Complex-valued Signal Processing for Condition Monitoring

This paper concerns the analysis of stationary complex-valued signals, and its application to the condition monitoring field. More particularly, it is confirmed that usual tools such as correlation function or power spectrum are insufficient to entirely describe the statistical and geometrical properties of complex-valued signals. In that case, additional time and spectral domain quantities, namely the pseudo correlation function and the pseudo spectrum, have to be used. They lead to new important information, and allow to entirely describe second-order properties of complex-valued signals. Once these quantities theoretically defined, this paper describes their use in several condition monitoring problems, where 2 dimensional signals are considered as complex-valued signals. The results obtained in these examples show that these quantities contain crucial information for condition monitoring not given by usual quantities such as the power spectrum, especially in terms of fault localization.