Turbine Tripped Due to Loss of Flame


Thread Starter

Sanjay Mallik

GE gas turbine tripped due to loss of flame after purge complete during startup. It has been observe that moisture and water found in the flame detector every time it tripped. After cleaning the water content from flame detector, the unit then detect firing and sustain. But every time after unit started this loss of flame tripping is occurs. After purge complete GCV and SRV also opening, and FSR also shows 20% but firing restricted as flame detector couldn't detect flame due to moisture deposited on the surface, and every time it need to be cleaned. The unit control system is Make-VIe. What might be the cause of water and moisture on the surface of flame detector?

Thanking you.
[email protected]
This is NOT a Mark VIe problem. This is a temperature/airborne moisture (humidity) condensation problem. And, if the flame sensors have throttling valves to adjust the flow of cooling water to the sensor cooling water jackets then it's also a mis-adjustment problem (mis-adjustment of the throttling valves). This isn't even a controls-related problem--it's a temperature/airborne moisture condensation problem.

I'm going to skip right past the questions about exhaust temperature spreads and the Frame size of the unit and the type of combustion system in use on the unit (because we are clairvoyant and can telepathically intuit all of this information over the Ether(net)).

Flame sensors (we don't know if they're UV or SiC) with water cooling jackets that are improperly valved-in usually suffer from these kinds of problems, particularly in humid environments (another thing we have to intuit). Usually, there are valves in the cooling water outlet (sometimes in the supply) to each flame detector cooling water jacket that are <i><b>supposed</b></i> to be adjusted to limit the amount of cooling water flow to prevent moisture condensation on the flame sensor lens(es). How to adjust the valves? It's a trial-and-error thing, but generally the valves should not be fully open, at least not until after all the air has been bled from the system supplying the flame sensor cooling water jackets.

If your flame sensors are not water-cooled, then you are going to have to find the source of moisture condensing on the flame sensor lenses. And, you're going to have to find how the flame sensors are getting cool enough to cause the condensation on the lenses.

Without LOTS more information there's not too much more can be said.

If the cooling water system is left running after the unit is shut down in a humid environment, then ambient humidity will be drawn into the unit during cooldown (natural draft created by the hot exhaust/stack) and will naturally condense on flame detector lenses which are too cool. Or, if the cooling water system is started before a turbine start in a humid environment then these kinds of problems can occur.

If the source of airborne moisture is NOT ambient air, and the unit has some kind of inlet air cooling system (evaporative coolers; chillers; foggers; wet compression) then there is something amiss with the configuration and operation of the system(s) during or after a unit shutdown that is allowing a large amount of water to evaporate and make it's way into the draft through the unit and condense on the flame sensor lens(es).

Please write back to let us know how you fare in resolving the issue--and if you require more assistance, please provide lots more information about the unit, it's combustion system, and the locale where the unit is being operated.

Sanjay Mallik

Flame sensor is UV and is not water cooled, then what might be the source of moisture and water? kindly give some view.

Thanking you
>Flame sensor is UV and is not water cooled, then what might
>be the source of moisture and water? kindly give some view.

Ambient air is drawn into the axial compressor and flows through the turbine and exhaust. As was previously written, if the unit has some kind of inlet air cooling, and it was not operating or being operated properly then it's possible the moisture source could be the result of improper operation of the inlet air cooling system(s).

If the unit burns liquid fuel it's possible that the atomizing air pre-cooler is leaking cooling water into the atomizing air system and that is the source of the moisture,

OR, if the atomizing air temperature regulating valve is not adjusted properly or is not working properly then the axial compressor discharge air could be cooled excessively causing the moisture in the ambient air to condense in the cooler shell.

EITHER problem can be checked by opening the pre-cooler shell drain and observing if water flows out of the drain. If the water has coolant in it then the tubes are leaking. If not, then the temperature regulator is not working properly or is not properly adjusted. Usually there are low point drains in gas fuel and atomizing air piping that can also be opened to check for moisture in the piping.

