Gas Turbine Flame Detection


We are having GE Frame 6B DLN Gas turbines. The original flame scanners (Reuter Stokes) which came with these turbines were facing repeated fluctuation problems in their readings and these fluctuations utlimately were becoming so low that we had to take turbine shutdowns to clean the scanners. We usually have high humidity in our environment and according to some documents, the high humid air finds a cooler surface on the scanner glass in the combustion chamber and thus condenses causing foggy conditions on the glass resulting in flame reading fluctuation. This cooler surface is present because these GE scanners are provided with a water cooling provision so that they can survive in a temperature excess of 200 degC.

With this in mind, we installed a 3rd party scanner with remote electronics so that we do not have to do any water cooling on the scanner. The hot end of this assembly is rated for 350 deg C, and through which a HT rated fibre optic cable runs to a cool end electronic enclosure which converts the reading to 4-20mA and provides to Mark VI control system.

We installed this scanner as a trial as one of the four secondary scanners. The scanner test proved to be very successful and we observed, during operation of the turbine, that it behaves well in comparison to the other secondary scanners and we found no lagging response when the secondary flame was ON and OFF.

We need to do a risk analysis as whether using these 3rd party scanners will compromise machine safety in any respect or not. I see immediately two main risks here:

Risk#1) What if the 3rd party scanner does not detect the flame?
If this happens then the machine will simply trip or will not start and the machine safety will be intact. What do you think?

Risk#2) What if the 3rd party scanner false detects the flame (means the flame is not there but even then it detects)?
This is a debatable scenario. I need your help here. Please let me know if there is any other way in machine parameters which can confirm that the flame is there or not. How machine can be saved in this regard?

I'm sure that the moment we will discuss this matter with turbine OEM, they will take it as a business case and will try to convince us that using non-OEM scanners will pose machine to the safety risk. But I want them to come to technical terms. If my scanner false detects the flame, how my machine safety can be still ensured. By the way this will be very rare case (almost unimaginable) as I have told you that we have tested this scanner on the same machines. But still, can you help me in figuring out any further risk which may get raise?

I don't know the details of turbine combustion, but in general, I have to ask,

1) is the replacement technology rated for continuous duty?

Some UV detectors will latch "ON" (fail un-safe) after continuous exposure to a flame. A mechanical shutter mechanism is used to block the UV and the electronics "proves" that the detector (correctly) drives to an OFF state in absence of flame exposure.

2) Does the replacement flame sensing technology match the original flame sensing technology?

Some flame sensor technology senses flame flicker, to distinguish a real flame from heat and emissions from the combustion chamber.
Dear David

yes it is rated for continuous duty. It has many references in GE turbines installations. It senses the flame flickering frequency to detect the actual flame. As i said that we have testedby operating gas turbine in different modes where flame come and go and every time it responded well.

If there is humidity condensing on the Reuter-Stokes Flame Trakker lenses, the problem is most likely that the cooling water flow to the sensors is not being properly regulated. There are supposed to be needle valves in the cooling water discharge tubing which are to be adjusted to control the flow of cooling water to limit condensation on the sensor lenses.

Is this "difficult?" Well, it's not easy, because there's no way to know at what cooling water flow the condensation decreases--except by trial-and-error. Many sites have started the unit, transitioned to Premix Steady State and then adjusted the needle valve until a solid flame intensity is achieved. Then by increasing the cooling water flow it can be relatively quickly determined when condensation is forming because flame detector intensity begins to decrease. Then, the needle valves are opened slightly from that point and locked in place (they usually have--or should have--some type of jam nut arrangement which can be used to keep the needle valve from moving due to vibration.

Now, there are MANY sites which have switch, successfully, to the type of flame sensor you are describing. I'm not aware of any which have regretted the decision. Yes; if more than two secondary flame sensors fail to detect flame then the unit would be tripped--which is what the logic says should happen in the Mark* turbine control system.

