摘要
提出了基于神经网络的结构地震反应仿真方法,探讨了仿真基本步骤中样本集的准备、目标函数的选取、网络拓扑结构的构建、隐层神经元数目的确定、训练方法的选择以及提高泛化精度的措施等若干实际问题,并通过算例分析验证了本方法的可行性。
A method for the simulation of structural seismic responses based on neural networks is presented in the paper. Several problems, involving collection of samples, selection of target functions, construction of neural networks' topology, determination of the number of neural cells, choice of training method, and measures to improve generalization precision, are discussed. The results of a structural simulation show that the method presented is feasible.
出处
《世界地震工程》
CSCD
2003年第3期144-149,共6页
World Earthquake Engineering
基金
国家自然科学基金资助项目(50008017)
关键词
建筑结构
地震反应
仿真
神经网络
neural networks
simulation of structural seismic responses
BP neural networks
