Hi Bhaskar,
Thank you for your quick response.
Are you saying that you can’t transform an RMS acceleration value (g) into an RMS velocity (mm/s), even if you have the frequency of the sinusoidal vibration (which is the case here)?
Could you please explain why? All the sources we have looked at have allowed us to find the conversion equations A(mm/s/s) to V(mm/s). But it’s quite possible that we missed something 
For example, on the reference site https://www.stiweb.com/Vibration_Calculator_s/104.htm, if we enter the values of the sensor (A=1g, f=100Hz) we get the same velocity value that we get with our equation.
But I repeat Bhaskar, it is possible that we missed something, in which case I am quite open to read your explanation on this subject.
Thank you for your support
Bert
Hello Bhaskar,
I may have a clue. The sensor documentation includes a table describing the sensor bandwidth (see screenshot)
This table indicates that with a data rate of 25600 and the filter active, a bandwidth of 3700 Hz is obtained. This is more than enough for most applications.
However, with most vibration sensors, it seems to me that when you change the data rate and thereby change the bandwidth, you also change the range of vibration frequencies that can be measured. Let’s take the following values from the datasheet of another vibration sensor as an example.
We can see very well that if we use a data rate of 12800, we cannot read vibrations with a frequency lower than 200 Hz. On the other hand, if we use a data rate of 6400, we can read frequencies from 100 Hz.
Could it be possible that by using a data rate of 25600 with our sensors, the range of vibration frequencies we read is erroneous? Maybe this causes the high harmonic frequencies coming from the bearings to be too heavy in the calculation of the average vibration.
This is a hypothesis, but it could be interesting to have a frequency response table of our sensors to verify (like the one above). I remind you that the fan has a rotation speed of 6000 RPM, so a dominant vibration frequency of 100 Hz.
As you can see, even on a Sunday I’m obsessed by this issue 
Thanks for your help!
Bert
Hi Bert,
the accelero sensor in this product is this one
https://www.kionix.com/product/KX222-1054
there use to be a tool which i used to calculate ODR vs BW. i will see if i can find it. as you mentioned as we increase the ODR we will loose the lower freq.
We spend quite a bit time to integrate the velocity but due to limited on board resources we were never able to get it work reliably.
keep in mind there is also be noise we we increase the ODR ( datarate).
The noise is Noise (µg/√Hz) = 600
RMS can go high if there is an extremely peak at one of the freq ( could be machine harmonic freq).
You might consider getting time domain that will help you to see if there is any high peak and if there is whats the freq and whats the amplitude.
i worked on this few year ago. will need to study more about velocity vs vibrations.
Thanks
Hello Bhaskar,
Thanks for the info.
I’ll try your recommendation. I’ll read the FFT curve to see if there’s a high amplitude harmonic in the high frequency range. This could be consistent with the fact that the fan on which the sensor is installed is emitting a sound that sounds like a worn bearing.
It would be really nice if you could find the tool which you used to calculate ODR vs BW. This would allow me to choose very precisely which ODR to use to capture the full range of vibration frequencies of the equipment.
About the transformation of acceleration into velocity. The formula in my last message is completely independent of the vibration frequency. In fact, this formula uses the ROTATION frequency of the equipment. So, with a fixed speed equipment, we can easily calculate the velocity by applying the ROTATION frequency to the acceleration. With variable speed equipment, things are a little more complicated. Every time you apply the formula, you have to know the rotation speed (rotation frequency) of the equipment.
I share an idea with you. It might be a bit of a crazy idea, but you never know Let’s imagine a new vibration sensor with a 4-20mA or 0-10V input. We could connect an RPM meter to this input. The electronics of the vibration sensor would have an integer register that would store the rotation value of the equipment provided by the RPM meter. In this way, the firmware of the vibration sensor would have access to the RPM in order to transform the acceleration into velocity. Not being a specialist in sensor electronics, I don’t know if in fact this is feasible, but it seems to me that it could be interesting.
As soon as I have the results of the tests, I will rewrite on this message to inform you. On your side, thanks in advance for your reference to the tool to calculate ODR vs BW.
Have a nice day!
Bert
Hello Bhaskar,
My client and I have just completed our tests.
It appears that ODR has a very significant effect on RMS vibration amplitude.
By modifying it, we have reduced the vibration converted to mm/s RMS from 50 mm/s to 10 mm/s. Not bad att all !
Therefore, I think that for serious predictive maintenance users, it would be really nice if you added a table showing ORD vs BW in the sensor manual.This would allow users to get the most accurate vibration measurement possible. Of course, I myself would love to have this table to set up my current customer’s sensors.
Thank you for your support
Bert
1 Like
