摘要
为进行大跨度PC斜拉桥、悬索桥等复杂结构的优化设计、可靠性分析或模型修正,提出基于神经网络的结构快速分析方法。通过对比分析不同样本集构造方法对结构分析精度与效率的影响,认为均匀试验设计法是构造网络训练样本的最优方法。基于Matlab工具箱函数newrb建立招宝山大桥平面分析的径向基函数网络模型,该模型含45个输入层节点和2个输出层节点。根据均匀试验设计法生成180个训练样本,利用有限元分析软件ANSYS进行参数化批量分析,得到样本的模拟试验结果,采用OLS法对径向基函数网络进行训练,用训练好的网络预测结构响应。结果表明:该神经网络模型满足结构快速分析的精度要求,与有限元分析结果吻合良好。
The neural network-based fast analysis method is proposed for optimal design, reliability analysis, and model updating of complex structures such as the large scale pre-stressed concrete cable-stayed bridges and suspension bridges. The accuracy and efficiency of neural network models based on different sampling techniques are evaluated, including techniques based on experimental design theory, random selection and rotating sampling, etc. , and the uniform design is verified as the optimal sampling techniques. The toolbox function newrb of Matlab is used to establish the radial basic function （RBF） network for plar nar analysis of Zhao Bao Shan Bridge based on the presented method. The proposed RBF network with 45 neurons in input layer and 2 neurons in output layer is trained by 180 simulation samples which sampled by the uniform design algorithm. The general finite element program ANSYS is used to get the simulation results of the samples and the Orthogonal Least Square （OLS） learning algorithm is adopted in present work. The differences between the structural response from the prediction of neural network and the finite element analysis are unconspicuous, which verified the prediction capabilities and the computational advantages of the proposed neural network.
出处
《中国铁道科学》
EI
CAS
CSCD
北大核心
2007年第1期33-39,共7页
China Railway Science
关键词
径向基函数网络
神经网络模型
均匀设计
结构分析
斜拉桥
Radial basic function network
Neural network model
Uniform design
Structural analysis
Cable-stayed bridge
