RMS in vibration analysis

[italo2008]

Hi (and sorry for my english..)

I have to analyze 2 graphs returned by a vibrometer; there's time on the x-axis, and, on the y axis, peak and RMS respectively (see graph in attachment). But RMS isn't Root Mean Square? Shouldn't it be a number (and not a function of time)?

 

[GregLocock]

I suggest you divide the peak spectrum by the RMS spectrum. if the answer is not 1.414.. for every frequency line I'll eat a pizza.

However there is one complication. Clever engineers sometimes use peak hold instead of ensemble averaging, in which case there is no fixed relationship between the two, tho peak >= ensemble.





Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[chicopee]

RMS should stand for root mean square and should have units of m/sec^2 according to your graph. The x-axis shown as time could be correlated to let's say rotational value in radian or some other unit if you are recording some type of rotating equipment.

 

[AlbertFM]

Yes, you're right. It should be a number but if you're measuring continuously you will get a value each sample. If the value is very different from time to time may be is because your system is not stationary.

(I don't know if I've understood your question).

 

[FeX32]

I agree with Albert about the RMS.
Interesting data you have there. I wonder what it looks like in the frequency domain.

Greg, I am not sure what you mean by peak hold? If it means that we simply hold the peak value of the time response then I am not sure how this can help you. Maybe you can enlighten me.



Fe

 

[GregLocock]

Peak hold retains the highest level in each frequency bin, over the entire duration of the test. It has many practical uses and few theoretical ones.

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[BobM3]

Is every point being plotted? Or maybe only one point in 10 is being plotted and that point is the peak (and the RMS) of each measured 10 points. Interesting time units - what is this data from?

 

[FeX32]

Thanks Greg. Don't think I could use that one in my work.


Fe

 

[SomptingGuy]

I seem to recall from the days when I actually held equipment and got my hands dirty that there were RMS filters on analogue devices. For a steady-state input they'd give the same value as if you'd done it numerically with a load of samples. "Fast" and "Slow" seem vaguely familiar, but it was a long time ago.


- Steve

 

[GregLocock]

That sounds like the damping setting on a 2209 sound level meter, which also had a peak hold function thinking about it.

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Strong]

RMS is root-mean-square. Your plot looks like a high speed strip chart with amplitude in acceleration. I guess it was recorded on a machine tool operation or some other process that had varying vibration levels over the several minute time period. I use DasyLab to make a plot like this. Each data point would come from the Peak (maximum value) and RMS computed value for each time record. A Time Record is a set of samples(such as 1024 samples) that depends on the Sample Rate and Buffer Size. The relationship between Peak and RMS values would only be 1.4 if the vibration waveform were a Sine Wave. I suspect that it is not.

Walt

 

[GregLocock]

Good catch Walt, I misinterpreted it as frequency spectrum, you are right, it is a plot of amplitude vs time. In that case the peak figure is likely to be the peak amplitude seen in the sampling interval.

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.