Cyclostationary Processing of Vibration and Acoustic Emissions for Machine Failure Diagnosis

Abstract

The use of vibrations and other variables to infere the mechanical condition of machines is a common practice nowadays. Several equipments for measurement and processing are available. Most of them offer FFT spectrum as the main tool for analysis, thus allowing the assessment of the stationary part of the signal only. Although this might be sufficient in some cases, it is certainly inappropriated in others. Recent advances in signal processing have opened the possibilities for analyzing a special type of non-stationary signals, called cyclostationary signals. It has also been shown that the behaviour of machines can be highly cyclostationary in some cases. Moreover, being stationarity a special case of cyclostarionarity, the advantages of the cyclostationary approach become evident. Still, even under consideration of these facts, the use of cyclostationarity appears to be still restricted to the scientific community, being its use in the industry far from being a reality. This chapter presents the concept of cyclostationarity, its terminology and its relation with traditional signal processing tools in a descriptive way. Two examples of real data analysis from the cyclostationary viewpoint are also presented.