Flame not stable at startup Frame VI GT on gas


Thread Starter


During the start-up of GT on gas (frame 6B with Mark V control) flame appearing in all four flame scanners A, B, C & D and immediately disappearing flame in A & B (Can 3 & 4). after the igniters off, again it will appear after 30% speed. C & D is ok all the time.

This problem started just after the MI and same problem appeared during the subsequent four start-up on gas. Other than this problem unit is running without having any problem. Spread and vibration all are normal.

What could be the reason? This GT is having two igniters in Can 1 & 10 and the Igniter ON time is 10 sec on gas start-up.

Thanks in advance for all reply.

Flickering flame indication after establishment of flame is typically caused by low fuel flow-rate when the Speedtronic cuts back to warm-up FSR after flame is detected. Firing FSR is higher than warm-up FSR, precisely to help establish flame. And once flame is detected, the fuel is cut back to try to reduce the thermal stresses on the internal hot gas path parts for approximately one minute (warm-up) and then fuel is increased to help with acceleration. If the problem isn't too severe, as fuel flow-rate increases the flame will re-establish itself, as you are experiencing.

Was the Gas Control Valve disassembled and refurbished during the outage, resulting in a need to recalibrate LVDT feedback prior to start-up? If so, how does the most recent calibration (position vs. voltage) compare with any previous calibration? In other words, do you have records of previous calibrations, with actual physical positions and voltages, to compare with the calibration done after the refurbishment of the gas control valve?

Or, did someone just "calibrate" the Gas Control Valve feedback as part of the outage activities? If so, did they record the as-found position vs. voltage readings before calibration, and have you compared them to the as-left calibration data? Can you compare it to previous Gas Control Valve LVDT calibrations?

Because the biggest cause of this after a maintenance outage is related to how the LVDT "calibrations" were performed, and a comparison of the current data to previous data will usually reveal the problem.

Now, if the internals of the gas valve were replaced or resurfaced/reseated, it may be necessary to make some adjustment to the warm-up FSR Control Constant (increasing it slightly), and maybe even the firing FSR Control Constant, though it seems the unit is able to establish flame easily enough (you didn't say how long it takes for flame to be detected after the SRV and GCV are both open and the ignitors are energized).

What are the ambient conditions this time of year at the site? Are they warmer or colder than normal? Colder ambients will require slightly more fuel flow because of the increased air density (colder air is denser), which might also help to explain some of the issue.

But, if the unit is establishing flame in all combustors during firing, then losing flame in a couple (or more; the exhaust temperature spread will be very high during this period depending on exactly how many combustors lost flame: just the two with flame detectors, or possibly more), and then re-establishing flame at about 30% speed, that's not really too hard on the machine. The desirable thing, and it should be easy to achieve when starting on gas fuel, is to maintain flame in all combustors during firing, warm-up, and acceleration, but it doesn't always happen and sometimes even ambient conditions can affect that if the Control Constant values were "on the edge" to begin with.
excuse me how you know 10 second as i know the ignitor establish the flame for 1 minute for all cans? also the flame transfer to other cans by crosses tube.

this i know info. i hope to share in your topics

Many thanks

maint makes a good point. Typically, the firing timer is set for 30 to 60 seconds, during which time the ignitors (spark plugs) are kept continually energized just in case there are low or intermittent fuel flows there is continuous spark available in the two cans with ignitors for duration of the firing timer.

Also, flame is propagated from the two cans where the ignitors are located through the cross-fire tubes. The pressure in a can where flame is present is higher than in an adjacent can where there is no flame. Hot combustion gases flow through the cross-fire tubes into cans where there is no flame, until all cans are lit. When all cans are lit, the pressures in the cans are relatively equal and there is no flow through the cross-fire tubes. (There are two ignitors in two cans for redundancy purposes, in the event that one ignitor is not working, though there is no way to determine one ignitor has failed when the unit is firing.)

It is important to note that while it is generally believed that if flame is lost in a combustor when the unit is operating at rated speed that hot combustion gases flowing through the cross-fire tubes from adjacent cans where flame is present would re-ignite the flame in a combustor where flame has been lost that this does not generally occur. At rated speed the air flows through the unit are so high that the air flowing past the open ends of the cross-fire tubes effectively prevents hot gases from an adjacent combustor from igniting the fuel, as well as acting to greatly cool the hot gases exiting the open end of the cross-fire tube. When this happens when the unit is running, there will be very high exhaust temperature spreads, and the cross-fire tubes can get so hot they actually "glow", turning red hot. This is a very bad condition and can result in catastrophic failure and loss of life.

However, maint, not every site is the same and there are sites at which, for some unknown reason, the ignitors are not energized for the typical 30- or 60 seconds and are de-energized shortly after flame is detected. It mostly depends on the packager of the GE-design heavy duty gas turbine or the requirements or dictates of the site- or commissioning personnel. But most sites do keep the ignitors energized for the duration of the firing timer (K2F, or 2F).
During the MI outage we did only calibration, verified command and position feedback, not compared with the old values. Gas Control Valve and Stop/Speed Ratio Valve not dissembled during this MI.

