest / second / third spread
allowable spread
compressor discharge pressure and temperature
fuel flow and nozzle supply pressure (per manifold)
combustion mode / staging status
CO and NOx at the stackA repaired machine shows a spread that is stable at FSNL and narrows as load comes on. If the spread reappears at the same thermocouples after the repair, do not repeat the same disassembly - go back to Check 2 and prove the measurement chain with injection, because the same fault reproducing after correct hardware work is usually electrical.
When to Stop and Escalate
Stop if the borescope shows cracking in a transition piece, liner, or first-stage nozzle, or if the spread stays outside the allowable band after the fuel and instrument checks are clean. Continued running with a confirmed cold sector risks hot-section damage well beyond the cost of the outage, and the combustion troubleshooting curves, allowable spread constants, and repair limits for the unit come from the turbine OEM's service organisation. Contact GE service with the trends, the twist map, the borescope images, and the spread data before you fire the machine again.
FAQ
How do I tell whether a Frame 7EA exhaust spread alarm is a bad thermocouple or a bad can?
Count how many channels moved and where they sit. One channel stepping down with steady neighbours is a sensor, extension wire, or input channel fault; three or more adjacent channels dropping together in the exhaust annulus is a combustion or hot-gas-path problem. Confirm by reading the suspect channels cold with the unit down and by injecting millivolts at the field end.
How do I map an exhaust thermocouple back to the combustion can that caused the cold spot?
Use the exhaust twist mapping from the unit's combustion troubleshooting documentation. Gas swirls as it passes through the turbine stages, so the cold thermocouples sit circumferentially offset from the can that produced them, and the thermocouple count does not match the can count.
How do I check whether the control system is causing three thermocouples to drop at once?
Trace the three inputs on the control drawing and see whether they share an input card, a terminal board, or a common return. Then verify the cold junction / ambient compensation for that board - a loose or mis-terminated compensation sensor biases a whole group of thermocouple signals in one step.
How do I confirm poor combustion rather than a measurement fault at full speed no load?
Log CO at the stack across the FSNL soak alongside the spread values. A CO rise coincident with the spread event confirms incomplete combustion in one or more cans; flat CO with a widening spread points back at the thermocouple loops or the control hardware.
How do I verify the repair before returning the unit to service?
Hold at FSNL for at least as long as the run that alarmed, trending every exhaust thermocouple plus the highest, second, and third spreads against the allowable spread. Then load in stages - a correctly repaired machine holds a stable spread at FSNL and narrows it as load increases.