Complaints of GCV Oscillating but Calibration and LVDT OK

T

Thread Starter

topcat

The customer I am on site for is complaining that the GCV is oscillating. The calibration has been checked ok and the LVDTs seem fine. There are no issues for the speed pickups, p2 pressure transducers etc.

At FSNL the GCV is around 15% but oscillates up to about 3%. As they have also complained of the Liquid fuel oscillating, I have been looking at the speed control loop for FSRN.

At 22MW the oscillation/control is reduced at around 1% and looks ok with FSR around 38%.

SRV is rock solid also just controlling 0.5-1% movement.

This is a FR6 unit with a MK5 control. The speed control loop uses an FSRNV1 block. The constants in the big block are as follows:<pre>
FSKRN1 16.7% FSKRN2 12.7%/% FSRKN3 3.13%/%

FSKRN4 16.15%/% FSKRN5 0.06 and FSKRN6 0.5%</pre>
My question is regard to these constant. Can anyone compare and comment on the settings and can the correction factor constants FSKRN2, 4 be used to dampen this oscillations?
 
Hi there,

I believe it has nothing to do with the control constants, but rather the current gain of the regulator constant in I/O Config could be too high. Can you tell me what are the values of the current gain and current bias that you have downloaded into Mark V after calibration?

Rgds,
Gomez
 
topcat,

Hmmmm..... So, this is for a paying job--you're getting paid and we're not?

FSKRN1 and FSKRN2 are the Droop Speed Control setpoints; changing them will have an effect on loading/unloading rates and droop. Small changes might not hurt, but large changes will definitely be noticeable.

When did this problem start? Most turbines, when at FSNL have a little bit of instability. The fuel control valves (especially for gas fuel) are right on the edge of their controlling range. And, the more "open" they go the more stable they become.

Grid stability can also have a LOT to do with this. If the grid is small(er), then the grid frequency can be oscillating which will cause the speed to oscillate which will usually cause problems for both the SRV AND the GCV. TNHA is the value that best indicates actual speed instability. It should be relatively smooth, with short peaks. If it's not smooth and/or has large peaks, then the actual speed/grid frequency is not stable.

Have you tried gagging FSR to see what effect this has?

I have seen Droop settings be used to increase stability, but only once. And only on a small, islanded grid. I checked with GE factory personnel, and made the change under their direction--with the admonishment that this was a one-off occurrence and that any changes to Droop settings should always be done only with factory review and concurrence. You may not have GE factory access, but, in my experience (30+ years) the Droop settings were only changed once. I changed them <b>BAC</b> to factory settings several times--when personnel had changed them and caused unintended knock-on effects and consequences. And, of course, that was always the Speedtronic's fault--that awful, mystical, unintelligible Speedtronic.

NOT!

If this is one machine at a site, I'd be looking at other things like L.O. cleanliness/quality; age of servo; etc. I'd also be looking at the regulator gains for the fuel valves--the SRV, the GCV, and the LFBV. Many times people change these values--or they don't get set properly during commissioning. These are another group of variables that should rarely, if ever, be changed. And, usually, again only with GE factory concurrence.

You will find the regulator gains in the I/O Configurator. These values should be relatively the same for just about every Frame 6B. The Droop settings will vary for just about every Frame 6B--because fuels are not the same and sites are different. Droop settings should be similar--but not the same (except coincidentally).
 
CSA

Hello again. Its the same job I spoke with you about once before.
And yes it is a small Island. When I described my issue only another two units were running for the entire grid along with my running unit.

The unit was installed last year but it is not new. It is an older unit that have been given a refurbishment.The person who commissioned this unit is knowledgeable and would have a done good calibration.

The current gain he set was 1.8 and the current bais 2.38. I know bias is normally around 2.67 but I,m sure that would have been based on a calc across the coils.

I guess I will go and check the calibration again tomorrow.

It sounds as though I should steer clear of droop speed settings. I looked at another FR6 settings for another job location and FRKRN1 was 14.7 and FRKRN2 was 13. That's why I was looking at this area.

I do like your statements regarding less instability as the GCV goes more open as that is what I see here. But not sure how to get this argument over to the customer who have already shown me the valve LVDT movement on there older MK4 FR6 units which are quite stable in comparison.

Would it help me any to see what speed droop constant settings are installed for the other older site units with MK4?
 
Hi there,

I think you should verify the current bias. Normally the null bias will change gradually over the years on a servo.

