Fault identification in rotating machinery using the correlation dimension and bispectra (Q5960055)

This paper reports on the application of nonlinear dynamics and higher-order spectra, with particular regard to the correlation dimension and bispectra in rotating machinery fault identification. The performance of the two methods is evaluated from a pratitioner's viewpoint. The correlation dimension of a nonlinear dynamical system is of value because it provides an estimation of the number of degrees of freedom that the system possesses. It can provide some intrinsic information on underlying dynamic systems, and can be used to classify faults. Therefore, correlation dimension is helpful as an automatic fault detection procedure. Bispectral analysis offers a method for determining the nonlinear coupling and energy exchange between Fourier modes, and explains the origins of spectra peaks at certain values in frequency spectrum.