It could be this problem started after an off-line compressor water wash and the piping low point drains were not opened properly to drain moisture out of the piping during or after the rinse,

If the unit burns natural gas fuel and there is a natural gas compressor supplying gas to the unit it's possible the gas compressor has a cooler that is leaking coolant into the gas fuel, or that moisture is entrained in the gas fuel flowing into the unit and is condensing in the unit (a lot of gas fuel supplies are not very "clean" these days and if the unit is located close to the gas source (wells) that could be a possible source of moisture).

But, yet again, you have not provided any useful information other than the flame sensors are UV and they are not cooled.

BUT, that doesn't change the problem of moisture in the airflow through the unit is somehow condensing on the flame sensor lenses and preventing the sensors from detecting flame intensity. And that analysis is based entirely on the information that was provided in the original post: that water was found on the flame sensor lenses after a loss of flame during acceleration. We have no idea what Frame size the unit is, what the exhaust temperature spreads were or what the exhaust temperature was immediately before flame was "lost", we don't know what part of the world the unit is located in, we don't know anything about the ambient temperature or humidity during starting (it could be monsoon season or the unit could be on on off-shore platform or on the shore of an ocean or lake), we don't know what the alarms (process- and Diagnostic) were prior to and during the start attempts, we don't know the type of combustion system (conventional (diffusion flame) or DLN (Dry Low NOx), we don't know know what fuel is being burned, we don't know when the problem being reported started (has it been ongoing since commissioning, or did it start after a maintenance outage or a trip from load), we don't know how old the unit is or what condition the inlet duct is in. There's just a lot we haven't been told.

But, based on what little information has been provided (and it's very little information), the Make VIe is not the cause of the problem (even if the Make VIe was recently installed to replace an older turbine control system), and the moisture found on the UV flame sensor lenses is condensing on the lenses. The source of the moisture has to be determined on site based on the unit, type of fuel and combustion system, ambient conditions, condition of machine and auxiliaries, source of fuel, etc. Every gas turbine is not like every other gas turbine. You asked what could be the source of moisture? We offered the most likely, but then you say the flame sensors are not water-cooled. So, other possible sources have been suggested.

Please write back to let us know what you find. And if you require more assistance you will need to provide much more information, including the Process- and Diagnostic Alarms prior to and during the start attempts. You will need to tell about the fuel being used during starting, the ambient conditions during starting, how recently an off-line water wash was performed, the exhaust temperature and exhaust temperature spreads immediately prior to the "loss" of flame, etc. But, based on the information provided moisture is condensing on the flame sensor lenses causing the water being found on the lenses. If the amount of water is excessive that could prevent the sensors from detecting flame. That's all that can be said based on the information provided. Full stop. Period.

sanjay mallik

GE gas turbine is frame-V and during the flame loss the temperature was around 110 degree and ambient temperature is 34 degree. Combustion system is conventional diffusion flame type. Fuel used is natural gas, and no leakage of coolant from gas compressor cooling system as other unit was running from the same gas compressor manifold. Offline water wash was done two years ago. This problem occurs after the modification of flame detector which is 30 Vdc during the up-gradation of make-VIe two years ago. In other unit 350 Vdc flame detector is being used, and such problem doesn't occur frequently during monsoon season in G.E gas turbine having mark-IV control system.

I can understand that loss of flame is not related with control system. This is just for your information.
Some flame detectors require air purges to protect the lens assembly, If that is the case in your unit, make sure you are using instrument air (dried air), and not the plant air.

We've had flame detection devices in other services that required filtered instrument air not only to cool the electronics, but two air purges on the furnace side to protect the lens assembly.

Single port purges to protect the lens can actually draw in furnace gases and moisture into the lens...
sanjay mallik,

So, when the turbine control system was upgrade from [Mark* ?] to Mark* VIe you say the flame detector was modified. Mark VIe turbine control systems are capable of powering <b>335</b> VDC and 4-20 mA flame sensors (the former are UV type and the latter are usually Sic type). The UV type flame sensors would connect to the TRPG terminal board, which would get the nomimal 335 VDC voltages from PSFD I/O Packs, usually mounted very close to the TRPG.

