Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment  
MarkL

Avatar / Picture

Sr. Member
Registered:
Posts: 1,121
Reply with quote  #1 
Hi, Guy's I am starting to explore alarms a bit more sophisticated than the regular overall level, namely envelope and band alarms.

I had dabbled with statistical overall alarms, but it made a mess of some databases (some too sensitive a mouse farting would set them off and others way too generous..Saturn 5 at take-off wouldn't make them budge!).

I have lately also found, particularly in Vacuum pumps, screw compressors and Roots-type blowers,  high overalls in demod/gE and I can spend time on them 'chasing ghosts' in the raised noise floor...which is mainly down to the process I think.

I need to make better use of my time as I have pre-assigned time blocks for analysis and find I am taking a bit longer than I think I should due to 'chasing ghosts' 

Are any of you using Band or envelope alarms? I have tried some band and envelope alarms in the last week just to get a feel for it. But just curious of others experiences with them. I am aware some even don't bother with alarms, how does that work out for you? I'm sure you have other methods.
dnk

Member
Registered:
Posts: 63
Reply with quote  #2 
I have used band alarms and statistical alarms.They both take time to get right. I started with recommended
value I got from T/a of Charlotte. From there I customized the alarms if I got alarm and found no problem.
Statistical alarms are the same, they take longer to get them right. Also have to recalculate statistical alarms periodically. 
Some places % of change works well.
 
Noknroll

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 843
Reply with quote  #3 
Mark,
I'm in the don't bother with alarms camp. I use trend lines and look deeper if an elevated or increasing trend catches my eye.
Edwin

Avatar / Picture

Sr. Member
Registered:
Posts: 131
Reply with quote  #4 
I have alarms just above the trend of overall values. I am convinced that when you measure velocity, acceleration and demod, with the right frequency and filters set, you will always get a trigger of at least one of the three for whatever change will occur. From that point I zoom into the spectra and look for changes. Works out fine for years now. I hardly miss any failure ever.

You can use statistical alarms for a start and finetune later. But the trick in this is to set the statistic parameters right to get them at the right levels.

__________________
= Cost Effective Asset Management =
MarkL

Avatar / Picture

Sr. Member
Registered:
Posts: 1,121
Reply with quote  #5 
Nok-
Does that mean you evaluate every single point taken on a route when you analyse?
I used to do that at the start out of fear of missing stuff but that even are up more time.

Edwin- your an aptitude user if I recall? I know I made a mistake somewhere on setting up stat alarms but I just went back to basics after I realised my mistake. That's a good point regarding setting it just above the trend.

How do you guys decide when it comes to high background noise/continuously raised noise floor machines I.e the compressors and blowers I aforementioned?
OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #6 

First I start with ISO standard old style levels, 2.8    7.1 mm/s as soon as I have some minimum data I go for "current for a while trend avg." + 10% and 2 x times that unless it's way hi then i fix it and start over.
Yes we have analysis bands depending on system but minimum a couple of general bearing condition bands. I guess where it is noise and it is process defined, the levels get higher.....

 


__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Edwin

Avatar / Picture

Sr. Member
Registered:
Posts: 131
Reply with quote  #7 
Quote:
Originally Posted by MarkL
... How do you guys decide when it comes to high background noise/continuously raised noise floor machines I.e the compressors and blowers I aforementioned?


Indeed those are harder because of the always higher overall values. Small components will drown in the noise. But still I experienced that if really something is going on, it will effect the overall value of at least one of the measurements. But there is less time left when you notice it. (2 months instead of 9 or more with a centrifugal pump worn bearing)
And above that, we also do periodic oil analysis that sometimes triggers us earlier than vibration. but not everyone has that possibility.

compressor.png


__________________
= Cost Effective Asset Management =
Big Al

Sr. Member
Registered:
Posts: 102
Reply with quote  #8 
I've been using envelope alarms with Emonitor for years and believe that they are very effective.

Typically I use the last 15 months worth of data for any given measurement point and pick out the data outliers manually (change to No Stat). For fixed speed machines this gives excellent spectral alarms, which are based on how the machines actually run. It gets a bit more complicated with variable speed. I believe that there is a way to enter the machine speed each time and the alarms will move accordingly, but I've never done it. I just use a percentage width on the alarm instead of a constant width and it spreads the higher frequency components of the alarm over a wider area.

There is a little bit of time required to keep the alarms updated (done during analysis), but It's worth the effort.

As a rule I have enough confidence that if the overall alarm and spectral alarm aren't breached then I don't look at the data. Of course you will get plenty of false positives. Regardless of what alarm you use, there will always be something that just manages to break through.
dnk

Member
Registered:
Posts: 63
Reply with quote  #9 
I agree with Big Al. Have used Emonitor for 20 years. Had good luck with band alarms. Also seeing which band is in alarm gives you good idea what is the problem. Emonitor will also calculate statistical alarm for you.

MarkL

Avatar / Picture

Sr. Member
Registered:
Posts: 1,121
Reply with quote  #10 
Appreciate all the feedback guys.
Noknroll

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 843
Reply with quote  #11 
Mark,
yes I look at all points, with trend, spectrum and waveform on the screen at same time,  it only takes a second to scan the trend, look at the spectrum for pattern recognition and a glance at the waveform.
Even though alarms aren't triggered I think you can spot subtle changes and make a note to self to "watch this space" for next survey
MarkL

Avatar / Picture

Sr. Member
Registered:
Posts: 1,121
Reply with quote  #12 
Hi Nok
I’ve actually had the habit of having twf trend and fft on the screen at one time and mainly watch the twf when I scan through and then look closer is I see any lumps or splashes in the twf. If it looks smooth and uniform and it’s in alarm I mainly put it down to a lube issue(unless some other criteria points to more.
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.