Exhaust Over Temperature and Abnormal Heat Emission During Startup


Thread Starter


Hello Experts,

One of my gas turbines, a GE Frame 5, 25MW capacity with MK5 Control system suddenly tripped on exhaust over temperature trip. Prior to this trip the unit was running on 20MW with a maximum exhaust temperature of 498 degrees Celsius. It has been stable at this performance for over a year before the sudden trip on exhaust over temperature. Maximum operating temperature of the unit at base load is 501 degrees under ISO condition.

The trip was observed to be instantaneous with exhaust temperature spiking to as high as 800 degrees Celsius. There has been no history of exhaust over temperature on the unit and no surge of gas flow occurred before the trip. After the trip, the unit was restarted with but failed to accelerate beyond 70%TNH with exhaust spiking as high as 600 degrees Celsius. Also there was high emission of heat from load gear and turbine compartment melting the the plastic lighting fittings.

The following inspection and maintenance actions have been carried out so far to correct the issue:

-Replacement of all failed exhaust thermocouples

-Replacement of failed intervalve pressure transmitters

-Inspection and cleaning of fuel nozzles

-Cleaning of main air inlet filters which improved the CPD from an 1.2bar to 2.45bar at 70%TNH

-Calibration of gas control valves

-Calibration of IGV

-Boroscope inspection on hot gas section (no major defect was found)

-Inspection of exhaust thermal insulation

Following these inspections and maintenance actions, several startups were attempted but the unit failed to accelerate beyond 72% TNH with exhaust temperature getting as high as 600 degree Celsius.

Please I need to know what could be the possible cause of the high exhaust temperature and heat emission during startup, and what further actions we should take to correct it.


WOW! There's a LOT going on here. You mentioned replacing "failed exhaust T/Cs"--but you didn't say how many were failed and how long they had been failed, and how they had failed (high or low).

You mentioned replacing "failed intervalve pressure transducers"-plural, so more than one. Precisely how many P2 pressure transducers were "failed" and how had they failed (low or high), and how long had each one been know to have failed?

Prior to "calibrating of gas control valves" (plural) and IGVs, was the "calibration" of the gas control valve(s) and the IGVs verified, and was the as-found condition (indicated position and actual, measured, physical position) recorded? Was the as-found indicated LVDT position(s) on the operator interface screens found to be so far different from the actual, measured valve position(s) that LVDT calibration was deemed necessary? ("Calibration," using AutoCalibrate or even manual calibration using Demand, or User-Defined, Displays ONLY affects LVDT feedback scaling--nothing else. And, as with any other control device like a pressure switch or a pressure transducer or a limit switch one first checks operation by recording the as-found condition and comparing it to the specification before performing any adjustment ("calibration") if necessary. LVDTs are no different than any other control element in that sense. And AutoCalibrate does NOTHING to anything except LVDT feedback scaling. Nothing. Full stop. Period.) This happens all too often--people just "calibrate" valves and IGVs without measuring the actual physical position and comparing it to the indicated position on the operator (the scaled LVDT feedback) and when they do this they inadvertently mess up the LVDT feedback scaling (which is all that AutoCalibrate does is scale LVDT feedback) and that can be part of the problem. One should ONLY "calibrate" valves and IGVs when it's indicated--after verifying the as-found indicated position against the actual, measure physical position and only if it's necessary should AutoCalibrate be used--and the as-left conditions should be recorded for future troubleshooting, to see if anything has changed during the next verification of LVDT scaling/feedback.

