SINGULAR VALUE NOISE IN RSVD ALGORITHM FOR DAMAGE DETECTION

Horacio Valadares Duarte¹; Lázaro Valentim Donadon²

¹ Departamento de Engenharia Mecânica da Universidade Federal de Minas Gerais; ² Departamento de Engenharia Mecância da Universidade Federal de Minas Gerais

doi:10.20906/CPS/CILAMCE2017-0281

Resumo

The Robust Singular Value Decomposition algorithm (RSVD) was used to identify damaged structure using frequency response functions (FRFs) data from an aluminium plate. The routine uses robust statistics and the analysis was done in a data set from healthy and damaged structure. The effectiveness of the RSVD method was noticeable linked to optimum singular value basis reduction. The main problem experienced was to identify the correct noise level that has shown to be critical for data obtained in this work. A review for this kind of application has shown that there is no theoretical procedure to determine the optimal reduction for singular value basis. From experimental data this work will outline a practical procedure that presented good results.

Palavras-chave: Singular value decomposition; Structural health monitoring; plate structure; damage detection; noise level