Mark6e (MK5 migration) Exhaust Thermocouples Read Low

2x1 Combined Cycle, 7FA/D11 with Marke 6e (Mark5 migration)

While checking exhaust temperature thermocouples we found that all readings on one unit are low ~20 Deg F. We disconnect the Thermocouple at the Mark6e input card (TBQA) and inject 1000 Deg F with a simulator and get readings of 980 Deg F. I suspect an error with cold junction compensation since all the readings are low ~20 Deg F. On another unit we get the expected 1000 Deg F reading.

Both units show close cold junction values (2 different locations) around 78 Deg F.

I'm not sure, but it looks like the cold junction compensation is done on the terminal board, TBQA. So I may change that but wanted to ask to see if anyone has seen this?
The cold junction compensation in the Mark V was done on each TCQA card (not a typo). I'm not sure if it's done in the Mark V Migration on the equivalent of the TCQA IR if it's done in the I/O Pack(s) of the TCQA equivalent (I think the TCQA equivalent has an I//O Pack; early Mark V Migration prototypes I saw had them).

The TBQA equivalent is common to all three <Q> processors. But I believe it is individually cabled to each TCQA equivalent--just as was done in the Mark V for each TCQA's group of exhaust T/C inputs.

The big questions though are:

1) What Diagnostic Alarms are active for the TCQA equivalent cards/packs: and,

2) When did this problem start?

Finally, the I/O terminal boards of the Mark V (such as the TBQAb (not a typo) did NOT have any active components (such as would be necessary for T/C cold junction compensation.

Hope this helps!
CSA, your reply does help. Thank you for all the support!

There are no diagnostics on the MVRA (or in the Mark6e at all- which is nice.) This issue was discovered as part of scheduled testing. Unfortunately there is no known history of this check in the past so this may have been an issue from the beginning (2004.)

From the manual: MVRA is the functional replacement for TCQA and IOMA boards. Ribbon cable connects to the same Mark V terminal boards TBQA/B/C or F.

Functional Description: The Simplex Thermocouple Input (TBQA) terminal board accepts 45-type E, J, K, S, or T, thermocouple inputs, as well as three cold junctions. These simplex inputs are sent through ribbon cables to one of three MVRA boards. Each MVRA processes 15 thermocouples and one cold junction.

Operation: The 45 thermocouple inputs can be grounded or ungrounded. The I/O processor performs the analog-to-digital conversion and the linearization for individual thermocouple types.

I did look closer and found the above info in the manual. Regarding the cold junction, the book states, "as well as three cold junctions" so it does look like the Cold Junction is physically located on or connected to the TBQA. Appendix D for TBQA does show an odd symbol/device on the card with the note 1uA/Deg K so there may actually be a device on the terminal board TBQA for cold junction. However, I no longer suspect the TBQA as there are 3, one for each core, and they would all have to have some malfunction. It does seem that the MVRA is doing the data "conversion and linearization" and since I can see a normal range value for cold junction, on all 3 cores, there may be something funny going on in the MVRA, but even in that case, there are 3 MVRAs. There may be some other config or internal issue. I did verify that the MVRA is set properly, and matches the other unit that is functioning properly.

If you think of anything else, please let me know. Thanks!
It turns out that we were taking the readings from the HMI screen. There is a .98 coefficient from the input to the screen. The inputs were reading exactly 1000 Deg F and the screen was reading 980 Deg F.

I did learn more about my system, how the cards are arranged, where to find some information, and of course this coefficient.

Thanks for looking at this!

GOOD ON YOU for reading the fine manual! The Mark VIe has remote cold junction compensation capability--the ability to sense the temperature at a remote point (other than where the "local" I/O Pack is located). The Mark V did not have that capability/option. I think that's what the excellently-written manual is referring to (there were "extra" terminals on the TBQA in the Mark V, and I think the Mark V Migration designers used those for the remote cold junction compensation inputs.)

I still say the cold junction compensation is accomplished by the MVRA or its I/O Pack or I/O Pack equivalent--NOT by the TBQA. I may be wrong; I've been wrong before. Hopefully demigrog2 will chime in with the real answer (I think he works for GE, at least in some capacity if he's not a full-time employee).

It's still not clear how the unit ran with the exhaust T/Cs reading the way you are describing. I don't know if you are looking at the CIMPLICITY or PROFICY display, or in ToolboxST. Have you tried using a Watch Window or Trend Recorder?