So, usually, when GE-design heavy duty gas turbines experience higher than normal exhaust temperatures during acceleration it's because of problems with either the starting means (induction cranking motor; diesel engine; torque converter) or excessively low air flow through the turbine (caused by dirty inlet air filters) resulting in higher-than-normal exhaust temperatures--which usually hit the maximum allowable exhaust temperature limit, Control Constant TTKn_I, where 'n' is a whole number from 0 to 7. 600 deg C would probably be pretty close to the maximum allowable exhaust temperature limit (the "I" in TTK_I means Isothermal exhaust temperature limit) during acceleration. So, it seems like something is really amiss with the amount of air and/or hot gases getting through the unit and into the turbine section, where it is converted to work/torque to help accelerate the turbine-generator shaft. In your case the starting means should be disconnected from the turbine-generator shaft before reaching 70% TNH (usually well before that). So, it appears its more likely there is and air flow problem or a serious leak (see below).

"High heat emissions" and melted plastic fittings (there shouldn't be plastic electrical fittings in any compartment...) are NOT caused by the turbine control system--but by leaks from the compressor discharge casing and/or turbine shell or exhaust casing. And, if there's that much heat during start-up/acceleration then there's a pretty significant leak or leaks and that could also explain why the unit isn't accelerating--not enough heat is getting to the turbine section after the starting means is disconnected from the turbine-generator shaft to allow self-sustained operation and continued acceleration. So, if you haven't already starting looking for leaks in the turbine compartment--you should. Doesn't sound good, at all. I would expect leaks of that magnitude (to melt plastic electrical fittings) would audible--and loud. You can have people with long sticks with rags tied securely to the end of the stick waving around joints and fittings looking for leaks which would make the rag blow. (I recommend having the turbine compartment vent fan(s) NOT running for this test--and only for a brief period of time during the test, especially if there's that much hot air/combustion gases leaking into the turbine compartment. It could be dangerous to personnel to be there for very long--but there's really no other way to identify where exactly leaks are coming from than this method when the unit is running. (People have tried using infrared heat "guns", but it's not a very good method--certainly higher technology that the rag tied securely to the end of a stick--but sometimes old technology can't be beat.)

If there were more than one failed P2 (gas fuel intervalve) pressure transmitter that could explain why there was a sudden spike in exhaust temperature. If a second P2 pressure transducer failed indicating a low P2 pressure then two (2) control processors would open the SRVs to try to raise the P2 pressure and that can definitely cause a spike in gas fuel flow-rate through the gas control valve and result in an exhaust temperature spike.

If the unit is slowing down it's acceleration rate and the speed is leveling off instead of increasing and the exhaust temperature has peaked at the Isothermal limit (TTKn_I) during acceleration then there's something wrong with the amount of hot gases getting to the turbine section to provide the work/energy (torque) necessary to accelerate the turbine-generator shaft. There's either an obstruction at the entrance to the compressor (you say you "cleaned" the inlet air filters (???), but you also said you "calibrated" the IGVs...). Or, the leak in the turbine compartment (from the compressor discharge casing or the turbine casing or the exhaust shell) is so bad that not enough hot gases are getting to the turbine nozzles/blades. Trying to increase the fuel isn't the correct answer--and it can cause serious problems with hot gas path hardware (combustion liners; transition pieces; turbine nozzles--AND the exhaust diffuser, which is very sensitive to exhaust temperatures above the Isothermal limit).

Also, can you tell us what the three exhaust temperature spread values are (TTXSP1, TTXSP2 and TTXSP3) when the unit is not accelerating above 70% TNH?

Please write back to let us know what you find!

By the way, I've rethought my decision to remove the question about Diagnostic Alarms during the operation prior to the exhaust overtemperature trip. Please tell us what Diagnostic Alarms were present and active prior to the original exhaust overtemperature trip from load.

And, please provide, if possible, the list of alarms from the Trip History Display when that trip occurred.

And, please provide the list of alarms (both Process and Diagnostic) being annunciated when the unit is failing to accelerate during starting.

Finally, please explain what you are doing to abort the start when the unit is failing to accelerate--are you initiating a STOP from the operator interface, or a trip from the Emergency Stop Push-Button, or is the unit automatically shutting down (it would (should) be accompanied by a Process Alarm to indicate the reason), or is it tripping (again, that would (should) be indicated by a Process Alarm)?

