基于遗传-神经网络的无粘结部分预应力高强混凝土梁延性预测模型  

Prediction Model of Ductility of Unbonded Partially Prestressed Concrete Beam with High Strength Based on Genetic Algorithm and Neural Network

作　　者：李哲[1] 胡立黎[1] 

机构地区：[1]西安理工大学水利水电学院,陕西西安710048

出　　处：《西安理工大学学报》2005年第3期281-284,共4页Journal of Xi'an University of Technology

摘　　要：尝试利用遗传-神经网络模型,对无粘结部分预应力高强混凝土梁的延性进行预报。选用23个试件为学习样本,3个试件为测试样本。模型得到了很好的预测结果。模型中使用了最优保存策略和实数编码。同时,为了防止“早熟”现象的发生,提出了一种新的改变适应度值的方法。In this paper,the genetic algorithm-neural network model is used to predict the ductility of unbonded partially prestressed concrete beam with high strength. 23 beams are selected as learning samples and 3 beams as test samples. The better results are obtained from the model. Elitist strategy and the real number coding are used in the model. Meanwhile,in order to prevent “remature convergence” a new way to change fitness value is suggested.

关 键 词：遗传算法 神经网络 高强混凝土梁 预应力 延性 

分 类 号：TU378.2[建筑科学—结构工程]