New features for diagnosis and prognosis of systems based on empirical mode decomposition

We present a new procedure to generate additional features for system diagnosis. The procedure is based on empirical mode decomposition of measured signals obtained by monitoring the relevant state of a system. This procedure is different from the conventional procedures for defining features, which are generally obtained using the statistics of the measured signal, the matched filter outputs, and the wavelet decomposition of measured signals. Features derived by this new procedure complement the existing features for diagnosis, and therefore they should improve performance of the classifier used to diagnose systems. We illustrate the procedure by generating new features for diagnosis of the AH64A helicopter transmission assembly.