Dear CSA,

Thanks for your reply.

I would like to take you through the performance and status of the unit prior to the trip on exhaust over temperature.

The unit was recovered in November 2016 after a forced outage due to diesel engine starting means. After recovery the unit has been operational on an average of 20MW without any outage due to hardware component or system failure or control system malfunction until the trip on exhaust over temperature, on December, 2018.

Furthermore, In response to your questions:
All T/C's were changed after the trip as more than five were observed to be reading low (-84 degrees Celsius).

All three P2 pressure transducers (96FG-2A, 96FG-2B, 96FG-2C) were observed to have failed; reading low as indicated on the HMI. However, at same time low P2 pressure was observed on the HMI, P2 pressure on the local pressure guage was at 16bar max. on base load, a 2bar greater than the set point of pressure P1 of 14bar. The unit did not experience any abnormal exhaust temperature in this condition. After the replacement of the failed P2 transducers with new ones, P2 pressure was

Prior to calibration of the gas valves (SRV and GCV) LVDTs, as-found conditions were duly recorded. however the calibration in this case was the Manual calibration using the user-defined to stroke the valves physically observing the stroke comparing the with the LVDT feeback. Valve positions responded accordingly with changes in the manual setpoints.

Similar to the gas valves, the VIGV was calibrated to modulate it and watch to response to the manual setpoints. However the IGV response could not be ascertain since the unit could not get to up to 85%TNH (the speed at which the IGV begins to open)

The GE MK5 Control in this case on displays one exhaust temperature spread (TTXSP-1) on the HMI. This value was over 700 degrees when the unit stopped accelerating at 70%TNH.
Dear CSA,

No alarm was recorded prior to the trip. The trip was a sudden occurrence with no build up of preceding events annunciated. When the unit failed to accelerated beyond 70% the "Exhaust Temperature High Alarm" followed by the "Exhaust Overtemperature trip"

The first start after the initial trip saw the unit tripping on "exhaust overtenperature trip" with exhaust rising. Subsequent start attempts were aborted my shutting the emergency shutoff valve due to the excessive emission of heat from the turbine and load gear compartments.

Again, I am forced to say: WOW! Some good information; but you didn't provide the list of Diagnostic Alarms (which must have been <i>HUGE</i>--<b>more than five</b> failed exhaust T/Cs, and three failed P2 pressure transmitters, at a minimum). It's a real credit to the designers of the GE Frame 5 heavy duty gas turbine that they take such a beating and continue to run, sometimes for long periods of time (such as in this case!). They can be hard to start under these conditions--but once they get running they will just keep going and going and going until <i>something</i> trips them, and then they can be hard to start.

It's difficult to pick a place to start--so I'm just gonna dive in.

Many GE-design heavy duty Frame 5 gas turbines prior to, and in the early days of, Mark V production did NOT have Combustion Monitoring. Combustion Monitoring is the protection scheme that detects combustion problems by comparing exhaust T/Cs to each other and initiating a Process Alarm if conditions indicate a problem, and a trip if conditions worsen. Combustion Monitoring has been covered many times before on control.com, so I'm not going to explain exactly how it works (AND your unit doesn't seem to have it--or it's been forced "off", which is a possibility). So, that could explain why five exhaust T/Cs were reading negative values and the unit continued to run (from the information provided). (By the way, even if the unit had Combustion Monitoring, it is not active during starting and acceleration or during shutdown--effectively it is disabled below approximately 95% speed by the Mark* sequencing/application code.)

(I will add one more thing about the Combustion Monitor--it's really composed of two functions in sequencing--a Big Block that calculates the spreads and checks adjacency AND a bit of ladder logic (rungs) that check the conditions from the Big Block to trip the turbine (because a trip requires a high spread AND adjacency. So, it's possible the CSP in the Mark V at your site has the Big Block, but it doesn't have the rungs to generate the trip. Frame 5s, for MANY years (decades) were pretty bulletproof (robust); and in some parts of the world they were considered critical to the local economy. So, they had very few trips configured--and often there was no Combustion Monitor tripping function. It would be necessary to review the CSP at your site to be able to say for certain if the unit has it or not.)

