Updating structural parameters: an adaptive neural network approach

摘要

An adaptive neural network (NN) method is proposed for the mode l updating and the damage detection of structures. The NN model is first trained off-line and is retrained during iteration if needed. An improved back-propagation learning algorithm with a jump factor and a dynamical learning rate is developed to facilitate the training. The concept of orthogonal array is adopted in this study to reduce the number of training samples required. Two examples illustrate that the proposed technique is quite useful for the model updating and the damage detection of structures.