Diagnostic Alarms at time of discovery:

previously in:

0 R VOTING MISMATCH FPG1
0 R VOTING MISMATCH FPRG
0 R VOTING MISMATCH FSRSR

currently in:

1 S VOTING MISMATCH FSGR

The problem that is being reported by those operating the turbine is as follows:

A sharp downward spike in megawatts, followed from seconds to minutes later by a sharp upward spike. This happens intermittently, sometimes in bundles of 3 spread over a few hours. With a day to several between them.

For pretty much all noted occurrences, the turbine has been in speed control.

The problem, based on chart analysis, appears to be a drop in fuel flow caused by a sudden closing down of the SRV. The GCV position signal shows no initial action, though immediately afterwards FSR reacts to stabilize the situation. Pressures before the SRV rises at the onset of the event as the intervalve pressure drops. The flow usually only drops for a few seconds.

These are the values I found on Datalist 15 GAS RATIO VALVE CONTROL:<pre>
NAME MEDIAN UNITS R S T

FPRG1 252.9 PSI 253.3 252.5 252.9
FPRGOUT -4.22 V DC -4.22 -4.21 -4.22
FPRGBAK -4.20 V DC -4.21 -4.18 -4.20
FSGR_v -3.15 V DC -3.14 -5.75 -3.10
FSGR 23.28 % STR 23.29 60.48 23.28
FAGR 2.51 mA 2.51 -4.40 2.69
FPG1 251.9 PSI 253.7 251.9 251.6
FSRSR 4.10 V DC 4.07 4.13 4.10

FSKGRG V/% 0.0699 0.0715 0.0715
FSKGRO V DC 1.51 1.43 1.4

We have 2 LVDTs associated with this valve: 96SR-1 (4A) and 96SR-2 (4B).

According the the actual jumpers on each HSAA card and our Elementary diagrams, each HSAA card plays the following roles

R - Provides LVDT excitation to 96SR-1 [through AIO.1 terminals ATB EA(5)&(6)]
Receives 96SR-1 (4A) output [through AIO.1 terminals ATB EA(3)&(4)]

s - Provides LVDT excitation to 96SR-2 [through AIO.1 terminals ATB EA(11)&(12)]
Receives 96SR-2 (4B) output [through AIO.1 terminals ATB EA(9)&(10)]

T - Receives 96SR-1 (4A) & 96SR-2 (4B) outputs (HIGH SELECT) [through the terminals noted above]

I added the terminal numbers because I took voltage readings between each pair to get a comparison.

ATB EA (3) & (4) / 0.634 VAC
ATB EA (9) & (10) / 0.707 VAC

ATB EA (5) & (6) / 1.18 VAC
ATB EA (11) & (12) / 1.34 VAC3</pre>
Though I was a little confused to not see 7 Volts for excitation, I figured it was due to the 3000Hz and my meter. As a side, any help on understanding this would be appreciated.

However, the outputs from both LVDTs seemed proportionally similar to their inputs.

This leads me to think that it's the HSAA card on S that needs replacement, and not the field wiring or LVDT.

The problems with 96SR-2 along with the HIGH select on T seems to start to explain the previously cleared VOTING alarms on the R computer, as 4B must have been the active input to T at that time, rendering R as the 'odd man out'. Initially that had me looking to the LVDT or field wiring as the problem, or even excitation to the LVDT, as the one LVDT is common to S and T, given the list of DIAG alarms. But the voltages have me thinking that I need to replace the HSAA card on S. So...

Can I shut down S, replace the HSAA, and start it up without a trip? I'm assuming that FSKGRG and FSKGRO are stored elsewhere in the computer... where?

Ideally this would be best done with a shutdown and a calibration. But during the summer heat, with a lot on the line, is having a spare HSAA card (jumpers pre-set) standing by and waiting and hoping for the best the best idea? What can go wrong?

Am I missing something, or heading down the wrong road in my assessment? Is there more I can look at while running?

 

Chris,

You've written a fine description of the details of the problem, but I'm not so sure I'm convinced that this is the cause of the problem of momentary disturbances of the fuel flow-rate. I say that because <T> should be reading the high-selected value of SRV position feedback and if 96SR-2 is in fact high as shown in the data below then <T> should be agreeing with <S>. I'm really concerned by the fact that <T> seems to be <b>momentarily</b> agreeing with <S> which is what's causing the SRV to close, the decrease in P2 pressure, the increase in gas fuel supply pressure, and the increase in GCV position during the event.

From your description, I agree that the problem is likely in <S>. I'm just concerned about why <T> is "momentarily" agreeing with <S>.... That's the odd part for me.