The fact that TTXSP-1 is 700 (deg C!) during starting (at around 70%-80% speed) indicates that one or more combustors does NOT have flame in it--which would definitely help to explain why the unit is not accelerating. If all the combustors are not lit (and only a small number have flame detectors--not ALL combustors have flame detectors) then there won't be enough energy in the combustion gases for the turbine to convert to torque to help accelerate the unit to FSNL (Full Speed-No Load). This is NOT GOOD for the turbine hot gas path components--and can cause the Mark V to add additional fuel to try to help with acceleration, which can increase the average ("median") exhaust temperature to the point that exhaust temperature control is limiting the amount of fuel. And, limiting the fuel because of excessive average exhaust temperature (TTXM) when the flame does not exist in one or more combustors means there is a LOT of fuel going into the combustors--more than should be.

It also means that unburned fuel is going into the turbine and through the exhaust diffuser into the exhaust duct. If there is a spark or ignition source in the exhaust, that can result in a pretty serious BANG (explosion).

The Mark* (on older, conventional flame combustion systems) just wants to see one flame indication during operation. Some older units only have two flame detectors, and they are redundant--only one is needed to keep the unit from tripping. Again, that means the remaining combustors (as many as eight (8)) don't have flame detectors--so flame can be lost or missing in one or more combustors. Especially during starting and acceleration.

"Emissions of heat" from the turbine is NOT monitored, controlled or protected against by the Mark V (until very recently). If the heat in the turbine compartment (or any compartment) is high enough to melt plastic conduit components--there HAS to be either a leak of hot combustion gases or a fire somewhere. The wiring insulation used in most GE-design heavy duty gas turbine compartments is only rated for 90 deg C, some of it has a slightly higher rating, but it's not impervious to heat and will eventually melt or degrade. This has NOTHING to do with the Mark*. Nothing whatsoever. The Mark* can't be used to understand where the high temperature is coming from, and it doesn't control the "emission" of high temperatures outside of the turbine--because ALL high temperature combustion gases, including axial compressor discharge air, should be passing through the turbine and into the exhaust. Any high temperature "emissions" would be degrading the performance and increasing the heat rate of the unit, meaning that fuel is being wasted because the hot gases are not passing through the turbine to produce work (torque).

The GE "standard" is: ANY condition that results in a trip (immediate shut-off of fuel to the combustors) will cause a Process Alarm to be annunciated. 99.88% of the software that comes from the "factory" in the Mark* is properly configured to do this; only a VERY small number have one, maybe two, un-alarmed trip conditions--and I have personally only encountered these coming from the factory when there is unusual Customer- or site requirements that were not properly followed by the requisition engineer and not properly demonstrated during the FAT (Factory Acceptance Test).

The reasoning for this is clear: The operator(s) need to be made aware of the condition that resulted in the Mark* tripping the turbine. Operators should NOT have to guess, and this helps to prevent guessing.

Having said that, I have been to several sites where commissioning personnel, and sometimes site technicians, have added trip inputs to the Mark* and NOT configured a Process Alarm to indicate why the turbine was tripped. An oversight, and a serious one.

Let's say a unit was running along fine and suddenly something happens in the fuel supply system to interrupt the flow of fuel to the unit causing the flame to be lost in the combustors--essentially a "trip" condition, but not initiated by the Mark* for any protective purpose. The Mark* is programmed to detect a loss of flame when NO protective condition was sensed by the Mark*--and it will annunciate a "LOSS OF FLAME TRIP" Process Alarm. That is an indication to the operator(s) that something caused flame to be lost OTHER than a protective condition detected by the Mark*. So, even outside factors which can interrupt fuel flow and cause a loss of flame can be detected by the Mark*--it may not be able to say exactly what caused the reduction or stoppage of fuel flow to the unit, but it can say, "Something caused the flame to go out--and it wasn't me!"