The clue that led me to this idea is that you mentioned the FSKRN1 constant is 16.7% but FSNL GCV opening was only 15%, so there seems to be a 1.7% in positioning error by the regulator. Improper current bias constant would normally result in a steady-state error. This steady state error might cause the regulator in a more aggressive way in order to minimise the error but caused overshoot...and these repetitive actions causing the oscillation. Only a guess. Double checking the null bias can also tell you whether the servo is still good for use. If the calculated rated null bias current deviates too much from the specs then it is recommended that it be replaced.

Can you confirm this by manually stroking the valve?

If you think the oscillation is due to grid instability, it should be observable on the grid frequency when the unit is synchronised, and oscillation should not have happened during FSNL.

Cheers,
Gomez
 
CSA/Gomez

Stroked the GCV from 0% to 100% open. Noted the LVDT positions in DIAGC for 0 and 100% stroke. May need small adjustment to installed input in IOCFG.

Coil resistance measured at panel (r)1140ohms (s)1157ohms (t)1136hms

Set stroke to 50% and checked polarity. All ok.
Checked DC mA on each coil. Found (r) 0.24mA (s) 0.37mA (t) 0.34mA

Calculate that I would have to change current bias from 2.38 to 3.16.

Do you concur?

Thanks
Topcat
 
Topcat/Gomez,

During LVDT calibration--which should include actual physical measurement of device position--if it's found that the calibrated LVDT feedback is equal to the actual measured physical position but the feedback/actual physical measured position is not equal to the reference value, then and only then should an adjustment be made to the null bias current value.

Null bias current only adds a fixed amount of current to the output to be used to make the actual measured physical position equal to the reference. The implication here is that the calibrated LVDT feedback is equal to the actual measured physical position. So, it can also be said that null bias current is used to make the calibrated LVDT feedback equal to the position reference--<b>BUT >>ONLY<< WHEN THE CALIBRATED LVDT FEEDBACK IS VERIFIED TO BE EQUAL TO THE ACTUAL MEASURED PHYSICAL POSITION OF THE DEVICE.</b>

So, in other words, if the null bias current value was 2.67%, and the calibrated LVDT feedback were verified to be equal to the actual measured physical position of the device, BUT the position (actual AND calibrated LVDT feedback) were less than the position reference, then one could increase the null bias current value to add a little more negative current to the servo-valve output to make the position (measured AND calibrated LVDT feedback) equal to the reference position. That's when it's appropriate to change null bias current values. And only then.

And, in my experience, the amount to change the null bias current value by or to <b>CANNOT BE CALCULATED</b> using the formula in the Control Spec. And, if it's necessary to change the null bias current to be anything less than 1.33% or more than 4.00%, then there's something wrong with the servo or the actuator or the calibration method. In other words, adjusting the null bias current value is a trial-and-error process, and does not lend itself to any kind of calculation. And, even if a new value is calculated, it still has to be verified. And, if it needs changing, how does one decide how much to change it? Another calculation? A smidge here or there? Then, what's the use of the formula? (It's really just some misguided person's attempt at presenting a method for calculating something that doesn't lend itself to calculation. And, it causes a LOT of problems for many people who don't understand what they're trying to do and what the limits of what they're trying to do are.)

If no physical measurements of device position are being taken, then there's no point in adjusting the null bias current. Full stop. Period. If calibrated LVDT position is not equal to actual measured physical position, then what is the point of making the LVDT feedback equal to the position reference? Again, if the LVDT feedback is not equal to the actual physical position then what's the point of making the LVDT feedback equal to the position reference?

I've been to too many sites to count where people used that formula in the Control Spec to adjust null bias current trying to make calibrated LVDT feedback equal to position reference when the calibrated LVDT feedback was not anywhere near actual physical position. And, because they couldn't get the LVDT feedback to equal the position reference as they made changes to the null bias value, they just kept changing the null bias until they hit the limit of adjustment (10 mA, for most Speedtronic panels).

If the calibration procedure doesn't compare actual physical position to calibrated LVDT feedback, then what's the use of making the LVDT feedback equal to the reference when the actual physical position won't be equal to the reference? It's useless. Full stop. Period. End of discussion.

<b>Changing null bias has absolutely >>ZERO<< effect on device stability.</b> None. Nada. Niente. Servo current gain or regulator gain affects device stability. Oil quality and servo condition affects stability. LVDT condition affects stability. Valve plug/seat condition can affect stability. Grid conditions can affect stability. Gas fuel supply conditions can affect GCV stability (though from the sounds of it the SRV is pretty stable so that shouldn't be an issue).