When measuring LVDT excitation or feedback voltages, one must use a True AC RMS voltmeter. Anything less than that will produce erroneous results. Some very expensive Fluke multimeters say they are RMS--but only up to 100 Hz or 1000 Hz, which means they are NOT True AC RMS. Only a True AC RMS voltmeter will yield correct readings with excitation frequencies of 2.8- and 3.2 KHz (which the Mark provides, each LVDT gets a different frequency, though both should be at the same voltage, approximately 7.0 VAC RMS). (Most True AC RMS voltmeters will say so on the front of the meter; they charge more for True AC RMS capability so they usually print it on the front of the meter!)

"The Mark IV Manual" says the turbine will continue to run with the loss of any single control processor and it's associated I/O--as long as the I/O is configured properly, there are no other Diagnostic Alarms indicate that any of the other processors thinks the turbine should be tripped, AND the servo-valve polarities of all the servo-valves are known to be correct. I'm always leery of shutting down a single processor of a TMR control panel unless I'm absolutely sure the servo-valve polarities are correct for every coil of every servo-valve, even if there are no Diagnostic Alarms and no Process Alarms.

I would say if you can't afford a turbine trip caused by trying to replace the card with the turbine running, then having a properly configured HSAA card at the ready is the next best alternative if you get a chance for a shutdown or if the turbine does trip because of the problem. There's still that pesky business of why <T> is only momentarily agreeing with <S> that is unresolved.

Wish I had better news.

However, great problem description, listing Diagnostic Alarms and voltages and understanding how the LVDTs are used by the three processors. Most people miss a lot of the details you provided!

Please write back to let us know how the troubleshooting progresses and what the problem resolution is.

 

Turns out that an (dare I say) unrelated problem has taken down the turbine today... I should have known, it's my son's birthday party, but here I am earning him a bigger cake.

We had a failure of the main gas supply pressure regulator and this turbine tripped on low fuel pressure. This did not seem to coincide with the spikes I described above. Very odd.

I've got the jumpers (FB and REF2) in each of the HSAA cards. I have a new card in <S>.

I've talked on the phone with someone also very familiar with our plant, and MkIV. He installed our HMI computers, and I've seen you recommend his system here in the threads. Initially after finding you here, CSA, I've thought you and he were one in the same. Ok, back to why I'm not with my son.

I'm going to stroke the valve down and up to see what LVDT position does with R S and T. I agree for T to agree with S momentarily baffles me, but I don't see another possibility. Hopefully stroking the valve and watching will uncover any dynamic problems with the feedback.

I'll post more, I have to read through your reply 1 more time before I head out there.

Thanks a ton for taking time out of your weekend to reply. What are the odds I'd be tackling this so soon...

I will provide feedback as things progress.

 

CSA,

I've now had a chance to stroke the valves and perform a 0% to 100 % calibration.

The first thing to note is that, with the new HSAA, all Position percentages agreed within a few percent. my 'as founds' were<pre>
R S T
3.95% 5.13% 4.86% FSGR FULL CLOSED : GSADJ -10.00
88.60% 89.22% 88.40% FSGR FULL OPEN GSADJ : 10.00</pre>
I noticed another thing that concerned me though, since we're still down and considering starting up, I'm going to hold on the detailed charts but FAGR for <S> goes to 0.00 during mid travel while <R> and <T> are about .3 to .45 mA.

I calibrated and now the position feedback % is pretty repeatable as I have stroked the valve a few times (for R S and T).

But from about 80% to 20% FAGR <S> goes to zero. once position goes to 100% or 0% and the forcing ref goes beyond, FAGR <S> catches back up to R and T.

Is that a servo polarity problem? Could it be related?

 

Ok here is what it looked like as I stroked the SRV:<pre>
Forcing Ref Position Current {FAGR} (mA)
GSADJ (V DC) FSGR (%) R S T
-10.00 -0.16 -6.87 -6.80 -6.64
1.83 0.08 0.11 0.05 0.06
1.97 1.23 0.27 0.23 0.34
2.2 5.15 0.25 0.23 0.34
2.51 10.07 0.25 0.23 0.38
2.79 14.95 0.25 0.20 0.36
3.09 19.98 0.25 0.20 0.36
3.38 24.94 0.27 0.18 0.36
3.7 30.30 0.29 0.16 0.36
3.97 34.85 0.32 0.14 0.36
4.26 39.79 0.38 0.11 0.32
4.59 45.34 0.41 0.09 0.34
4.88 49.84 0.41 0.09 0.32
5.16 54.74 0.45 0.07 0.32
5.47 60.21 0.45 0.05 0.32
5.76 64.95 0.47 0.05 0.32
6.05 69.93 0.45 0.05 0.32
6.41 75.73 0.47 0.05 0.34
6.67 80.22 0.47 0.02 0.32
6.96 85.13 0.50 0.00 0.32
7.25 89.87 0.47 0.00 0.34
7.56 94.85 0.47 0.02 0.34
7.85 99.91 0.47 0.00 0.34
8.05 99.99 0.95 0.52 0.84
8.51 100.00 3.48 3.03 3.41
10.00 100.00 5.53 4.92 5.28</pre>
LVDT % values agreed very well.