Unfortunately, without being able to see the CSP (Control Sequencing Program) running in the Mark V at your site, it's IMPOSSIBLE to say for certain if there is a condition in the Mark V that will result in a shut-off of fuel to the turbine (using the fuel stop valve(s)) that does not have a Process Alarm to tell the operator(s) why the turbine was tripped. I am going to say the likelihood of there NOT being a Process Alarm for a trip condition detected by the Mark is extremely low, as in practically non-existent. About the ONLY way I could say there was not a Process Alarm--ESPECIALLY IF THE UNIT WAS EXPERIENCING EXTREMELY HIGH EXHAUST TEMPERATURES, HIGH ENOUGH TO TRIP THE TURBINE-- is if the operator interface and/or the Alarm Logger (the dot matrix alarm printer) had lost power or the printer didn't have any paper in it or the paper was jammed or the printer ribbon was worn through or torn. So, either the operator interface monitor couldn't display the alarm (because there was no electricity to do so), or the printer couldn't print the alarm (because it didn't have power, or it wasn't maintained properly).

I will grant that GE makes it VERY DIFFICULT to interpret alarm lists/printouts after trips because when a unit is tripped by the Mark* they don't block any subsequent trip indications. For example, quite often when a turbine is tripped, say for high-high L.O. temperature, the fuel stop valve may take a little longer to close than the Mark* thinks it should take. Or, the Trip Oil (or Control Oil) pressure may take a little longer to bleed down than the Mark* thinks it should take. And, it thinks this is a reason the turbine should be tripped--when it was already tripped previously on high-high L.O. temperature. And, it annunciates this "secondary" trip to the Alarm Display and the Alarm Logger (alarm printer). And, this might happen for one or two or more other conditions during the coast-down, and they all get "alarmed." And, that can make it difficult to understand what was the FIRST condition that resulted in a turbine trip. <b>BUT,</b>, it doesn't make it impossible to determine exactly what tripped the turbine--it just makes it a little more difficult than it should be.

How can you be certain the turbine tripped on exhaust overtemperature if there was no alarm indicating high exhaust temperature or an exhaust overtemperature trip when the trip occurred??? You say that during subsequent start attempts the Mark* is annunciating high (excessive) exhaust temperature and is tripping the turbine on exhaust overtemperature--so that indicates that sequencing is working properly. So, why didn't it work when the unit was running (for approximately two years)? It's just not logical to assume the unit tripped on exhaust overtemperature if there was no Process Alarm to indicate that was the reason the unit tripped.

If all three (3) P2 pressure transmitters were failed--there would have been a <b>BOATLOAD</b> of Diagnostic Alarms just related to those three failed transmitters that weren't being attended to. <b>AND,</b> the SRV would have been 100% open and told to go even more open by the servo current being applied to the servo coils. Which means, among other things, the P2 pressure was uncontrolled and higher than it should be (presuming the gas fuel supply pressure was normal (at least a couple of barg higher than expected P2 pressure at rated speed). If there was a spike in gas fuel supply pressure the Mark* would not have been able to respond to it, because the P2 pressure transmitters were not working AND because the servos were trying to drive the SRV even more open (supplying saturation current to the servo coils--ALL THREE OF THEM!) and even if the P2 pressure transmitters had been working it would have taken some time (a second or more) for the SRV to actually close. So, a spike in the gas fuel pressure would just make it's way through the SRV <i>AND</i> the GCV and into the combustors, and that could have caused a high exhaust temperature (which would have been annunciated with a Process Alarm!). And if high enough would have resulted in an exhaust overtemperature trip. (We know that bit of sequencing works--because it's working when you're trying to start the turbine!).

