Condition monitoring and failure prediction in chaos

Condition monitoring relies heavily upon signal processing techniques which were developed to deal with linear systems and are dominated by Fourier analysis methods. This dominance was founded on the Cooley-Tukey fast Fourier transform algorithm, which allows the computation of the spectrum of a sampled signal in real time. As systems are stressed beyond their linear regime, complex and unforeseen behaviour can result. In the context of condition monitoring this nonlinear behaviour may be benign or may indicate system deterioration or failure. Unfortunately, linear spectral methods applied to the outputs of nonlinear systems can result in featureless spectra which obscure the potential simplicity of the underlying nonlinear mechanism. What is needed is an appropriate nonlinear approach to signal process ng providing information upon which useful decisions can be made. The article describes, by way of an example, a model-based approach to nonlinear signal processing.<
>