I failed to ask previously, but how much is the load changing when the fuel control valves are oscillating? And since this is a "soft" grid (from the sounds of it), how is that affecting grid frequency?

What is TNHA doing? Since this is a Mark V, you're going to need to use VIEW2 to capture high-speed data on TNHA. And then plot it somehow with one of the many third-party software applications, or MS-Excel, to see it graphically represented on a "trend." You should look at TNHA, DWATT, TNH, DVAR (if there is a MVAr transducer), FSG, FPG2, and FPRG.

Have you used AutoCalibrate 'Verify Current' or 'Verify Position' to determine if the LVDT feedback is linear over the range of travel--or if the servo current required to move the device is constant over the range of travel? (These two features have to be run AFTER an AutoCalibrate, when there are new RAM values for LVDT calibration.) If the LVDT(s) is(are) worn in the area of FSNL/low load, then the feedback could be unstable at that point. Or, if the actuator cylinder is worn at the FSNL/low load point then it could require unusually high amounts of servo current to maintain desired position, which could result in instability. Even one LVDT being worn in a certain range can cause feedback problems.

Now, FSKRN1 is calculated for "straight" Droop Speed Control (which is what FSKRN1 & FSKRN2 are used for) assuming that the machine is at ISO conditions (new and clean; nameplate ambient temperature and -pressure; rated exhaust duct back-pressure; clean inlet filters; etc.), <b>AND</b> that the fuel characteristics match those in the Control Spec. It also assumes that the GCV LVDTs have been calibrated correctly (meaning that calibrated LVDT feedback is equal to actual measured physical position--AND that 100% stroke is as per the Control Spec. value). It's pretty rare these days that the fuel actually being burned matches the original expected fuel characteristics (sometimes it's better; sometimes not). And, it's pretty rare that LVDTs are calibrated properly.

But, mostly it's unlikely that the machine is started at nameplate ambient temperature and -pressure, which will have an impact on actual FSNL FSR versus FSKRN1--which is the <i><b>approximate</i></b> value of FSNL FSR. So, just because there's a mismatch between FSNL FSR (or extremely low load FSR) and FSKRN1 is not immediate cause for concern. Actual operating conditions have to be taken into account when assessing whether or not the value of FSKRN1 is right or wrong. We don't have enough information about the load or the value of FSNL FSR to be able to draw any conclusion--and it's not likely that the value FSKRN1 has much of an effect on GCV stability. FSKRN1 is an offset--a "bias." FSKRN2 is the gain, and would have much more of an effect on stability. FSKRN1 has to be some value that's near to the typical value of FSNL FSR--otherwise there will be problems with AutoSynchronizing taking an excessive amount of time.

The fact that the Customer also complains of LFBV instability when running at light load on liquid fuel is also pretty telling in this case--and my money's on the grid as the most likely cause of the problem of fuel valve instability at light load.

Now I'm not saying that changing the Droop setting(s) might not have a positive effect on load stability, but I am saying that changing Droop settings without understanding all of the knock-on effects may have a large effect on other aspects of operation. We don't know how grid frequency is controlled on this island--whether it's by running one machine in Isochronous control, or by some kind of third-part "power management system" which send signals to various prime mover governors all of which are operating in Droop speed control, or whether or not there is some kind of Isochronous load-sharing scheme with two or more units operating in Isoch load-sharing mode. There's just too much we don't know here. Too much.

But, one thing I do know: Changing null bias current values without knowing what the actual measured physical position of the GCV is with respect to the calibrated LVDT feedback--and using that formula in the Control Spec to decide on how to change null bias current values--is <b>>>NOT<<</b> going to affect GCV stability. Not unless there's a LOT we don't know about this unit and the site and the island and how things are being operated. And, there is a lot we don't know. But, I still maintain that changing null bias currents isn't going to fix this problem.

If you're going to change Droop settings (FSKRN1 and/or FSKRN2), I would recommend you do so with the consent and agreement of the grid operators. If you increase the Droop setting (say from 4% to 5%) that means the unit will load faster, and that may affect their operating scheme. It means the unit will load faster when directed to when the grid frequency is normal, and it will change load faster when grid frequency is not normal.

And, of you're going to change Droop settings, you should know what FSNL FSR is at the typical site ambient temperature (which isn't always near nameplate ambient temperature) and Base Load FSR (and Peak Load FSR if the machine has Peak Load Capability).