And, you say this is only happening on the T/C inputs to the TBQA or <R>? Not to the T/C inputs to the TBQA on <C>?

Is it possible someone enabled the remote CJ compensation in the affected unit and didn't connect anything to the inputs???? That's about all I can think of.

I'm SOOOO confused<i>!!!</i>
I know zilch about turbine controls, so I'm asking questions.

1. Wrong CJ?

If I understand the correct connection sequence it is: T/C > TBQA terminals > ribbon cable > MVRA

So I have to ask, is the 'system' using the Simplex forwarded CJ measurements at the TBQA terminals or is it using the CJ measurement made at each MVRA board?

If the CJ being used by the 'system' is the CJ measured is at the MVRA board, the error you see could be the difference between the TBQA CJ and the MVRA CJ.

Thermocouple thermometry says that the CJ measurement is the connection where the thermocouple connects to the 'copper' measurement wires. In this case, that should be the CJ temperature measured at the TBQA terminal blocks.

Could the temperature of the TBQA terminal blocks be 20 Deg F cooler than the MVRA CJ measuring point? I don't know. I do know that the heat generated by card racks can keep the electronics toasty warm. But I haven't a clue what the operating temperatures of these types of installations are.

From where the does the 'system' take the CJ it is using? TBQA or MVRA

2. Potential Ribbon cable error?

The statement "These simplex inputs are sent through ribbon cables to one of three MVRA boards. Each MVRA processes 15 thermocouples and one cold junction" brings up the point that each temperature measurement could have false junction errors.

Thermocouple extension wire is the same type of metal materials as the thermocouple itself. Copper is not a thermocouple extension wire except for the noble metals R, S and B, which might use one or both legs of copper extension wire.

Ribbon cables are typically copper. I suppose the manufacturer could have had thermocouple extension wire ribbon cables made for each of the E, J, K, S, or T types that are used. Or they could have gone overboard and produced a ribbon cable with different thermocouple type wire which is even MORE custom than a single type T/C ribbon cable.

But supposing those ribbon cables in use are just conventional copper wire ribbon cables?

Differences in temperature between the terminals for both legs of a thermocouple will introduce an error. The terminals should be "isothermal", all at the same temperature.

I doubt that this is the cause of the 20 degree difference problem which is much more likely to be CJ. The absence of CJ is typically a 35 to 60 degree Low error. A 20 degree Low error could be reading the wrong CJ.

I find it hard to believe one could get multiple temperature points to have a consistent difference of 20 DegF due to ribbon cable connections.

what simulator do you use? some kind of thermocouple injector? Check your injector whether it has cold junction compensation or not. Some injector like CA70 doesn't have internal cold junction compensator and should add external cold junction compensator.

Beside checking in the Mk V side, I suggest you to use injector that have internal cold junction compensation.

Just to share a bit more detail: There are differences in software (and hardware) between the 2 units. GT1 has the GE Opflex software with autotune, dynamics, and model based Ares software. The temperature coefficient, tkxcoef, is .9775. GT2, the unit I initially experienced the difference in reading, has older software, hardware, and 3rd party autotune and CDMS. It has TTKXCOEF at .98. Either way, they were pretty close. GT2 is getting the same software and hardware this Spring.

The self-created learning opportunity occurred when I went to do the original test on GT2. We were looking at the HMI (Cimplicity) screen and that’s when we were getting the .98 difference. I didn’t remember this coefficient existed and thought there was some problem. Then when I wanted to double check the other unit, GT1, I just happen to be looking at the I/O and it was right on.

I now remember hearing about this coefficient years ago but could not find it in my review of what I thought may be a problem. This was still good to run through the manuals, check jumper settings, and cables, and ask for help.

The cold junction values in the MK6e are labelled mvra_cjr/s/t so I would agree that this is done in the MVRA now. The exhaust thermocouples are in the R/S/T (TBQA) terminal boards and I did not check any others (on the C core.)

