Abstract
A neural network approach is proposed to predict the nonlinear hysteretic behavior and dynamic response of wooden tenon joints using measured static hysteresis loop data. A multi-layered back propagation neural network is employed to determine the restoring force at the next loading step as an output with the displacement responses and the corresponding restoring forces at the past few loading steps as inputs to the neural network. In the training stage of neural networks, measured hysteresis loop data are ajusted such that the hysteresis parameters of each joint become identical with each other. The prediction accuracy can be improved significantly by adjusting hysteresis loop data with different hysteresis parameters and by taking an average of restoring forces predicted by different neural networks.
