摘要
为了预测不同结构参数下RC柱的恢复力,提出一种泛化神经网络算法的RC柱恢复力预测方法。通过给出网络输入变量选取方法,预测不同轴压比、纵筋配筋率、体积配箍率和长细比下的RC柱恢复力。结果表明:采用该方法预测所得的RC柱滞回曲线与参考解几乎完全吻合,不同长细比时神经网络所预测出的恢复力均方根误差最大值为0.02535;不同纵筋配筋率时神经网络所预测出的恢复力均方根误差最大值为0.01079;不同体积配箍率时神经网络所预测出的恢复力均方根误差最大值为0.01597;不同轴压比时神经网络所预测出的恢复力均方根误差最大值为0.04916;不同截面尺寸时神经网络所预测出的恢复力均方根误差最大值为0.02347。该算法具有较高的恢复力预测精度和泛化能力。
The paper proposes a new algorithm based on generalized neural network algorithm to predict the restoring force of RC columns with different structural parameters.The method works by predicting the restoring force of RC columns with different axial compression ratio,longitudinal reinforcement ratio,volume stirrup ratio and slenderness ratio by selecting network input variables.The results show that the method enables the predicted hysteretic curve of RC column almost identical with the reference solution,and the neural network permits the prediction of the maximum root mean square error of restoring force of about 0.02535 for different slenderness ratio;of about 0.01079 for different longitudinal reinforcement ratio;of about 0.01597 for different volume stirrup ratio;of 0.04916 for different axial compression ratio,and of 0.02347 for different section sizes.The algorithm features a higher prediction accuracy and generalization ability.
作者
王涛
周天楠
孟丽岩
Wang Tao;Zhou Tiannan;Meng Liyan(School of Architecture & Civil Engineering, Heilongjiang University of Science & Technology, Harbin 150022, China)
出处
《黑龙江科技大学学报》
CAS
2020年第6期708-716,共9页
Journal of Heilongjiang University of Science And Technology
基金
国家自然科学基金项目(52078398,51978213)。
关键词
RC结构
恢复力模型
泛化能力
神经网络算法
reinforced concrete structure
resilience model
generalization ability
neural network algorithm
