Cyclostationarity in Condition Monitoring: 10 years after

摘要

Mechanical signals produced by rotating or reciprocating machines are remarkably well modeled as cyclo-stationary processes. The benefits offered by the cyclostationary framework in Condition Monitoring have been explored since more than one decade; research in this domain has resulted in the proposal of a quite general methodology and processing tools to analyze mechanical signatures of machines. The state-of-the-art in cyclostationarity, as it stood a few years ago, still faced a few strong limitations. The aim of this communication is to introduce recent solutions to these issues. First, the framework is generalized to angle-time cyclostationarity, an enlarged class which perfectly models signals of machine operating under varying regime. Second, a fast estimator is proposed for the spectral correlation density, a central tool of cyclic spectral analysis. Finally, solutions are given for the synthesis and the extraction of a cyclostationary signal. This is briefly illustrated on vibration signals.