I have read through all I can find and have reviewed the CSP generated by the MKV (TTXMV4 and TTXSPV4). Checked all the control constants associated with the generation of the exhaust thermocouple arrays, average calculations, spread calculations and spread limit calculations. A frame 6 unit suffered an instantaneous single exhaust thermocouple fail high event which resulted in the three spreads being exceeded and a trip occurring. All control processors were healthy, all other exhaust thermocouples were healthy and within a 15F degrees of each other. The rogue thermocouple stepped up to 2046F in 3 seconds, stayed there till the unit tripped then returned to a value similar to the other T/C. Adjacent T/C were stable. The unit operation was stable, flame detection count remained unchanged during the event, as did load. I conclude that the temperature being sensed by this single T/C was false.
I appreciate that the dominant fail mode for thermocouples connected to a MKV is to fail low, but when there is a fail high event this single sensor failure appears predictably to trip the unit. I have convinced myself that this scenario is correct and is part of the GE design, however I would appreciate a sanity check if there are any other views out there. I appreciate that the average is calculated differently with a low threshold acceptance and discarding highest and lowest so it is much more immune to T/C failures. Not so the combustion monitor.

 

sensors do fail, so a bit of trouble shooting may be needed.

 

sensors do fail, so a bit of trouble shooting may be needed.

Thank you for your response Dave. The point of my post was to ascertain agreement or otherwise on the MKV Combustion monitor being relatively immune to T/C which fail low, the dominant failure mode but not immune to a single T/C which fails high.

 

I've always wondered about this, because, if I recall correctly, the exhaust spread algorithm blocks exhaust T/Cs less than 500 deg F when the unit is running (I don't have any CSPs to look at now) which effectively blocks "failed" exhaust T/Cs which go open or much less than the expected exhaust temperature when operating normally. (Usually, even compressor discharge temperature is higher than 500 deg F so even with a complete loss of flame in a combustor the resultant combustor exit temperatures would be higher than 500 deg F).

But, yes, if one rogue exhaust T/c goes VERY high (which as you noted IS NOT the normal predominant failure mode) then there can be bad consequences (such as you, unfortunately, experienced).

I also seem to recall seeing that newer combustion monitor functions (for the Mark* VIe) have some kind of detection for very high, "out of normal range," exhaust T/C readings--but that's probably not available for the Mark* V.

Anyway, I hope my recollection is correct. And doesn't cause any confusion if it is.

 

T/C's don't normally fail high, they may drift over time due to inappropriate material selection.
When they fail abruptly it is typically the result of loss of continuity.

RTD's can fail high depending on the details of the sensor failure.

 

Brilliant WTF. Many thanks for sharing your experience and recollections. Yes MKVIe has a more sophisticated exhaust T/C monitor. The fail high had me scratching my head as this flies in the face of the MKV single point failure immunity tag.
My only comment is that the calculation of average removes any T/C less than 500F lower than second hottest. The combustion monitor does not do this but looks at all T/C which are active.

 

What constitutes "active"?

Yes; the average rejects the highest and lowest as outliers. But, it is extremely unusual for T/Cs to fail high; how many times have you seen it--before this once?

The only reason (most likely) that newer machines have the "exhaust T/C fail high" protective condition is because of trip reduction--and the fact that F- and H-class machines are so damn picky and DO NOT like being tripped unless it's really, Really, REALLY warranted. And, now it's likely being applied across the fleet. And, let's not forget: GE Belfort likes to grossly overcomplicate anything and everything. This is likely their doing (since, as we all seem to agree: it's pretty abnormal for T/Cs to fail high and GE Belfort like to change everything because their way is (yet to be proven) better).

And. let's not forget: GE has ALWAYS been about reliability. In fact, it's a design feature of the control systems that when a T/C input goes open circuit (again--the predominant failure mode of T/Cs....) the signal in the Mark* goes low, negative in most cases. (I think the value of "low" has changed recently....)

Anyway, this reliability thing has been a staple of GE heavy duty gas turbine controls philosophy. Predominant failure mode is part of it; always has been. It wasn't a miss. It was a calculated risk.

And, lastly, how many times have you seen this before?

I'm thinking ... never in my experience. Whenever I've seen exhaust T/Cs failing high--it was intermittent and didn't say high for more than 9 seconds. They burn out very quickly.

 

Thanks again. “Active” I used to describe the TCs that pass from RST via their selection on the TTXMV4.
I appreciate the way of the combustion monitor and its complications and that fail low is the dominant TC fail mode. What really made me pick up on this single point failure trip is that it is the second I’ve seen in 3 months. The first I saw was on another machine. Strange also is that the fail high only lasted about 9 seconds then returned to reading the same as the pack. Critically though it is good to know that a fail high will cause a trip.

 

You said 'pack.' The other occurrence was on a Mark* VIe?--which you said had the newer combustion monitor algorithm....

Were there any Diagnostic Alarms on the Mark* VIe pack?

Okay; I'm done.

 

Oops!!! Both trips were on MkV. I used the word “pack” to mean the rest of the thermocouples. No reference to MKVIe. Sorry to mislead. And no diagnostics from either of the MkV panels.

 

A lot of GE-design heavy duty gas turbines in the Mark* V era used the stainless steel-sheathed exhaust T/Cs which require a lot more patience to replace than most service people have when replacing them. Consequently the sheaths don’t get properly clipped into place resulting in lots of rubbing and worn sheaths which can lead to intermittent shorts and lots of nuisance problems (self-inflicted issues…!!!). In thinking about this post, I am reminded of one of my least favorite sites which had—and continues to have—lots of exhaust issues (spreads and trips). When looking in the exhaust compartment it was no wonder—NONE of the exhaust- or any other T/Cs in that compartment were properly clipped in place and some were severely rubbed through. Added to this mess there were exhaust leaks from the flex seal and the Load Compartment vent fan damper was closed. It was BLOODY HOT in that compartment. So, I’m wondering if the (self-inflicted) issues they were experiencing were the result of the problem highlighted in this thread. They ALWAYS insisted the highest T/C was not a problem because it always went back to normal after the event. And even after the rubs were pointed out to them they still insisted the Mark* V was not working properly.

Garbage in. Garbage out. This ain’t rocket science.

I AM NOT saying that’s what happened in this post. I am saying to others reading this post and experiencing what appear to be similar issues—check the exhaust T/C installation for problem. I’ve even seen damaged Frankenstein flexible T/C leads (caused by mechanical personnel not giving a crap about anything with a wire or wires….

 

T/C fail high, can only occur if you have improper ground connection to the negative lead.
That being the case, you can get ground fault disturbances for example when a motor with faulting windings start.
Check and perform tests of your T/C wiring.

 

The Mark* can accept ungrounded or grounded T/C inputs.

 

Thermocouple failure modes are: Open Circuit, which the Mark V should indicate as a low temperature; or Short Circuit, which offers multiple options:
1, short between the thermocuople leads, which the Mark v would read as the temperature at the location of the short.
2. short between either or both thermocouple leads and some external metal material, either a ground or a, say, copper wire with some voltage with respect to ground. Since the Mark V is expecting milivolts, this might result in a off-scale high reading. I would suspect this or some induced voltage on the interconnecting wiring, although, since the thermocouple wiring is normally bundled with other thermocouple wiring, I think an induced voltage would likely be seen by more than 1 thermocouple at the Mark V.
Regarding the comment that the Mark V can handle grounded as well as ungrounded thermocouples, I don't think that is true for the exhaust thermocouples.