Detection of structural alterations in metal bodies: An approximation using Fourier transform and principal component analysis (PCA)

Abstract

In this theme some advances have been developed, verified in the background, where attempts have been made to determine the existence of structural alterations such as perforations, defective welding and dents in metal structures; a pattern of mechanical vibration that allows to differentiate each alteration has not yet been clearly defined. In this work, the data taking was carried out taking into account the position of the sensors, two beams were added without alteration, in order to be able to interact with the five configurations, which were adopted for the experimental design. To the tests of repeated measurements, in each configuration, analysis (ANOVA) was used for the validation of NULL hypotheses, and thus to determine the number of tests to be treated. After having the defined matrices representing each configuration, in each anomaly, it is necessary to apply the principal component Analysis (PCA), to the data obtained by the calculation of the fast Fourier transform (FFT). And thus, determine the number of components by means of three Criteria (Jollife, Kaiser and PVA), using a classification algorithm, which evaluates the percentage of classification vs lower standard deviation. In this analysis the descriptors were not calculated but the main components of each criterion were taken as a description tool. The process of extraction of characteristics was fundamental to determine the proper configuration in each alteration (fissure, welded, perforated, deformed). On the other hand, statistical parameters were calculated (average, standard deviation, variation factor, Euclidean distance) of each anomaly. Taking as descriptors. Finally, it was shown that the Jollife criterion is the one that allows to better differentiate between components associated with each alteration studied