AutoCalibrate has a manual positioning ("stroking") feature which can be used to check, or verify, device operation and LVDT feedback versus actual, measured physical position ("stroke"). If you used a User-Defined Display to stroke the SRV, that requires changing the SRV I/O Configuration regulator from a Type 77 to a Type 43 (from a pressure loop to a position) loop prior to stroking in order for the valve to go to and maintain a reference position during stroking. And, that requires downloading to the three control processors, re-booting them, stroking the SRV, then changing the regulator type back to a Type 77, downloading to the three control processors and re-booting them. Did you do that? If not, when you put a SRV manual position reference value into the User-Defined Display it will just go wide open--because it's really looking for a P2 pressure and there is (should be!) no P2 pressure, so the valve just goes wide open against the fully open mechanical stop no matter what the position reference value is. Pretty hard to verify operation unless the SRV regulator type is changed when using the User-Defined Display method of stroking the SRV. (AutoCalibrate automatically reconfigures the SRV regulator type--making it much more convenient and easy to stroke the SRV.)

If the IGVs stroked (moved) when using the User-Defined Display, then they should move when the unit is accelerating--when the unit reaches the appropriate speed. I can't recall if that speed is 68-72% or 78-82% (and I don't have any Mark V documentation or CSP to look at at this writing)--but it's somewhere around there (around 70-80%).

There's some kind of serious leak in the turbine compartment (and other compartments?) causing this "high heat emissions." And, it could be serious enough to contribute to the unit's inability to accelerate.

If TTXSP1 is 700 deg C during starting and acceleration, then something's seriously wrong with the flame in one or more combustors. If the unit has only two, or even four, flame detectors, the flame in those combustors may be fine--but unless some of the exhaust T/Cs are still reading negative (which shouldn't affect the spread calculation)--one or more of the combustors doesn't have flame and that means two things: not enough hot combustion gases getting into the turbine section to help with acceleration, and unburnt fuel getting into the exhaust. Neither is good for the turbine.

You mentioned shutting the unit down by closing an emergency shut-off valve in the gas fuel supply. That should be resulting in a "LOSS OF FLAME TRIP" Process Alarm--again, because flame would have been lost before the Mark V detected a trip condition.... Is that happening?

I made a presumption about the value of TTXSP-1. That signal name is unusual; it's usually TTXSP1. And, if you can't find the other two signals (TTXSP2 and TTXSP3) using a User-Defined Display or the Logic Forcing Display, then it's possible and likely that somehow the Mark V was programmed to calculate the difference between two exhaust T/Cs, or somehow it chooses the highest and the lowest and than calculates the difference between the two.

If the Mark V is using one portion of the Combustion Monitor Big Block it would automatically reject any value less than approximately 400 deg F (if I recall correctly) when the unit is running. This prevents negative (failed) exhaust T/Cs from affecting the spread calculations and adjacency checks. If there is unique sequencing to calculate TTXSP-1 then it's possible it's not rejecting low readings as it should if using the Big Block.

It would be great to know how TTXSP-1 is being calculated at your site. If it's a unique bit of sequencing (that is, it's not being done with the Combustion Monitor Big Block), then it's possible there is flame in all the combustors and what I wrote about no flame in one mor more combustors isn't correct. Without being able to see the CSP running in the Mark V at your site, it's really difficult to tell.

Do you know when this Mark V was commissioned? Was it provided with a new unit when it was commissioned--or was it provided as an upgrade to an older Mark* Speedtronic turbine control system?

It's the average exhaust temperature calculation (TTXM) that excludes exhaust temperatures below a value--<b>NOT</b> the Combustion Monitor. (Excluding ANY exhaust temp T/C value would defeat the purpose of the Combustion Monitor. That's why when (typically) more the third exhaust temp T/C fails (low) the unit is tripped.)

Sorry for any confusion!!!