LVDT calibration has zero effect on device stability. LVDT feedback, if it's not linear, has an effect on device stability. And, null bias current has zero effect on device stability. It's time the myths and wives' tales and legends and falsehoods about "calibration" and servos be dispelled once and for all.

Lastly, I am working in the field at the present time and don't have access to some old jobs where I could look up and verify the GCV servo regulator gain value. But, you should be able to find it in the Control Spec., somewhere in Sect. 05.02.nn--same for the LFBV servo regulator gain value, except it should be Sect. 05.01.nn.

Write back to let us know how you fare!
 
topcat,

For a TMR Mark V turbine control panel being used on a GE-design Frame 6B heavy duty gas turbine with conventional combustors and a Young & Franklin combined gas valve assembly (SRV & GCV in a single assembly) using a Moog servo-valve rated for 1.0 gpm, the GCV servo regulator gain value should be 5.9. That's a pretty significant difference from 1.8 which you reported earlier. Unfortunately, I don't have the GE P/N for the 1.0 GPM servo-valve which would be used on the machine described above.

It could very well be that the servo is not a 1.0 gpm servo, or the person who commissioned the unit made some changes for "stability" reasons. Or, there's something "unusual" about the unit (it doesn't have conventional combustors; it has a non-standard gas valve assembly or non-standard servo/actuator assembly; something like that).

I've never tried changing servo regulator gain values on a TMR panel while the unit is running, so I don't know what would happen. And I'd certainly be leery of going directly from 1.8 to 5.9, which would mean even more downloads and re-boots while the unit was running which always increases the risk of tripping the unit.

I recommend you get the data requested and make some decisions. I'm also looking at the numbers you sent previously. For a machine with 4% droop and the Control Constants you provided, the FSR at Base Load for a machine in new and clean condition and at nameplate ambient temperature and -pressure would be ((4 * 12.7) + 16.7) = 67.5%. That's for a machine WITHOUT Peak or Peak Reserve. (If the machine had Peak Load capability, the FSR at Peak Load would be 67.5%; if the machine had Peak Reserve capability, the FSR at Peak Load would be 67.5%).

Let's say that the machine was rated for 40 MW (a typical Frame 6B). At 22 MW, the load would be approximately (22/40) 55% of rated. The FSR for that would be approximately ((0.55 * (4 * 12.7)) + 16.7) = 44.64%. And you reported an FSR of approximately 38%. That's close, and considering you said the FSNL FSR was lower than FSKRN1 that would account for some of the difference. And, presuming the ambient temperature is higher than nameplate AND that it's probably been a while since the last off-line axial compressor water wash and a while since the last maintenance outage (CI, HGPI, Major), and that the fuel may not be exactly equal to specification (since this unit was relocated), well that could account for some of not most of the difference.

I don't know what to say without more data. And more feedback.
 
Topcat/CSA,

Thanks CSA for your sharing. On the second thought after reading your comment I think you are right, current bias should not be the root cause to oscillation, actually i don't even think current gain could cause that too, in view of the fact that the system is a pure first order dynamic system (that's why type 43 regulator is a proportional feedback control loop only).

Topcat, did you notice the same oscillation happened when you manually move the valve in the mid travel position and hold it there? I am suspecting a bad hydraulic actuator (seal/diaphragm leakage maybe).

Gomez
 
topcat,

I've read and re-read the original post, and I only see a description of GCV oscillations at FSNL. And, there's no indication of how much speed (TNH; TNH_RPM) is changing.

Is it difficult to synchronize the unit with these GCV oscillations?

What is the SRV position at FSNL? At 22 MW? I know you say it's very stable, but, what are the positions of the SRV? Could there be some kind of upstream flow restriction which would be reflected in the SRV being open much more than usual?

You say the Customer complains of similar oscillations on liquid fuel. There isn't usually an LVDT on the liquid fuel bypass valve on a GE-design Frame 6B heavy duty gas turbine. Is this an older machine with a variable high-pressure liquid fuel pump? Some of the had LVDTs, or, rather, RVDTs, for position indication.

What is the servo regulator gain for the liquid fuel control valve/pump?

Without some actionable data (speed oscillations; load oscillations; etc.) it's really difficult to say if the fuel control valve oscillations are really a problem or not.
 
CSA

Can I get back to you. I am in Iran and internet is not so reliable so I have only just seen the last few responses

I need to digest your reply before responding.

thanks
Topcat
 
CSA

Thats a lot of reading! Phew you can talk mate. But it is appreciated.

