THE INTERPRETATION OF THE BISPECTRA OF VIBRATIONAL SIGNALS .1. THEORY

Bispectral analysis is emerging as a new powerful technique in signal processing, offering insight into non-linear coupling between frequencies and having potential applications in many areas where traditional linear (i.e. power spectral) analysis provides insufficient information. However, it is more difficult to interpret bispectral features than power spectral features, and this has hindered the applications of the theory. In this paper a normalised bispectral measure is used in the analysis of vibration signals containing periodic components and noise. In Part I the bispectral features associated with various signal types are derived, and practical considerations regarding the choice of sampling rate are considered. Part II describes the application of this theory to experimental data from a variety of vibration sources, and discusses the possible applications of bispectral analysis in machine condition monitoring.