Liquid carryover to centrifugal compressor

[tgmcg]

We recently conducted a performance test on a 1960's-vintage, 7,000 HP, high-pressure centrifugal H2 recycle compressor compressing saturated refinery H2 with MW=10 (approx). Compressor operating speed is between 7,500-9,500 rpm. Per factory PTC-10 test curves, the peak efficiency is only 73%. However, the calculated isentropic efficiency per the field test data is 90-95% which, of course, is impossible. The temperature differential across the machine appears to be reading about 13 F too low. After scratching our heads for a week or two, check calibrating all instruments, drawing new gas samples and whatnot, I believe we may have identifed a problem with liquid carryover from the suction KO vessel, which is then being evaporated by the heat of compression, thereby cooling the discharge by 13 F. The amount of liquid (assuming it's all C6) required to do this would have to be about 12% w/w (by weight), which at first glance seems high. But on a volumetric basis this equates to only 0.7%....which, assuming a finely divided spray, seems rather plausible. The effect of this hypothetical liquid volume on single-phase gas MW also puts the operating point very close to the factory head-speed curve.

Does anyone here have any quantitative insights/experience on the amount of liquid carryover a centrifugal compressor can tolerate for a prolonged period? This particular machine has been operating like this for about, well, forever. (Unfortunately we don't have any photos of the impellers during the most recent overhaul.)

We've recently conducted performance tests on about 8 other centrifugal machines using field instrumentation, obtaining better than 1-2% correlation with the factory head-speed curve. So, IMHO, test methodology and/or instrumentation are not contributing factors to the 13 F temperature error.

Any thoughts/comments?

Best regards,

Tom McGuinness, PE
Turbosystems Engineering

http://www.turbosynthesis.com

 

[zdas04]

Tom,
Your analysis seems quite plausible, and while you don't mention the volume being compressed or the suction/discharge pressures if I assume values then I can make the heat of compresson and latent heat numbers you do mention work out.

Looking at a drawing of the impeller and volute of a centrifugal compressor, the differences between it and a centrifugal pump are pretty subtle. Really the biggest difference are the materials and external piping changes required to handle the heat of compression. I can see some pretty significant mechanical loading issues if you hit one with a large slug of liquid, but a continuous 12% of mass in aerosol form doesn't create a damage scenario that I can see.

My experience with centrifugals is pretty spotty so maybe I'm just missing something here, but I don't see a huge compression concern (what the C6 is doing to the downstream processes is a whole 'nother issue).

David

 

[pmover]

ok, a matter to think about . . .

test your theory or conduct an analysis of the process streams before suction KO drum, after suction KO drum, and the liquid being removed by the KO drum. perhaps conducting an analysis upstream and dowmstream of compressor will support your theory.

also, think about this. a 7,000-rpm impeller suddenly being exposed to liquid is catastrophic. not much more to be said about that. the mass of the liquid is significantly greater than the mass of the gas, so when that liquid mass impinges/impacts on the impeller, an imbalance occurs and boom! saturated or semi-saturated vapor/aerosol is another matter and such a process stream should be thoroughly "scrubbed of liquids" before compression.

just curious, are there any "anomalies" or noticeable differences in vibration data?

good luck!
-pmover

 

[dcasto]

I can support your theory. We injected liquid propane into our compressor to cool off the propane vapors coming from the recycle system. Been doing that for 40 years, 12,000 HP unit! max rpm 6000.

 

[tgmcg]

zdas04, pmover, dcasto,

Thank you all for your helpful comments.

Compressor vibration signature is pretty clean. No indication of impact forces.

Yes, the numbers appear to hang together for the liquid carryover hypothesis. We need to find a means to confirm this, as we have a s/d coming up in early 2009, which would be the time to take corrective action.

Best regards,

Tom McGuinness, PE
Turbosystems Engineering

http://www.turbosynthesis.com

 

[pmover]

for clarification purposes, my "guess" is that the system dcasto relates to is a propane refrigeration system; whereas, the system tgmcg relates to is not (i.e. H2). if refrigeration, then the liquid propane injection is by design and can be found in other propane refrigeration systems throughout the world. in other words, the liquid propane will vaporize prior to entering the compressor - minimal possibility of liquid propane entering the compressor.

also, the "test your theory" suggestion related to the process stream analysis, not injecting liquid.

some other matters to think about . . . some you may have already addressed . . .
how about investigating the suction KO drum for proper operation. has the process stream changed, thus an increase of heavier hydrocarbons (i.e. C6) that are not removed in KO drum. any increase in flow capacity (i.e. velocity through KO drum)? do previous compressor test results indicate the same outcome?

good luck!
-pmover

 

[tgmcg]

pmover,

I understood the context of the propane compressor application.

We plan to investigate/confirm the liquid carryover hypothesis by focusing on the suction KO vessel. Data suggests there's been an increase in recycle flow over the years. Partial blockage of the demister might also reduce it's removal efficiency. There are some recent process anomalies that do not obviously relate to what we're observing at the compressor, but it's still early in the investigation. Once the process engineeers buy into the proposed carryover scenario, we'll be able to evaluate possible process-side corrective actions.

This was possibly the first performance test ever done on the compressor, so there is not a lot of documented history to go on.

Best regards,

Tom McGuinness, PE
Turbosystems Engineering

http://www.turbosynthesis.com