Abstract
Health monitoring is an important issue especially for a structural system being in operation for a long time. In this study, we deal with an approach to damage identification of structural system by means of a multi-layered neural network. The approach is based on the change in some specific characteristics caused by the damage of the structural members. The relation between the damage and the characteristics is implemented with a multi-layered neural network; the error back-propagation with fixed connection weight is adopted to solve the inverse problem for the damage identification. We discuss the type of characteristics used for the damage identification. Numerical experiments with truss structures are conducted and the feasibility of the proposed approach is demonstrated.
