APPLICATION OF NEURAL NETWORK TO EIGEN-PARAMETER BASED DAMAGE DETECTION IN MULTI-COMPONENT SANDWICH SHIP HULL STRUCTURES

摘要

Damage in sandwich composite structures is simulated using eigen-parameter analysis. It is shown that the fundamental displacement mode shape curvatures of sandwich structure can be used to identify damage sites, which show up as a sharp irregularity in the curve. A 3-D orthotropic finite element analysis is used to simulate damage reflecting reduction in stiffness at critical locations. Neural network is subsequently trained and used to detect multi-site damage in sandwich composite structures. This damage detection technique is shown to work well for both isotropic and orthotropic materials, such as metals, laminated and sandwich composites.