Abstract
In general, hysteresis models that are applied to a numerical analysis part of substructure on-line tests do not refer to an experimental behavior of members/subassemblage under loading tests on real-time basis. The objective of this study is to develop a new experimental technique for substructure on-line tests based on nonlinear hysteresis characteristics estimated with a neural network. New learning algorithms for the network applicable to substructure on line tests afe proposed focusing on input layer components and a normalization method for input data, and their validity is examined through several numerical analyses. The results show that the new algorithms proposed herein successfully reoroduce the dynamic behavior of model structures.