A few sites have found that by using a little CPD flow to the tube/pipe the Flame Trakkers are screwed to that this helps prevent condensation from forming on the sensor lenses. A small (1/8-th inch) weld-o-let is welded to a small hole drilled in the flame sensor mounting pipe, and then tubing (MORE tubing in the turbine compartment!) is used, along with a needle valve, to allow a small flow of CPD into the sensor mounting tube/pipe, preventing moist air from getting to the sensor lens. The angle of the hole drilled into the mounting tube/pipe should be such that it directs the flow of air to the combustor, not towards the sensor.

The question about flame sensing technology is best answered by the manufacturer of the flame sensor you are considering buying--especially since you DID NOT provide the name of manufacturer (but you had NO PROBLEM naming the OEM flame sensor manufacturer). Reuter-Stokes Flame Trakkers use SiC (Silicon-Carbide) technology for flame detection in a particular UV range. The other commonly used flame sensor on GE-design heavy duty gas turbines is manufactured by Honeywell and uses a different type of UV sensing technology (search for Geiger-Mueller flame sensing technology)--in the same UV range. The flame sensor manufacturer you are testing should be able to tell you the UV range of their sensing technology and they should also be capable of answering this question.

When flame is lost in a GE-design heavy duty gas turbine CPD drops very sharply and suddenly. There have been tests over decades trying to devise a more precise method of "flame detection" using exhaust temperature (which also drops quickly) AND CPD rate of change (in conjunction with each other). But, those tests don't seem to have been very successful, though I'm told other GT OEMs do something similar with some of their machines (smaller units; much smaller, though).

Look, in the end, if the units are not under any kind of manufacturer- or LTSA (Long-Term Service Agreement) warranty you, as the owner/operator, are free to use whatever technology you deem necessary and reliable--after doing due diligence. You might ask the OEM, but it's pretty certain they are going to tout the effectiveness of the sensors they provided (Reuter-Stokes, is owned by GE... or was; I don't know if it was part of Baker-Hughes or not...?). And, there are--or should be--needle valves which have proved serviceable for decades when properly adjusted and maintained (they can become choked over time with poorly maintained cooling water).

Hope this helps!
Dear CSA

You may be well aware that cooling water tubing poses an additional risk of leakage when used with these flame scanners. Secondly, GE has a TIL in which they suggested to use a larger bore of anti condenser nipple which will drain any collected moist air outside of the flame scanner. But this modification did not work well with us.

The new scanner which we have tried on our machine (remote electronics type) is from a German manufacturer called BFI Automation. The range of UV which we have selected is 270nm to 450nm which is a narrow band that covers the turbine hydrocarbon peak output of 310nm.

Nowadays, I know that GE also offers a similar remote electronics type flame scanner called Flame Tracker Dry 325. Do you have any experience with its installation? Was it able to resolve the famous fluctuation issue?

As far as I know, GE presents a solution in which they ask to replace the existing Primary Scanners with Flame Tracker Dry 325 and replace all secondary with synthetic sensors which means through logic and no hardware will be required for secondary. Anyone has any experience in such solution, please let me know?

But this solution is way more costly. That is why we opted for a well reputed 3rd party sensor.

Thanks for the information about the CPD rate of changing. This is what I was looking for as Im going to sit with GE and discuss total technical aspects that why they think that using a 3rd party scanner will compromise machine safety.

There are over 100 threads Flame-Detection files in the original, but I'm at a loss to explain how to extract them!
Regards, Phil Corso
GTEngineer... Sorry for the confusion !
The original site used 12-digit numbered threads. Using the Magnifying-glass icon, reveals 141 threads ! Unfortunately my "library" lists them using the original 12-digits !
Perhaps I can help if you provide the logic you seek. For example... 1 out of 1, 1oo2, 2oo3, 2oo4, m of n ?
GTEngineer... I suggest a search for Thread # 1239015160, "Boiler Flame Failure" !
It discusses the catastrophic explosion of a 16-burner boiler with 32-flame detectors, because of the wrong logic !