David_2, I apparently know less about turbine controls than I thought, but am always willing to learn, and re-learnâ Connection sequence is terminal board (TBQA) ribbon cable to MVRA. The original Mark5 did have cold junction done through the card, the migration to Mark6e puts the CJ in the MVRA (processing) card, as CSA described. The issue I had with a possible CJ error is that the 27 exhaust thermocouple are split into 3 and connected to each core, so there would be 3 different cold junction errors causing this. The cold junction is still done in 3 locations, each core’s MVRA, so either way CJ error didn’t make sense, especially since I could see the CJ values and they were close to each other in value and matched another turbine control panel. But I was considering everything at that point. As you can see, the issue was in reading the data at the wrong spot. The input has the value as measured on the thermocouple, as I simulated, but based on stratification testing of the actual exhaust, and since the thermocouple is measuring 1 point in depth, GE has determined the overall corrected value would be 98% of the input. So there is a .98 coefficient in place before the value hits some logic and the operators screen.

briq, the simulator is a Fluke725. I used 2 different simulators trying to figure this out. I checked output of one into the other, tried both on the input, and I also went to another turbine and was getting the correct reading. Again, the problem was self inflicted.

Thanks for all the review on this. If anything it was a good learning exercise!

Ah, the plot thickeneth.

TTKXCOEF has a lot of "history" and mythology behind it. I've never heard a description of TTKXCOEF that I really believe or trust is truly correct.

I do know that it's a "multiplier" that's used to bias the exhaust T/C inputs. They are originally labeled as TTXD1_nn, after they are biased they are put into another array, TTXD2_nn. And, someone decided for some reason when operator interface displays were being created at GE in Salem, VA, USA, that the TTXD2_nn array values should be used for the displays--which causes MUCH confusion and consternation for a lot of installations and sites. (Mostly because there is no consistent and logical explanation for TTKXCOEF--and it gets used for many things by GE engineers (not so much field service people, but factory-, performance- and combustion engineers).

I have seen problems caused at multiple sites with Mark V panels where the output of the control compartment air conditioners was "aimed" directly at the top of the Mark V panel. This caused the cold junction values of <C>, <R> and <S> to be different from each other, and VERY different from <T>. I don't believe the cold junction values are averaged or high- or low-selected, so this can cause issues for sites with DLN combustors, in particular (because the average exhaust temperature (using all 27 scaled and biased exhaust T/C values!) is what's used to calculate FSRT when the unit is operating at or close to Base Load. And if the unit was tuned very close to the emissions limit, or ambient conditions change and cause the unit to be very close to the emissions limit that can cause issues with the real firing temperature (NOT the calculated firing temperature!) and result in emissions exceedances. Simply placing large pieces of cardboard in front of the upper vents of the Mark V panel would result in changes to emissions, and, of course to load....

I have been to sites where the previous tuning or performance test had resulted in TTKCOEF being changed to some value for some reason, and then the DLN tuning people wanted to change it to some other value for another reason--and complained the previous group had made unauthorized or incorrect changes. Such is life in a BIG corporation, where's there's only ONE exhaust T/C bias value.... (There's others, but that's a different story and discussion.)

It's usually a mistake to use CIMPLICITY displays to test Mark* inputs and outputs--for many reasons, but you stumbled across a BIG one. I STILL don't understand why the TTXD2_nn values were reading negative... and why that wasn't noticed LONG ago when the unit was shut down. That is difficult to understand, or explain--with the current information we have. So, while you seem to be satisfied you've discovered the root of the problem, I wouldn't be so content.
I may have added some confusion, I do that... The screen readings (TTXD1_nn) read lower than the input by 20 Deg F (at 1000.) I wrote that they were low by approximately (~) 20 Deg F, not negative (-) 20 Deg F. They were exactly 20 Deg F low, or .98 of the input (of 1000.) Yes, using the readings from the screen is not a good practice for this very reason; I won’t do that again, and glad to share the experience! It is interesting that the 2 units have different values, but that may be a result of different tuners and engineers in the system. It may be part of the new model based control. I'll keep digging to learn why it is .9775 on one unit and .98 on the other. We need to keep a closer eye on what is going on in the system during tuning and other service and gain a better understanding of the different logic schemes going forward.

It is interesting that the new Opflex and Ares Model Based Control does not actually calculate firing temperature, no TTRF. They now use a dimensionless number that is based on GEs long history (of "acquired") models to produce a combustion reference index. It's a 0 to 100 value and kind of correlates to what the old TTRF was but uses modelling to fine tune the firing "reference."
old software uses a single control constant (TTKXCOEF) as opposed to "new" MBC software that employs an array for the exhaust temp coefficient, using control constants TTKXCOEF_Y[1 through 15], that vary as a function of compressor pressure ratio TTKXCOEF_X[1 through 15]