Vibrations and sounds generated by reciprocated machines or by their parts
strongly depend on the rotation speed of the main shaft of the tested recip
rocating system. At the testing or at common performance of the reciprocate
d machines, their rotation speed is usually changing. With regard to this f
act, signals produced by reciprocating machines are non-stationary ones. Th
erefore, conventional time-invariant methods of their spectral or bispectra
l analysis are frequently unable to provide meaningful results. In order to
solve this problem in the field of polyspectral signal analysis, the order
bispectrum analysis is proposed in this contribution. This approach is bas
ed on the bispectrum estimation from the signal which is a function of the
angle of roll of the main shaft of reciprocated machine. A digital represen
tation of this signal can be obtained by resampling of the signal convenien
tly sampled in the time domain. The advantages of the order bispectrum appl
ication in comparison with that of the conventional bispectrum approach is
illustrated based on the example of an engine set classification. (C) 2000
Academic Press.