Using Multivariate Energy Distance on Frequency-based Damage Detection

Resumo

The Frequency-based Damage Detection Method using FRFs (Frequency Response Functions)
in a high-frequency range can identify a structural failure of about 0.16% and 0.34% of the total mass.
These results demonstrate the effectiveness and reliability of the method to track the problem in its

initial stage. The Robust Singular Value Decomposition algorithm (RSVD) is employed in Frequency-
based Damage Detection Method to find out a damaged subset from a set composed by reference and

damaged data. Although effective, this procedure has some drawbacks, it is time consuming and the
method performance is linked to optimum singular value basis reduction. The main objective of this
work is to select an algorithm to overcome these less acceptable features. In order to achieve this goal
there are a short review of the main methods employed in data classification, parameter extraction. The
selection criteria is to pick out those that present simple algorithms and that are based on metric distances.
From those review the Multivariate Energy Distance Correlation was selected. Using experimental data
this Energy statistic method are compared with Robust Singular Value Decomposition (RSVD).