基于RBF神经网络的结构动力响应预测

介绍了径向基函数(RBF)神经网络学习速度快,动态仿真性强,具有较强的输入输出映射功能和全局最优逼近的结构特点.针对快速预测结构动力响应有助于克服结构振动控制中时滞效应的特点及BP网络存在的问题,应用RBF网络对结构的位移、加速度进行了预测,并采用BP网络作对比研究.仿真结果表明RBF神经网络训练速度快,精度高,可及时为主动控制建筑结构响应提供较为准确的优化性能指标,从而为实现在线实时控制结构响应提供优良的保证.

Prediction of structural dynamic response based on RBF neural network

Constitutive features of radial basis function(RBF),such as fast neural network learning,strong dynamic simulation ability,good input/output mapping function,and global optimal approaching,were introduced.Taking into consideration of the feature that fast prediction of structural dynamic response is useful for overcoming the time-lag effect in structural vibration control,and aiming at the existing problems in BP network,the RBF network was employed to predict the displacement and acceleration of the construction.The RBF network was also used to conduct comparison investigation.Simulation result showed that there were fast training and high accuracy with RBF neural network,so that it could provide more accurate optimal performance indexes for timely active control of structural response of the buildings and good guarantee for implementing on-line real-time control of structural response.