Firstly, a comprehensive discussion about the vibration theories for structural health monitoring, Radial Base Function (shorted for RBF) neural network and its dynamical improved algorithm were presented in this paper, and a program of RBF neural network was finished;Secondly, on the basis of Fuzzy logic theory, a novel Fuzzy Radial Base Function neural network and its training algorithm were presented. Thirdly, three SW-210 glass fiber cloth reinforced composite beams were fabricated, and the composite beams'modal frequencies were measured by an experiment method. Fourthly, delaminations were modeled by pairs of nodes with the same coordinates but different node numbers, while the modal frequencies of these beams with different delamination location and size were computed by FEM. Moreover, a novel method combining computational mechanics and neural network was demonstrated for composite health monitoring; The first six flexure modal frequencies obtained by FEM were modified by a primary revising approach and were used to train the RBF neural network and Fuzzy RBF neural network, respectively. Finally, the first six flexure experimental modal frequencies were input to the RBF neural network and Fuzzy RBF neural network to predict the demalination location and extent respectively. The predicted results of the above-mentioned networks demonstrate that Fuzzy RBF neural network was more robust and better than RBF neural network in the field of health monitoring for composite structures.
|