Utilização De Estatísticas de Ordem Superior Na Detecção De Falhas Estruturais Em Vigas

Abstract

In this article, it is proposed a method for detection of structural faults or damages using Higher-Order Statistics (HOS). Taking vibration signals measured using an accelerometer, characteristics are extracted by cumulants and selected with the Fischer's Discriminant. Subsequently, the classification is carried out with the aid of Supporting Vector Machines, aiming to determine the presence or not of faults in the structure. The method was experimentally tested in a cantilever beam, providing high detection rates. Advantages presented are low computational cost and no requirement of information about the damaged structure.