Hi
I guess the repetition frequency of the coin taps can be found from the FFT autospectrum?
If you want a triggered evaluation, you can define a trigger event and use only a selected time portion thereafter for evaluation.
A suggestion - coin tapping tests are sometimes used to identify defects by the occurence of 'raw sound'. Such defects often are hard to detect by spectral analysis but, surprisingly, turn out to arise from zero-crossing distortion and to, even more surprising, be detectable 60 dB below the signal level for most listeners. This was empirically found found in the late 1960's by BBC and other radio stations and resulted in the ITU-R 468 weighting filter to enhance such defects.
See http://en.wikipedia.org/wiki/ITU-R_468_noise_weighting for more information. The BBC paper is referenced in the wikipedia and makes for quality reading.
Anyhow, the ITU-468-R filter is implemented in Sound & Vibration Measurement Suite and may or may not be what you want in quantifying the defects. Perhaps it is worth giving it a try?
About finding the actual location of where the tap occurs, this may be more difficult as the dominant noise will be reverberantly radiated from the plate which happens at various angles for various frequencies due to the coincident nature (Snell's law) of sound radiation from bending waves, i.e. there may not be a clear wave front from which to detect the tap.
http://en.wikipedia.org/wiki/Snell%27s_law
Should you have a clear acoustic impact to use, you can try trilateralization using 4 microphones and the difference in arrival time between the wave fronts.
Sincerely
Claes