Repeatability and reproducibility of characteristic features measuredby laser Doppler vibrometry for on-line diagnostics

摘要

It is analyzed the statistical dispersion of characteristic features measured by Laser Doppler Vibrometry (LDV) in on-linediagnostic applications, with reference to on-line detection of mechanical defects of washing machines. The paperpresents two complementary approaches: a) experimental evaluation of repeatability of measured features according tothe Guide to the Expression of Uncertainty in Measurement-GUM; b) Montecarlo simulation of uncertainty propagationacross the on-line test station. Experiments consist in a test bench which simulates the vibration of a washing machine,by playing back on a shaker a real signal acquired on-line and taking repeated measurements, so that a statistical analysisis performed about dispersion of diagnostic features. The analysis is repeated by varying the scattering characteristics ofthe vibrating surface, so to evaluate the effect of signal quality. The Montecarlo approach consists in modeling thepropagation of uncertainty across the various elements of the measurement chain, up to the computation of features. Theinfluence of LDV, Data Acquisition device (DAQ) and processing software have been taken into account. Results allowto estimate Repeatability and Reproducibility (R&R) of a typical set of characteristic features used in industrialdiagnostics and to discuss uncertainty of similar diagnostic procedures.