Vibration Monitoring on a Steam Turbine and Axial Compressor 4

[gordonl]

I'm looking at reusing the existing proximity sensors on a steam turbine driven axial compressor for a new monitoring system. There are 4 sets of x-y probes, one set on each bearing, a thrust probe, and a key way probe for speed and phase. Is there any value in having accelerometers added to the cases of the machines?

The system:
- 3800rpm
- 30,000 BHP
- Journal Bearings
- Existing pickup IRD Mod 403 with IRD Mod 14538 Signal sensor
Turbine:
- 700lbs steam
- 27 stage
Axial Compressor:
- 175,000SCFM of air at 15psia
- 13 stage with variable stator vanes

I've sees suggestions that if blade problems are going to be detected accelerometers are required.

Questions restated:
- What types of blade problems can be detected?
- What freq response is required from the system?
- Would one accelerometer on the machine case be enough?
- Should I worry about turbine and compressor or only compressor?
- Any other comments or life experience.

Thank You in Advance,
Gord

 

[MachineryWatch]

Gordon,

I guess no one else is going to respond to your post, so I'll give you my opinion, just to keep your post near the top of the list!

I suspect what you have heard about detecting blade problems with accelerometers is a reflection of frequency range considerations for the sensors/system you are using. The number of turbine blades times the RPM of the machine means that a pretty high fundamental blade pass frequency (BPF)will result. Probably beyond the frequency range of the installed proximity probes. Of course this would also move any harmonics of BPF out well beyond the ability of the probes to "see" them.

That said, I have no personal experience with using accelerometers to detect blade problems in this type of application. Detecting blade problems (cracked or broken blades) on turbines has long been a real challenge that few people have been successful at. There have been some attempts with mixed success to derive from torsional vibration data an indication of cracked blades. When I say mixed success I mean that experiments in labs are successful with real machine simulators, but to my knowledge this has not translated to success in the field. Some of this may be due to simply not having the system deployed on machines that have developed blade cracks. There are only a couple of these systems installed and running, so coming up with success stories from the "real world" may take some time.

Finally, I would include accelerometers for reasons other than blade problems. While it is true that the proximity probes give valuable information about the shaft/bearing/rotor dynamics relationships that are very important for this type of machine, the accelerometers can provide a back-up function for the embedded sensors that may develop problems. They can be an important "reality check" for abnormal data indicated by the proximity probes. They are relatively inexpensive. They can provide information not seen by the proximity probes. For example I recently collected data on a steam driven boiler feed pump that had bearing housing vibration at 1X RPM that exceeded 1 inch/sec (peak) and the proximity probes were reading within their "normal" range (as specified by the machine manufacturer).

Skip Hartman

http://www.machinerywatch.com

 

[hacksaw]

you are not going to gain much benefit from accelerometers in an application like this, simply because of the masses involved.

one visit to a compressor deck when the machine is operating normally, will clearly demonstrate why a case mounted accelerometer is not really useful. they're great on fans and such, but nt a multistaged, multicase machine like thise

your thrust bearing seems to be under instrumented for a machine of this size.

 

[magicme]

hello
i doubt very much that you would identify blade "problems" from casing accelerometer signals. you might be able to identify blade passing frequencies but these would not indicate blade problems....they are the driving frequencies as the blades pass the stator vanes (or "nozzles" or "diaphragms"). except for the extreme case of having thrown a few blades and creating a great unbalance (if the machine is still running ! ) in order to identify a cracked blade problem you would need to measure the blade vibrating frequency (which would be lower due to the crack)....i can't see how you could do that from casing measurements. in my experience with gas and steam turbines we never attempted anything like that....on development engines, we'd put strain gages on individual blades but never on a production machine.

as a side story...one of the smoothest running test engines i ever monitored was found to have *the entire last stage of buckets gone after teardown * ..... go figure.

daveleo

 

[tiffy023]

Hi, I work on a relatively new alstom power plant and our 250MW steam turbines do utilise accelerometers for shaft supervision.
These are mounted on top/in the same plane as the eddy/proxy probes on the journal bearing housings.This is used to provide an absolute value, not just shaft relative.

 

[gordonl]

tiffy,

Do the accelerometers provide any useful data not available from the proximity probes?

 

[tiffy023]

Gordon,
The proxy's purely look at the shaft movement but when the steam turbine is running, the bearing shell is also moving. So, the accelerometers are compensating for this bearing shell movement which in turn corrects the proxy signal through some math that is used in the software,thus providing you with the shafts absolute vibration reading.

On a seperate note,our gas turbines use a velocity sensor on the bearing housing to provide machine protection i.e.trip

Hope this helps

 

[ronfrend]

I had a problem some time ago with a Solar Centaur centrifugal gas producer on the Gas Turbine. The unit had a DYMAC system with a permanently installed accelerometer and levels started to peak at about 7 Gs causing trips but at blade passing frequencies. At the time I was young and naive so I asked our "expert" back in holland who promptly faxed (long time ago) back and said we should ignore the vibration as it was way above the natural frequency of the accelerometer.

I objected that the levels were still rising so something was happening. While this discussion was going on the GP collapsed. Turned out that the tie rod holding the stack together had come loose and we WERE actually seeing blade pass frequencies.

Sorry for the long story but the point is that YES, accelerometers can work at high frequencies but be aware that they may well be above natural frequency. But even so trending is useful.

Good luck,

Ron Frend

Ron Frend
tel: 011 44 1253 400541
ron.frend@predicon.net

http://www.predicon.net