Skip to main content

Dear all, 

In case studies and technical notes, I have seen that the decision regarding machines' health are made based on overall (rms) vibration level. I wanted to know the common practice of vibration monitoring for industrial machines - do we measure overall vibration (10 - 1000 Hz velocity), compare with threshold limits, then we go for detailed spectrum analysis in case of alarm/danger ? 

Should not we directly monitor spectral components, e.g., 1x (90% of overall vibration), 2x, etc., because there can be situations when overall level is normal but spectrum indicates dominating frequency components. I think overall vibration level would be effective only for few cases!

Regards,

Akashri

Regards,

Akashri

Original Post

Good question.

Different people have different practices, objectives of measurements and also capabilities.

There are cases where calls are made purely based on a single value (overall rms for example). At formal acceptance tests, this can be the case. Equipment tested at a factory can be rejected if the amplitude is above the rms limit. Simply if this has been agreed on, this should be the way to do it.

 

However, many analysts are in-house and have a different objective which is to detect problems early, diagnose them and hopefully resolve the problem with the least complications. These analysts will try to employ as many tools as possible to succeed. The limits are in general very good tools to screen cases and save time. Can I do this by having limits to the key vibration frequencies like 1X, 2X etc? If I can, I would do that. However, not every vibration instrument tool has that feature. There very basic tools that are based on an overall value readout.

I would say, most of the vibration analysts whose role is fundamentally focused on vibration analysis would use screening based on overall and spectral limits. Some online protection systems do that but most of the time, based on my limited experience, these functions are not enabled as alarms.

I hope this helps a little.

Regards- Ali M. Al-Shurafa    

Basically no but it is a minimum cost solution and does give the lowest cost effective machine insurance you can buy. You will know that there is something wrong if somebody is looking and any alarms are set etc. etc. If your level goes up it is a very crude monitoring, better than nothing but still. If you also want to know what is wrong you need more analysis. You would increase usefulness very much by also monitoring a bearing fault parameter. In most cases it is bearing fault that shut down your machine unplanned. Having only 10-1000 Hz RMS will give you a short decision horizon or none in those cases but it will shut down before failure. Some machines, slow would need to go down to 1.2Hz, all electrical motors would benefit from increasing to 12-1500Hz for electrical faults and if you have expensive hi speed gearboxes you would need frequency analysis and up to 2xgearmesh. So you need to analyse your case but in the boiler PO it says "vibration monitoring included" so you get the cheapest the supplier can get away with unless you specify. IRL it is very unlikely that a peak in the FFT goes up that is significant that is balanced by another peak going down thus keeping the "ISO" value unchanged. I maybe seen something like that once on a gbx but it was monitored by full FFT. Only my view.

@Akashri posted:

Dear all, 

In case studies and technical notes, I have seen that the decision regarding machines' health are made based on overall (rms) vibration level. I wanted to know the common practice of vibration monitoring for industrial machines - do we measure overall vibration (10 - 1000 Hz velocity), compare with threshold limits, then we go for detailed spectrum analysis in case of alarm/danger ? 

Should not we directly monitor spectral components, e.g., 1x (90% of overall vibration), 2x, etc., because there can be situations when overall level is normal but spectrum indicates dominating frequency components. I think overall vibration level would be effective only for few cases!

Regards,

Akashri

Akashri,

I personally think it is a choice for each person to decide.

I decided, very early in my career, NOT TO DO IT......... Depending on an overall's calculation value.

Since 1988 I have looked and analyzed every point's data which I or someone else had taken and for which I was responsible for making a decision on.

Once a person gets to the point of, what I call "pattern recognition", the looking at every spectrum/waveform is really fast and simple, as to whether or not a more in-depth analysis is needed.

Just my personal opinion.

Judging from your last paragraph of your post, you seem to be a little leery of using it, "overall value".

But, "to each his own" belief.

Thanks,

Ralph

What are you trying to find?  What causes your machine trouble and has value?  These are the items you should monitor for.  You need the capability to detect these items, and you will need the signals that can detect these.  If overall measurements suffice, good, if not you have to decide if you will miss something important and how important it is if you miss it.

Add Reply

Post
×
×
×
×
Link copied to your clipboard.
×