BP NEURAL NETWORK MODELING OF A MAGNETORHEOLOGICAL DAMPER

摘要

Magnetorheological (MR) dampers are regarded as one of the most promising devices in vibration mitigation of civil structures against severe earthquakes and strong winds. To implement vibration control via MR dampers,dynamic models that can accurately capture the dampers' inherent nonlinear behaviors should be developed. In the present paper,using the results of mechanical property tests of a MR damper,both forward and inverse models of the damper are developed based on back propagation (BP) neural networks,respectively. In addition,the generalization ability of the two neural network models is validated,and the sensitivities are analyzed on the assumption that noise arises from additive and multiplicative perturbations.