How much Vibration Data is enough for analysis

Enough Data to:

Identify which machine has a fault

Identify what the fault is and its cause

Determine severity of the fault (not just vibration severity) and machine reliability severity with fault present

Make practical recommendations for maintenance and operations decisions and corrective actions

Is this what you are looking for?


In real world vibration analysis, we make calls based on incomplete data and take risks. It's impossible to have all data/information. The more data the better.

In my personal practice, I insist to have the "primary" data and ask for the "secondary" data. Many times, a site vist and talking to operators (not classic vibration data) help a lot.

Regards- Ali M. Al-Shurafa 

I'm starting to develop a picture here, I might be wrong but maybe your in the wrong profession, you apparently are a level 2 VA,  yet you've been asking questions a level 1 should know, you blatantly asked (in more than one form ) for 'an easy way' to pass your level 3 Va.


Now your asking about the equivalent of 'corner cutting' by how little data can you get away with making an analysis. Forgive my skepticism but have you considered any other line of work? Something less involving perhaps.



Ali M. Al-Shurafa's comment is important.  As an example, proximity probe orbit/timebase and spectral data may very well support some form of misalignment.  If we then acquire reliable shaft centerline data you might be able to establish that the shaft is in and incorrect position within the bearing, thus supporting a conclusion.

Knowledge of the machine is essential in this process.


In most industries the first priority is routine compliance. The "if we had all the time and resources in the world" answer would be to analyze and study to no end however in the real world we are time and resource constrained. I refuse to use the words common sense, but there is a certain amount of judgement to be used by analysts in how much data is required to make calls based upon analysis. There are really two tiers at play in condition based monitoring programs. The first tier is collecting routine data in order to find those anomalies which indicate there is an issue (can you say narrow band alarms). We should look at operational, basic care, vibration, IR and UE routines which are used according to identified fault type (and as a reconciliation method) and consequences of such faults. Once we do find a possible issue using the methods above (and more) we as analysts and reliability program coordinators, do enough to qualify and detail the required intervention for maintenance. If we have a small gearbox of low cost which ran for 20 years which has a bearing defect do we need to invest time to find out which bearing it is inside the gearbox and hold root cause sessions? Of course not. If it is a critical piece of equipment in terms of safety, the environment, production, limited access for maintenance etc... then yes let's invest the time and find out the cause(s) and solutions as well as answer the inevitable question of how long will it last. In other words there are many answers to your question with the foremost answer being "it depends". Now if your only talking about tier one routines and vibration data there are well defined collection specifications and timing for data collection available from an ever growing number of sources. This is of course where taking and properly understanding a category III vibration course comes into play. Buy some books and enjoy the journey.

I have encountered many people with data of unknown quality and without the ability to interpret it. Too much and useless data is like diarrhea, and you know what that is like! You should be results-driven and not data-driven, and you may become a good vibration analyst.


! ! Shabir Ahmad KHAN® posted:

@DBTCMP no any question is meaning less. If you don't know the answer just ignore my question and wait & see for answer from other members.

This forum is not identical for a specific role of area. 

Anyone having problem can ask here. 

I think you understand & next time you will answer specific to question. Any avoid from Crassness.



No Crassness,

just observation, Offense is taken never given.

It took me, and im sure most on here, a lot of had work to get where We are, and while I'm happy to assist anyone I can when I can, I've seen far too many on this forum 'post and run' looking for help and no gratitude or follow up on whether advice worked etc, so most posts like that I Ignore but when you posted a multiple of such posts in quick succession it's a bit hard to ignore and not voice what I am sur either s are thinking and may be a bit more reserved in commenting on. But as I said it's just an observation but hey if it walks like a duck and talks like a duck it's seldom anything else.

To your original question I would echo what CKYJohn commented, and add it might help you to do a criticality evaluation and then establish from your own 'learned' experience how much data you need to make an evaluation. 



This is one of the basic things to do in setting up a route on a machine and it requires some rudimentary knowledge of the machine such as speed, type of bearings and gears if applicable. The good thing about todays software, data collectors, and computers is you can collect much more data in various forms and not overload anything. More resolution, more averages and special time wave forms are all done as routine now where back in the good(?) old days you had to watch your setup or you would lock up the data collector mid-route. I'm a firm believer if your not taking a high resolution spectrum, a high frequency spectrum and saving the TWF you  aren't getting enough data to have a good indication what you have. As mentioned there is a wealth of knowledge out there if you have the desire to look. 

Add Reply