The following are the existing control constants

K2FGAS 10 sec
K2FLIQ 30 sec

FSKU_FI 17.6 % for gas
FSKU_WU 11.3 % for gas

Flame detected immediately when the igniter on, initially there was no delay for detecting flames in all four detectors. Ambient temperature was 27-28 Deg C.

This is happened after the MI. Is there any thing can wrong in the combustion side or cross fire tube? gas fuel nozzle? or in the can 2 & 3 where the flame detectors are fixed?
This is a <b>PERFECT</b>example of "calibrating" something that likely didn't need calibrating, and screwing things up in the process.

When someone goes out during a maintenance outage to "calibrate" a working pressure switch, the first thing they do is record the as-found condition of the switch. They apply pressure and record the changes of state as they increase the pressure, then decrease the pressure. Same for a temperature switch, or a limit switch, or a pressure transducer or a temperature transmitter.

If the pressure switch (or other device) functions as it should, changing state at the specified settings within the expected tolerance or producing the proper output within tolerance, then that person usually doesn't do anything else to the device other than apply a "calibration" sticker to it! The device has been checked and deemed to not require any adjustment; in other words, it's "calibrated."

Or, if the setting needs to be adjusted slightly, an adjustment is made and then the operation is checked and the as-left condition is recorded. In the worst case, if the device can't be made to function correctly within the desired tolerance, then it's replaced.

Which is the primary function of "calibrations" during maintenance outages in the first place: To identify devices that aren't working properly, adjust them if possible, and replace those that don't work or can't be adjusted so that when the unit is running we can have a high degree of certainty that the devices providing us with data are working properly. But, I'd venture to say that the majority of devices tested during maintenance outages don't require any adjustment at all. (Sorry, controls guys! Your secret is out!)
So, why, on GE-design heavy duty gas turbines do people feel that it's necessary to calibrate LVDT feedback without first checking to see if the LVDT feedback is out of calibration???? Without first comparing the feedback to actual physical, measured position????
A pressure switch can tell us if a pressure is normal or not, and a pressure transmitter can tell us exactly what the pressure is at all times.

A limit switch can be used to tell us if a valve is open or closed, or even in mid-stroke. An LVDT is a device or instrument for indicating physical position through the entire range of travel of the device it's attached to.

An LVDT is not a magical device or instrument. It's a device that produces an output that is proportional to physical position. And that output must be calibrated to accurately indicate physical position.

It's simply amazing the number of people that will just accept the LVDT feedback on a GE heavy duty gas turbine as being 100% correct for the position of the device it's attached to, without ever physically verifying the feedback with the actual position.

If the machine was running fine, and there was no indication that anything was wrong with the calibration of the LVDT feedback, why change it? And why "calibrate" it without first checking to see if it's not calibrated properly by recording the as-found condition?

Because, people <b>almost never</b> physically measure strokes and positions when calibrating LVDT feedback on GE-design heavy duty gas turbines!

Why not? Because, the Speedtronic is just that damned good, I guess. The Speedtronic automatically knows what the physical stroke of every device is, and it knows what the 100% stroke (effective stroke) position from the Control Specification of every device is, without being "told".

I give up, folks. I surrender. The Speedtronic is a good, even great, turbine control system. But it's not this good, or even this great. And, yet, turbines still start and run. Though, as this thread will attest to, they don't start and run as well all the time.

Now, the controls guy is blaming a loss of flame at low speed on mechanical problems.

If the unit gets to full speed-no load with flame in all the combustors, without high exhaust temperature spreads, and then produces torque without any high exhaust temperature spreads and with flame in all the combustors, then it's not likely a mechanical reassembly problem.

It's more likely that the SRV and GCV "calibration" ain't right.

It could be that the fuel nozzles aren't flow-matched properly, or that the combustion liners aren't flowing exactly the same amount of air through the head end or through the cooling and dilution openings.

But if the only loss of flame and high exhaust temperature spread is occurring at very low speeds after initial firing and then during acceleration if flame re-establishes itself in all the combustion cans and the exhaust temperature spread goes back to normal, then it's likely a self-induced controls problem. Pure and simple.

Hell, if flame is not indicated in two of the four combustors with flame detectors, that doesn't mean it's only out in those two combustors! It could be out in any of the other six combustors, as well. (It's just that there aren't flame detectors in all ten combustors, so the other six aren't usually considered.)

And, see, this site only fires for 10 seconds on gas fuel; thirty on liquid fuel. Not typical, but, not abnormal, either.

And, fuel is being cut back by more than one-third during warm-up from the firing value. That seems a little excessive for most applications. And probably too excessive for this current LVDT "calibration".

Which must be correct, though, because the Speedtronic is just that damned good that is just "knows" the actual physical position matches the LVDT feedback for the effective stroke value in the Control Specification.