基于神经网络的高强钢管砼柱抗震性能研究  

Study on seismic properties of high strength CFT columns based on neural network

作　　者：李兴权[1] 魏华[1] 冯艳峰[1] 高华国 

机构地区：[1]沈阳工业大学建筑工程学院,沈阳110023

出　　处：《沈阳工业大学学报》2005年第6期682-685,共4页Journal of Shenyang University of Technology

基　　金：教育部留学回国人员科研基金资助项目(20046293)

摘　　要：利用人工神经网络的自适应性、容错性、模糊性,建立了一个三层BP神经网络模型对高强钢管混凝土核心柱抗震性能进行了研究.在试验研究的基础上,用30组数据作为学习样本进行训练,模型的误差收敛快,学习效果很好.用另外8组数据作为测试样本,对柱的位移延性进行预测,结果和试验值吻合良好.与其它文献的柱位移延性公式计算值相比,大大提高了预测值的精度.该方法可作为实际结构设计的一种辅助手段.Artificial neural network is of self-adaptability, fault toleration and fuzziness. It is suitable to solve the seismic properties of high strength reinforced concrete columns with concrete filled steel tube core （HRCCFT）. A three-layer back-propagation network model was build up to study the seismic properties of HRCCFr. The model was trained according to 30 sets of experimental data. The network convergence is fast. The model was verified by 8 groups of experimental data. The results show the predicted values of displacement ductility are in good agreement with test values. The precision of model is better than that of formula from other reference. This method is good enough to be used as an auxiliary method for structure design.

关 键 词：人工神经网络 容错性 高强钢管混凝土柱 延性 抗震性能 

分 类 号：TU201.4[建筑科学—建筑设计及理论]