Dear Experts,
I had a resonance suspect at running speed in a feed pump which cannot be turned off to do a bump test (Equipment off), so I had to go with the bump test (Equipment running), I am using CSI 2130 analyzer, I have a confusion about how the negative averaging works, I read about it, although I am not sure about the results.
I have done the test first with the following settings

Expert analysis: (Bump test – equipment running)
Fmax = 400 HZ
LOR = 1600
Averages = 12 (one hit per average)
Pre trigger = 10 %
Trigger level = 1 G’s
Window = Hanning

The resulting spectrum was very similar to the normal spectrum before negative averaging but lower in amplitude.
I thought maybe I didn't leave it the enough time to settle in the second step (live spectrum) before I press (Stop) button,

So I repeated the test again twice; one with the default setting in the analyzer, I left it for long time before I pressed stop, the result was a flat spectrum (Just some noise - see attached)
Second, I used a high resolution (Fmax = 200 HZ, LOR = 3200), I tried this high resolution to seperate between natural frequency (If exist close to 1X), the result was a similar spectrum as before but 1X almost disappear, leaving little peaks @ 2X, 3X.

Could anyone please tell me what wrong I am doing and the validity of the tests done.
- How long should I wait in the second step (Live spectrum) before I press stop?
- What if the natural frequency was exactly the same as 1X, and I waited for a long time taking too many averages, wouldn't it cancel each other leaving a flat line? the "valley" shape in the peak as described in CSI manual, should it always appear if I have resonance @ running speed?
- In the last trial, If I don't have a natural @ running speed so that 1X completely gone, why the spectrum still showing 2X & 3X after negative averaging?
Please refer to the attached file, spectrum photos.

Thanks in advance, appreciate your help.


Original Post

You can't do negative averaging in Route Mode, at least not on my CSI2120-2! I have to use the Analyze Mode.

You really only need about 200-Hz Fmax and LOR of about 400 to 800.

You need to use a hammer with soft impact tip and make your impact and response locations/directions where the vibration is highest. 

Don't do this test if machine has a permanent vibration monitor/trip, unless precautions are taken!!

If you cannot see any vibrations added to the spectrum during the hits (like your data), then something is wrong (like a better hammer).


Thanks Walt for your reply.

Yes, I used expert analyze mode, I meant by (Route) spectrum = what I collected before I go to expert mode.

I tried using low as well as high resolution , but both gave almost same result.

I used a hammer with a soft tip.

Maybe the hammer was not heavy enough to give the required force, I will try a larger one. I don't see the spectrum during the hits, If you please guide me how to do so?

Any suggestions, why the 1X disappear in the second trial while existed in the first one?

Thanks for your help.


As a practical matter, there must be enough force applied to the structure to excite natural frequencies within the frequency band of interest, and particularly near the dominant (problem) frequency, assuming they are actually there. The force is determined by the hammer weight, impact tip hardness, and hammer swing speed. The vibration created by the hammer should be visible in a live spectrum, whether using a trigger setting method or the rapid hit method. The is no point in proceeding with negative averaging if you cannot see the hammer vibrations in the live spectrum or have a clear visual difference between the averaged normal spectrum and the averaged impact spectrum. Negative averaging simply subtracts the normal spectrum from the impact spectrum to give the residual vibration produced by the hammer. It is not a perfect signal process due to time changes in vibrations, so it must rely on averaged spectra. The user manual describes the procedure, but you have to handle all of the practical issues to get useful results!


I second Walt's comments on the vibration introduced by the hummer force. If you're not confident that you're performing a reliable test procedure, you will doubt the  reults regardless of anything.

A possible check you can do is to test your procedure on another simpler case. Most likely you have small vertical motors on sump pits that can be manually started and stopped. Do a normal impact test while the pump is off to know the natural frequencies and have a "feeling" of the amount of force you apply to this structure. Then do the negative averaging while the pump is in service. You can play with the configurations to see their  effects. These two tests would help you if you're not a frequent user for the negative averaging.


Regards- Ali M. Al-Shurafa 


For any "regular" bump test I like to take a look at the time waveform behind the bump test spectrum.  If I don't see each hit I figure I'm not exciting the structure. The "ringing" waveform  in the TWF can often usually be probed ( cycles/second) manually to confirm the primary resonant frequencies.  I don't have any negative averaging CSi experience, but the comments by others suggest to me there would be some value for negative averaged tests as well.  

Add Reply