From 0-35% median value of FSGR agreed with <S> from then on it agreed with <T>. In any event it was only off by a few thousandths.

They're almost done with the gas yard regulator, so I'm pretty much done with the MkIV side of things for time being... I hope.

 

I guess my concern is, what if I have a problem with R or T at this point. My sum of servo currents may not equal 8mA +-4mA.

Also, in reviewing all alarms and the printout we have about 3 minutes and 4-5 pages of intermittent 125V DC grounds that came in about 40 minutes after the trip.

We've had ground problems for many years, and they've never been cured. Due to operational limitations, I haven't been able to find what that is about, but since it occurred during the course, I thought that I'd add it here.

 

In the first post, you stated:

T - Receives 96SR-1 (4A) & 96SR-2 (4B) outputs (HIGH SELECT) [through the terminals noted above]

I'm just tossing out simple ideas (sometime it's the simple thing that is the problem)...but if I read correctly, it appears that <T> may agree with <S> for a split second which could cause the commanded reduction in fuel flow/MW.

From my paste above, it looks like <T> gets input from both LVDT's...but <R> and <S> only get input from 1 LVDT. Perhaps this was discussed, but if <T> is flipping who it agrees with, there must be a reason.

The data posted shows that <R> and <S> are not agreeing on some readings. What if a glitch or momentary interruption in decent feedback from the LVDT causes <T> to "switch sides"?

Anyway, just a thought.

 

> I guess my concern is, what if I have a problem with R or T at this point. My
> sum of servo currents may not equal 8mA +-4mA.

The -0.8 mA, +/- 0.4 mA, is a guideline. If the algebraic sum of the three currents doesn't fit that guideline the turbine will still run. There are literally hundreds of turbines running just fine around the world that don't meet that guideline (usually because the servo polarity wasn't properly verified, if it was at all; or because of other unresolved Diagnostic Alarms).

Grounds are a problem. I was going to ask about them, but since you brought it up it's a moot point now.

Yes, they are difficult to find. And, yes; they can be a source of intermittent problems--and even card failures!

The data just shows that the LVDT connected to <S> still doesn't agree with the other LVDT. You might think about having another LVDT around, too, just in case. You may be fighting more than one problem.

 

Chris,

We recently experienced a load swing issue at our facility, which was directly attributed to a faulty HSAA card. We were operating on fuel oil as opposed to gas, but information can be gleaned from this experience. Here are some thoughts pertaining to this scenario: With regard to the Berg jumpers, verify Berg jumper configuration with your Control Specification as it takes precedence over the elementaries. Also ensure that the oscillator frequencies are not set the same, one should be set at 2.8 kHz and the other at 3.2/3.4 kHz. There are several data points which point towards a faulty HSAA card in the "S" processor. Among them are:FSGR,FSGR_v & FAGR. FSGR is most likely a problem with the LVDT. The HSAA card in each core contains an Operational Amplifier which generates the driving current that is responsible for positioning the servo valve and ultimately the SRVS/GCVs. Each processor generates its own current but they should be relatively close in amplitude. FAGR for "S" appears suspect (R 23.29 S 60.48 T 23.28). This may be a result of faulty position feedback from the LVDT or the Op Amp for this card may be bad. One other consideration is to verify the servo valve is intact. These valves contain three coils which receive driving currents from each of the respective core (R,S,T) HSSA CARDS. If you have Moog servos you may want to inspect the pencil filter integral to the valve.
As far as rebooting on line, the manufacturer says it can be done but that it is a roll of the dice. I hope this has offered some helpful insight.

 

We are running this turbine currently, and we are still experiencing the dips and peaks, just as described in the original post. We have replaced the HSAA card, we stroked the SRV (jumpers moved, etc.) and everything looked good (FAGR). That would seem to point away from the LVDT and perhaps toward intervalve pressure. We have the IBECS HMI, though it is an older version. I am trying to print to file the detailed alarm/logic history that it can provide, but I'm running into roadblocks there. As it stands, I've rebuilt most of the day's alarms and diags from printouts in an excel file. Sadly the printout from the the most important time (during startups) is nowhere to be found. I can fill in the blanks manually from the IBECS console, just as I have from the printouts.

