What collection specification settings do I need when collecting time

[mcdm]

What collection specification settings do I need when collecting time waveform, i.e. velocity mm/sec rms, acceleration g rms or displacement um rms.
(iso apply as I am in the uk)
Is it as simple as; when collecting data and I find a potential problem in a spectrum, should I apply the following: - low frequency problem use displacement, medium frequency problem use velocity and high frequency problem use acceleration?
I have taken some very basis recordings of 1024 sample 400ms acceleration and 1024 samples 400ms velocity on a desk top fan. (Poor I know but just for test purposes.) It seemed that I could understand the waveform in the velocity spectrum (far less peaks) a lot easier than in the acceleration spectrum (multiple peaks). The spectra are available on request.
I would like to add time waveform parameter setting to all the machines I regularly monitor, is this viable? Are there any guidelines available?
Any views or thoughts on this would be very welcome.
mike.cooney@dowdingandmills.com

 

[McVibe]

Mike, you can email me if you want more information - here are some thoughts:

1. Yes, displacement is for low, velocity mid and acceleration for high frequencies, however, this also depends on your sensor i.e. if you are using an accelerometer to take a low speed measurement, then you're better off viewing the data in acceleration as converting down to Vel. or Displ. filters out freq.'s below 10Hz.

2. There's no such thing as peak or RMS in a waveform, these apply when you convert to a spectrum, usually RMS is preferred.

3. For machinery monitoring its usually not necessary to collect waveforms on a route - the spectrum has more information as it can separate components that are overlapped in the waveform - and these can then be related to machine components and specific faults (these are called "forcing frequencies&quot

4. One problem with the FFT (to go from wave to spectrum) is that it can't differentiate between impacts and random noise - these items can mean different things as far as machine faults are concerned - this is one reason to look at the waveform. Again, usually just for special tests or to go back and confirm something you identified in the spectrum.

We have some good technical articles that deal with these and other basic subjects at

http://www.dliengineering.com

- feel free to download what you will and email me if you have doubts or questions: afriedman@dliengineering.com

 

[MikeyP]

McVibe Wrote
"2. There's no such thing as peak or RMS in a waveform, these apply when you convert to a spectrum, usually RMS is preferred."

This is simply NOT true!

The RMS of a signal x(t) of length T is DEFINED as the square root of the integral from 0 to T of x(t) squared dt.

"4. One problem with the FFT (to go from wave to spectrum) is that it can't differentiate between impacts and random noise"

The Fourier transform of a signal is unique. No two different time signals will have the same spectrum.

M

 

[McVibe]

I was referring to the RMS as opposed to the peak or pk to pk when looking at a waveform, not calculating the total RMS over a time period. If you draw a sine wave you can talk about its amplitude in these three ways...

4. This is true - I understood your original question to mean you wanted to do time waveform analysis of machines and was asking if this is a worthwhile thing to routinely collect - I was replying that the spectrum is usually adequate - although the waveform is useful if you would like to differentiate between impacts and noise.

 

[pjhype]

mcdm,

To complicate your problem a little more, you should also consider the effects of bandwidth on the vibration levels. Usually, if you are interested in finding resonances, you will take your data in narrow or constant 1 Hz bands. However, if you are interested in an overall level in certain bands (perhaps those levels and frequency ranges that are perceptible to humans), you might want to take (or convert) your data to 1/3 octave bands. The form of your data (displacement, velocity or acceleration) and the bandwidth you present the data have different meanings to different people depending on what you want to do with the data.

pj

 

[mcdm]

many thanks for your help I have received many emails on this subject it will take me a while to sort through them Regards mike