Peak-hold equivalent amplitude for transient vibration

[CmRock314]

I have a question regarding transient vibration data I received that was processed into a peak-hold equivalent amplitude (units = g). I have come across peak-hold before which is a type of "averaging" that retains the highest values from each estimate in random vibration overlap processing and FFT frequency 'bins' as opposed to linear averaging. Understandably, peak-hold is used in transient vibration as the vibration is changing in frequency content over time as opposed to stationary random vibration.

I have processed the same set of data into random vibration using linear and peak-hold averaging. The peak-hold (envelope) PSD is lower than the peak-hold equivalent amplitude processed by someone else. It doesn't surprise me since my peak-hold PSD is expressed in units of G^2/Hz and the peak-hold equivalent amplitude is expressed in G. My question is, how does one convert this transient vibration into the correct units of peak-hold equivalent amplitude?

I have never come across this before. I thought it may be as simple as multiplying by the delta-frequency and taking the square root, but this doesn't appear to come out the same.

 

[TheQuietOne]

...I thought it may be as simple as multiplying by the delta-frequency and taking the square root...

Well I would just keep trying different things and figure it out on my own. Otherwise I would just find a textbook or something (I found a lot of PDFs when I googled it)

 

[CmRock314]

Well I would just keep trying different things and figure it out on my own. Otherwise I would just find a textbook or something (I found a lot of PDFs when I googled it)

I appreciate the thoughts. I have found a lot of information. One of the things I ended up discovering is PSD was not actually used. It was simply just an FFT of a transient sine vibration that used multiple overlaps (90%) to capture the peaks of the transient; the peaks from each estimate are 'held' (i.e. peak hold) as opposed to averaging. My answer is still not matching exactly, but I'm thinking some data manipulation to the time history may have taken place. The important part is the peaks are matching up quite closely, but the other I'm comparing to has a little more dynamic range.