VA alarm levels and criticality

[analyst34]

We started a Pdm program from ground zero with baselines,etc..
We have gathered up to at least 5-8 sets of readings with CSI RBMWare software.
I am needing some advice as far as how to determine what the alarm levels should be set at and what steps to take to acheive great results.

 

[vanstoja]

A good source of guidance on statistical vibration alarm settings is:
Slemp,M.& Skeirik,R.(1999), "Using Statistics to Avoid Vibration False Alarms", P/PM TECHNOLOGY, Feb. 1999, pp. 26-30. The authors are from CSI and tried several methods of alarm settings. They found the best approach to be a combination of the "external alarm envelope method" based on a statistically derived composite spectrum and the "analysis parameter method" This approach was reported to correctly identify 100% of the machines in alarm and to appropriately exclude 93% of the machines not currently exhibiting a fault.

 

[dgallagher]

Check out the Proven Method from Technical Associates. Excellent way to optimize your spectral alarms bands, narrow band envelopes, and more. Not associated with these guys, but it is valuable info when starting a program:

http://www.technicalassociates.net/Proven_Method_4th.htm

Dave G.

http://www.reliabilitydirect.com

 

[analyst34]

Thanks for the info and I will heed your advice and look into the literature suggested by you two.
Hopefully I can start off the day with some progress with these good proven methods.

 

[vanstoja]

An excerpt of the report cited by dgallagher can be found in;
Berry,J.E.(1990), "How to Specify Machinery Vibration Spectral Alarm Bands", Sound and Vibration, Sept. 1990, pp.16-28.
This is probably from an earlier edition of the report (which is now in its 4th Edition and costs $69US) and may miss some important information added later. S&V magazine has been offered free to qualified engineers working in acoustics and vibrations for over 30 years.