Hello,

Any of you dealt with the fenomena of backflow? We where dealing with it. First instance i was was suggesting about a partial radial rub because this was seen in the reversed side of the full spectrum cascade plot. Becoming know of new knowledge which was standing in the orbit november issue of 2007 about slow roll: useless noise or rich data?? Slow roll is rich rich data!!! See the trend rpm vs time plot. Also adre daiu208P boxes in relation of bently 3300monitor have there bugs.The speed was going ramped fast and reaccelrated. The delta rpm was set to 17. increasing and decreasing. Other dataconfirmed the speed ramped to 0RPM and reaccelrated.

Q: was can be the root cause of backflow in an gas compressor train? What tremendous torque have been there on the couplings and rotors.
Next thing is it happened twice within 48hours.
I advised to build in the bently 3300/52 reversed monitor and do recalculations on these compressor train.

Anyone have experiences with this backflow fenomena??

I know this is for real.....

This message has been edited. Last edited by: monitor,

quote:

Also adre daiu208P boxes in relation of bently 3300monitor have there bugs.

Not sure why you make this statement. Unless something special is done, typically in the way of added hardware, the monitor can't detect direction of rotation or a change in the direction of rotation. However, as you've discovered the ADRE shows a reverse in the direction of precession from the full speed condition to something at slow roll. So ADRE was informative and pointed you toward the ultimate issue.

Keep in mind that ADRE has sampling limitations when doing rapid sampling. Henece you will often get a warning message to turn off real time plots when entering store enable. I don't know what you might have had in the way of real time plots, but it is possible that your RPM trend plot did not dip to "zero" RPM for this reason. Also, the setting in delta RPM of increasing, decreasing (or both) can affect the appearance of the data.

John from PA

Hello John,

Thank you for your reply.

I agree fully with your statements. When i posted it i was also in hold with the answer i get from the Benlty Nevada Experts. They have said the same as you did. There are limitations with ADRE. They also told me that 3500racks with S1 would better.

I did set the configuration with incr. and decr. 19RPM. The reason i missed data is due to overkill of samples during the 1 second even of 990RPM + 1330RPM = total 2320RPM (122samples).

In the nearby future the customor is going to shutdown the plant, do you have better suggestions to set the delta RPM? 23RPM? 27RPM?

regards Monitor

This message has been edited. Last edited by: monitor,

monitor, choosing delta RPM and time is somewhat of a balancing act. I don't know what version of Adre you are using but I'll assume you are using one of the latter versions that permit 2560 vector samples. I also notice that your machine speed is 8860.

A waveform sample is taken every 10 vector samples so that means that 256 waveform samples are a "full" database. Keep in mind vector samples are vector information and the waveform samples are digitized waveforms of the signal that are used for spectrum based information such as cascade plots, waterfall plots, direct orbits, spectrum plots, etc.

OK, as an example lets say you want to take steady state data for a period of (4) hours and capture a shutdown and pretty much maximize the data you can obtain. If you set delta time for 12 seconds, that will mean a vector sample will be acquired every 12 seconds and a waveform sample will be acquired every 120 seconds or two minutes. So for (4) hours you will use up 120 waveform samples and 1200 vector samples. That essentially leaves 1360 vector samples available for the shutdown (2560-1200=1360). It also allows 136 waveform samples.

OK, lets take the 8860 RPM and divide it by the remaining samples of 1360 which yields approximately 6.5 RPM. However, if during the rundown you still have delta time enabled, and say the coastdown takes two minutes, then an additional 10 vector samples and 1 wavefrom sample will be acquired. To play safe you should always be conservative so you would in actuality (for this example) set something higher for delta RPM, perhaps 10 RPM. That way you have some excess space in database and can do a manual sample at the end of the rundown for reference data purposes.

If the coastdown is very rapid, perhaps less than a minute for this example, ADRE just simply can't keep up and not all the anticipated samples will be obtained.
What you have to balance is the quality of the cascade and waterfall plots. If you sample every 25 RPM for instance, then a waveform will be obtained every 250 RPM, and a cascade or waterfall plot will be compromised of 35 samples. Is that sufficient for your purposes? That is a question you have to ask based on the machine and its behavior.

One last thing, again somewhat dependent on version, there is the repeating database feature which can help but keep in mind that when the PC automatically goes out to save a full database the sampling process will stop, so you will loose data for a period of approximately 15 seconds while the database is saved to the hard drive. Then sampling will automatically start over again, depending on your trigger settings.

Hope this helps, ADRE can do many things but you have to understand just what it can do.

John

Hello John,

Once again thank you for your reply.

Because this machine has also dealed with some compressor surge i had to set the increasing and degreasing speed setup in the configuration. With stable speed the delta time is sampling some data. I already uses the automatic sampling process when a database is full. This is a perfect function within ADRE 4.8

regards Monitor

quote:

With stable speed the delta time is sampling some data

Keep in mind that if increasing and decreasing is enabled then it is possible to rapidly pack a database full of data. For example, should you have a governor that hunts ± 20 RPM, and you have a delta RPM set at 25 and increasing and decreasing enabled, then sampling will occur due to the hunting. A workaround is to turn off delta RPM sampling while collecting steady state data and turn on delta time. I recommend settings of 6, 12, 18, etc. seconds. It then becomes easy to remember that 6 seconds gives you a waveform every minute, 12 every two minutes, etc.

The current version is 5.11 SP1 by the way. I've attached a document outlining the changes since version 4.8. If you think you will be using Adre for Windows for the long-term you may wish to update to the latest. 5.11 SP1 will not be upgraded in the future by the way, unless some major issue is determined.

John from PA

beyond_4.8.pdf (31 KB, 17 downloads)

Hello John,

I will keep in mind the settings. I am used to use prime numbers, cause an event is not often repeatable with a prime number. The first shudown i will be focused to the rundown and will use the delta RPM of 25 (incr. and decr.)

I will keep you informed.

In september we have a healthy discussion with the company about this issue.
Can you tell me what could cause backflow to this installation according to safety, reliability?
The slow roll orbit issue 2007 tells me about it but i want to confince the custumor how serious this is!!
Digging on the internet shows me a case wich was leading to which they called "catastrophic failure". This was in 1972.

regards

quote:

Can you tell me what could cause backflow to this installation according to safety, reliability?

monitor, I'm certainly not in any position to comment on the effects of reverse rotation relative to your specific installation. The Orbit article you site from 2007 does go into a particular situation with a compressor that reversed its direction and the effect on the ADRE data. If you need more insight with respect to how this was detected from the data let me know. The same article has some discussion of safety and reliability. But with respect to your specific machine I would think the compressor manufacturer can provide you with possible scenarios involving reverse rotation.

John from PA

John,

I did have contact with the writer of that document. Through this knowledge i strongly advised to place a 3300/52 reverse rotation monitor. Other advises were to ask the OEM's what could happened with the machine during this event. This is planned in september. I am looking forward to do this presentation because i want to set the safety and reliability to a higher/highest level.

A great discussion and interaction with the several parties did bring out new insights.. The custumor gonna check every valve during planned maintenance. The proces has to be finetuned and monitored during shutdown. Those answers must lead to new actions.

I'll keep you informed.

to be continued....