SiC (Silicon-Carbide) type flame sensors, usually manufactured by GE-Reuter-Stokes, would be connected to 4-20 mA inputs of the Mark VIe. That might be on one or more of several terminal boards which accept and power 4-20 mA inputs.

It's my understanding that the Honeywell flame detector interface module used in the Mark IV produced it's own nominal 335 VDC, and also "interpreted" the feedback to produce relay outputs (contact closures) when flame was detected. The Mark VIe, on the other hand, "interprets" the UV flame detector feedback as flame detector intensity (same for the 4-20 mA flame sensors) and that once the intensity exceeds a certain level a logic signal is set (goes to logic "1"), and when the intensity drops below a certain level the same logic signal is reset (goes to logic "0").

Yes; there might be a slight difference in the UV flame detector voltages of the two control systems, and probably in the range you noted (somewhere between 318 VDC and 352 VDC, according the GEH-6721, The Mark VIe System Guide, Vol. II).

But, many Mark VIe turbine control system upgrades were designed to work with existing flame sensors (if the Customer didn't want to upgrade to R-S 4-20 mA SiC flame sensors). So, the SAME UV flame sensors could be used with minimal wiring changes in a Mark VIe turbine control system as in older Mark* turbine control systems.

Again, this is a search-and-find mission. And, again, you have chosen to provide very little additional information, doling it out in dribs and drabs. Does the unit have any inlet cooling system(s) (evaporative coolers; foggers; wet compression)?

Has anyone examined the floor of the inlet duct, beneath the bellmout, after the unit is shut down, to look for puddled water? The water could be coming from a malfunctioning inlet cooling system, or from leaks in the inlet duct work.

Has anyone checked the flame sensor enclosures to see if they get wet during rainstorms? Usually, the flame sensors are mounted in special "boxes" which keep them from having to be mounted inside the turbine compartment, in part to protect them from the heat of the turbine compartment. These boxes usually have an expanded metal cover to allow for cooling, and that is open to atmosphere. Could the monsoon rains be causing the flame sensor assemblies to be cooled much faster than the turbine/combustors, and cause moisture drawn into the unit by the draft created by a hot exhaust and stack? (Immediately after a turbine is shut down or tripped there is usually a pretty good air flow caused by the temperature differential due to the residual heat in the exhaust and stack, and this draws ambient air (and humidity) into the axial compressor, through the combustion section, and the turbine section and through the exhaust diffuser into the exhaust duct and stack. So, a LOT of ambient humidity can be drawn into the unit, and if the flame sensors are at a colder ambient than the turbine compartment, that could cause moisture to condense on the flame sensor lenses.)

And, because the unit fires and then trips the combustion section is warmer, and after you dry the flame sensor lenses and reinstall them and re-start the unit there isn't much humidity or moisture to condense on the lenses--and so the next start attempt after drying the flame sensor lenses is successful.

Natural gas has a dewpoint temperature, and when gas flows through an orifice (like a control valve or two, and a fuel nozzles) each pressure drop results in a temperature drop, and that can cause moisture to condense. Usually, the condensation is natural gas liquids, but if the natural gas isn't clean and "dry" then it can also be moisture--similar to humidity.

We don't know what the conditions of the two units at your site are. Meaning, were they both installed and commissioned at the same time? Are the enclosures, including the inlet duct work (turbine air inlet filter house and duct work the air flow through to the bellmouth) in good condition? Has anyone gone inside the filter house and duct work during a bright sunny day and looked for rust holes which would allow sunlight (and rain!) to enter the duct work and eventually get into the axial compressor and turbine?

And, again, are the mounting arrangements for the flame sensors the same for the two units?

We can only provide so many tips and hints based on the (limited) information provided.