I uploaded the file to http://www.auto-ic.com/SRVissue.xls

Does anyone know if there is a complete list of diagnostic alarms (with drop numbers, perhaps) available anywhere?

I'll add to the list soon.

Grounds were continuously coming in right after shutdown and just prior to starting up. I plan to do some ground isolating in the fall.

Thanks everyone for the advice so far. This thread will eventually contain the answer.

 

Please consider loss-of-flame in a single combustor as the cause of the power dips. If the MW power dip is roughly proportional to the contribution of a single combustor-can flame-out diagnosis will be a fruitful trouble-shooting path. Note that flame-outs can spontaneously reignite by back-propagation or cross-fire tube ignition ... with no control system action. And, the control system and fuel system pressures react violently to large power dips.

 

I have had a chance to try several different things.

Upon inspection I did find where, in all likelihood, the HSAA cards for S and T were swapped. This resulted in jumpers being miss-positions, causing LVDT feedback for the GCV to be treated as LVDT feedback for the high--select computer (S). Since GCV is at ~55% whereas LVDT for SRV is ~22%, this resulted in T computer trying to drive down the valve by reducing its servo signal (minimum).

Since system is engineered to prevent this type of problem from affecting operation, I've been trying to wrap my mind over what could cause R or T to momentarily agree with R. No progress there.

Thank you Paul, that is a scenario that we hadn't fully considered. I'm checking on that again. Initially we spent a lot of time looking into Exhaust temperatures to see if something like that was the case. They all tended to agree, so we headed down a different path.

We had an outage and I replaced the jumpers and that problem is gone. I tested and calibrated the inter-valve transmitters all of the way to the mk4. We replaced both the SRV and GCV servos, polarity and null checked them. Even though they are rebuilts from United Servo, the seem to check out just fine. We plan to replace them with factory servos and we're checking into ones with cannon plugs at the valve so that we don't have to re-run the wiring each time.

We had ground indications throughout these spikes, and they were most prevalent during startups. The grounds went to 0K-ohms. So we ended up finding them in a conduit to the starting clutch limit switch and liq fuel pump clutch relay.

And the spikes still continue.

here is some the data I've collected with values from <C> only :

DW, FSR, FSGR, FSG, FQG, FPRG, FPG3, FPG1

http://www.auto-ic.com/F6Mk4spikes.xls

 

I also just wanted to express my thanks and gratitude for <i>everyone</i> who's contributed to this thread and provided their knowledge and insight. I've only been around gas turbines for 5-6 years, though I have taken a pretty good class, I'm relatively new at this. I've learned a lot in this process and I'm very grateful.

We've had one consultant look at this on-site and he was stumped as well. I'm looking to ask another GE Mk4 expert (and newsletter author) that I know of to come and take a look. I'm not sure if soliciting paid on-site consultation is against policy here, so I'll check before I make any specific requests. But I am looking for help.

 

I must have replied to the wrong post, but if you check back about 5 or 6 posts, you will see a post dated 31 October where I added some data. I just didn't want it to get lost...

 

Ok, my new bitscope came in the mail last week, and I finally got a chance to hook up to the mag pickups:
77NH1, 2, 3

This scope is going to give me the ability to record frequencies and voltages over a long period of time. The MW spikes have become much less frequent after I have made the most recent changes in our most recent outage, but <i>they still occur</i>.

I'm connected with 100x probes. I did that due to the warning given in the April 4, 1986 newsletter Number 5 Page 2 section 4 that warns that while voltages (RMS) should be 10V at crank and 30V at FSNL voltages may be as high as 220V with a zero clearance. Apparently at over 120Vrms the input comparator is over-voltaged causing a doubling of the HRPB output.

Upon connecting, and after setting up the recorder with the right probe impedances I have found the following:

-all frequencies were exactly equal at 5.102 KHz.

-77NH-1 220 Vp-p / 73 Vrms
-77NH-2,3 10Vp-p / 4.9 Vrms

Check out my quickly web page with screenshots (camera) for a better look at my data.

http://www.auto-ic.com/bitscope/bitspike.html

I was able to see, during the short period of time that I was watching, a drop off in freq for 77NH1 for a few seconds. Not entirely sure that was real data... but 2 and 3 stayed at 5.102 KHz.

 

Great feedback! Thanks!

Let us know how the troubleshooting progresses.