On the Efficiency of Different Dissimilarity Estimators Used in Damage Detection

- This paper deals with methods of interpreting the results of vibration measurement to identify structural changes in beam-like structures. We briefly presented an own developed damage assessment method, that considers a large number of frequencies for the weak-axis banding vibration modes; it allows first a precise localization and afterwards evaluation of the damage depth. The first step, recognition of the damage position, essentially consists in comparing a histogram obtained from measurement results with a series of histograms, namely the damage location indexes for all possible locations along the beam, obtained by calculus. To find the probable damage position, we tested some dissimilarity estimators like Minkowski-form Distances, Histogram Intersection and Kullback-Leibler Divergence. In order to avoid uncertainties, a graphic representation of the two histograms, one obtained by measurement and one by calculus that represent the probable damage location, are also presented. The method providing best results bases on the Minkowski Distance; it was tested for randomly selected locations, with real measurement results and with results artificially debased by noise, proving its reliability.