After performing an "online" compressor wash the following events transpire: Multiple diagnostics on either <r> or <s> computer (occurs on only one computer per instance) FSRS1, FSRSR, FSR2, FSROUT, FPRG, FSRACC, FSRSU, FRST, L4X1, L20CB1X, L20FG1X, L83RX, L4GA, L20WN1X, L60 WQDIR and the Master Protective Lockout.

Gas Valve Control Screen:

Funct name med units <r> <s> <t>

REFER FSR2 75.0 %fsr 75.0 0.0 75.0
REF OUT FSROUT 7.38 V DC 7.38 0.00 7.60
REF BACK FRSBAK 7.38 V DC 7.38 0.00 7.56
POS VOLTS FSG_V -7.34 V DC -7.19 -7.40 -7.34
POSITION FSG 74.53 %STR 74.53 73.86 74.55
CURRENT FAG 0.45 MA 0.54 -8.69 0.45

It appears to me that the <s> computer is trying to force the valve closed, thus the negative current value on FAG. Since this occurs following the online compressor wash could it be that the GCV is hanging up internally? Resetting the computer in question will correct the issue but hardly seems to be an viable solution.

Replaced the GCV LVDT and Servo yesterday (calibrated)and the unit ran at baseload throughout the night w/o any issues. It started approximately 3 minutes following the wash.

Thanks, Mark

 

The problem was finally traced to faulty relays on 45FTX and 45 FX1 in the MkIV. Also the main gas supply regulators were found to have accumulated grease on the strainer baskets (thus limiting flow) from gas valve greasing on the upstream valves.

 

Mark,

Thanks for the feedback!

I had sent a response and forgot about the thread since there hadn't been any activity, but I see the response never got posted.... These things do happen; sometimes things get lost in the Great Bit-bucket.

Anyway, to summarize, first, I think the list of Diagnostic Alarms was incomplete. And, second, when you see this group of alarms it generally means that the indicated processor believes the turbine should be tripped. The big indicators are the L20FG1X and L4X1; the first is the trip solenoid output and the second is the 'Master Relay' signal, for lack of a better term. But there should be a few more Diagnostic Alarms at the "bottom" of the list (since they would precede all the others, and the last alarm out is at the top of the list), and they would provide some more information.

If you look at the top of the relay output module on the back wall of the Mark V, you will find three relays: 4R, 4S and 4T. Those three relays are supposed to be energized when the turbine is running, and de-energized (dropped out) to trip. When <S> is trying to trip the turbine, the 4S relay will be de-energized. If you look at the 4G sheet (I think it's 4G!) you will see the two-out-of-three hardware voting using the 4R, 4S, and 4T relays.

You never indicated what Process Alarm was annunciated at the time of the trip. It is *supposed* to take two out of three processors to trip the turbine, and it does. But, in the Mark IV, there can be a little trouble with that. Read on.

For a Process Alarm to be annunciated, two of the three control processors must believe the <b>same</b> alarm condition exists to be annunciated. In most cases, all three processors will detect the same problem, annunciate the alarm, and if need be, trip the turbine by dropping out the 4R, 4S, and 4T relays all at relatively the same time.

But, in the Mark IV, if <R> processor thinks the turbine should be tripped because the IGVs are out of position it will drop out its 4R relay, and a lot of Diagnostic Alarms will be annunciated, including one that indicates the IGVs are out of position!

Now, let's say that nobody did anything about those Diagnostic Alarms and the turbine was left running with 4R dropped out, and <S> thought that the gas fuel trip oil pressure was low enough to trip the turbine. Well, <S> would annunciate a bunch of Diagnostic Alarms, drop out its 4S relay, and WHAM! the turbine would be tripped because two of the three 4 relays had been de-energized. HOWEVER, <b>NO</b> Process Alarm would be annunciated, except 'Loss of Flame Trip' because two of the three processors did not agree that the turbine should be tripped <b>for the same reason!</b>

So, it's really critical to understand why a single processor thinks the turbine should be tripped and resolve that issue, and then usually a Master Reset will reset the processor so it doesn't have to be rebooted while the turbine is running.

This problem was resolved in later versions of Speedtronic using SIFT, Software Implemented Fault Tolerance. The processors in later Speedtronic panels are all very tightly synchronized and they each individually vote on the inputs of all three processors and then use the voted value of the inputs in their execution of the CSP. In the Mark IV, the three control processors act independently and don't really know what the other control processors are doing.

Another way to look at why a particular processor might think the turbine should be tripped is to use the Auxiliary Display (presuming all the segments are working properly!). In the manuals, and on the magnetic sign on the front of the Mark IV, there are instructions about how to use the Aux. Display to look at the Process Alarms in each processor. So, if <R> has dropped out its 4R relay and the Diagnostic Alarms don't make much sense, use the Aux. Display to look at the Process Alarms in <R> and write down the drop numbers as you scroll through the list of Process Alarms using the Aux. Display.

You will find that the drop corresponding to the Process Alarm which that processor believes should result in a turbine trip will be in the list you obtained from the Aux. Display, and you can then go about resolving the alarm.

I'm very confused about how the 45FTX and 45FX1 relays can cause trips of individual processors; it just doesn't seem likely, but stranger things have happened. Usually, a contact from 45FTX is used in the hardware trip string on the 4G sheet (the same one referenced above) and if it is faulty it will act to trip the turbine by interrupting the hardware trip string.

Now, if that happened when one of the processors was thinking the turbine should be tripped for another reason, then that might explain what you were seemingly describing.