Acoustic analysis of vibration automated quality control by inspection of free vibration modes using autoregressive feature extraction (AR)

摘要

Inspection of eigenfrequencies ofa freely vibrating body is a classical tool of quality control. Tinyhidden irregularities and damages produce detectable spectral shifts. In the context of automatedinspection by use and pattern classification, the choice among available feature extraction methods isnot obvious. This paper compares two extraction methods - Autoregressive Analysis (AR) and FastFourier Transform (FFT). Within the AR family only Modified Covariance (MODCOV) wasconsidered, an AR model particularly adequate for inspecting weakly damped multiresonant spectra.