Adequacy of first mode shape differences for damage identification of cantilever structures using neural networks

Abstract

Damage identification of structures has attracted attention of researchers due to sudden collapse of in-service structures. Modal parameters and their derivatives have been widely employed in the proposed damage identification techniques. However, mode shape differences have been shown to be an ideal damage indicator when used as the input vector of neural networks. Since measurement of higher-order mode shapes is very difficult to be acquired reliably, this study investigated the adequacy of using only the first mode shape differences for damage identification using artificial neural networks. Results of numerical and experimental studies on a cantilever beam indicated that the first mode shape differences alone can accurately localize imposed damages. Damage intensity at the lower levels of cantilever beam was predicted with less than 15% error; however, prediction of damage intensity at the free end of the beam encountered large discrepancies. It was also found that damage localization was successful even when the first mode shape differences were measured at few points along the beam.