Title: Damage Metric Algorithms for Active-Sensor Structural Health Monitoring

The availability of a reliable damage metric is critical structural damage diagnostics and prognostics. Several methods, from overall-statistics of the entire spectrum, to features-based probabilistic neural networks (PNN) are illustrated on: (i) experimental data taken on simple-geometry calibration specimens; and (ii) experimental data recorded on aging aircraft panels. The electro-mechanical (E/M) impedance technique using piezoelectric-wafer active sensors bonded to the structure, was used to collect experimental data, in the form of very high frequency (hundreds of kHz) spectra, representing the structural dynamics in the sensor’s neighborhood. The overall statistics approach using the correlation coefficient difference damage metric could correctly classify near-field spectra using the dereverberated response and, but it could not classify medium field spectra. The features-based PNN could correctly identify both medium-field and near-field data, but required a two-stage approach: (a) features extraction; and (b) classification