I am going to steer clear of any changes to the speed droop constants. To be honest I really only wanted to know if mines were in the ball park of normal.

The LVDT actual position matches fine with calibrated feedback value. Any change to current bias was made after I had confirmed this was done.

I have run VIEW2 trends but here in Iran things that might be taken for granted are in short supply. Mainly floppy discs and USB a:drives that work. Still I have looked at prevote data etc and found that values for DWATT,DVAR FPG2, FPRG are stable. With TNH only changing by 2-3rpm.

Also you may remember or not that I could not get Autocalibrate to work here and use the User Defined display instead.

The unit has been run to 20MW and the oscillation is certainly reduced to about 1%.
But the hydraulic pressure still fluctuates by 2 to 3bar

With the Lvdts checked again for position and the new null bias setting of 3.16 the GCV stability was certainly improved on my next run up.

But I am most interested in your statement that current Gain would be more likely to improve servo and valve stability and that my values of 1.8 may be low and may usually be around the 5.9 mark.

These units are not running online at the mo so I do have some leaway to adjust these values. I will of course try to get the correct spec values of gain for these units and servo type but would you consider raising the gain to 5.9 worth trying to see the reaction?

Hope I have answered some of the many questions.

regards
Topcat
 
topcat,

This is new information--that the hydraulic supply pressure is fluctuating. 2-3 bar is a pretty significant amount, in my opinion. Does the unit have a hydraulic accumulator? Is it correctly charged and valved in correctly?

The hydraulic actuator for the GCV is pretty small; the servos are usually only rated for 1 gpm (gallon per minute). Even if it was moving +/-10% I wouldn't expect the hydraulic supply pressure to fluctuate very much, if at all. Even if there were no hydraulic accumulator, or it wasn't working properly or wasn't valved in properly.

I'm very surprised that changing the null bias value had any effect on stability--extremely surprised. Yes; if you have the ability to change the servo regulator gain to 5.9 I would suggest it would be worth the effort--but only if cause of the hydraulic pressure fluctuations can be sorted and resolved.

If the speed is only changing by 2-3 RPM, I really don't understand what the problem is. If the GCV has to change by that much to hold the speed relatively constant, that's not really bad. Unless it's jumping up and down to do that. My rule of thumb for valve stability was to take a coin with a flat edge and balance it on edge on the bar the LVDTs armatures are connected to while the turbine was running. If the coin didn't fall, then even if there was some movement in the valve that's stable. I learnt that from a very experienced (mature; old) field engineer. It's still as true today in the digital control world as it was back in the analog control world.

I think there are several issues all combining to cause this concern over "oscillating" GCV position. And, one of them is likely that they can't see a trend of the Mark IV-equipped units' GCV positions, but they can see that for the Mark V. Another is fluctuating hydraulic pressure. Another might be a problem with gas fuel supply restriction (some issue with the y-strainer or filters). Another might be actuator and/or servo condition. It's often found that a problem is the result of two--or more--causes, one of them being the availability of data that was never available before (this happens a LOT when the control system is upgraded).

I forget the name of the carrot soup (ash e jow, I think it is)--but have a big bowl for me!
 
Hi Topcat,

There is one site where I have worked before having Frame 6B and the current gain is 1.92 %/% ...

Firstly can you kindly answer one crucial question of mine asked previously: whether the oscillation is observable when you manually move the valve to a position where the so called oscillation is greatest observed during unit run. By answering this we would be able to rule out many uncertainties.

Secondly, it is not advisable to increase the gain by too much as it may adverse the situation if wrongly done. Please increase it gradually, say 1% per try, to see how situation turns out.

Do you also have information in the rated flow rate of the servo, surface area of the hydraulic actuator piston, volume of its cylinder and etc.? As with the information I might be able to calculate the best gain for you

Cheers
Gomez
 
Topcat and CSA,

Extending on the subject of hunting of hydraulic pressure (with main lube oil pump), could it be caused by (a?) leaking hydraulic actuator(s?) (which I suspected to be the cause for "oscillating" GCV)?

The inconsistent hydraulic (header?) pressure is supposed to affect other devices as well such as IGV and SRV...maybe the servos are regulating rigorously.

Gomez
 
Guys

Regarding my GCV and hydraulic fluctuation. Have been sent to another site in Saudi temporarily for a startup so will get back to this issue shortly.

Thanks for the help and advice so far
regards
Topcat
 
Top