首页|基于RF建立的双层桥墩矢量式损伤极限状态能力模型

基于RF建立的双层桥墩矢量式损伤极限状态能力模型

扫码查看
为了确定双层桥墩的抗震能力,基于随机森林(RF)算法构建了双层桥墩的矢量式损伤极限状态能力模型.将地震动激励角视作为一个符合均匀分布的随机变量,通过大量推倒(Pushover)分析构建双层桥墩的能力样本数据库,以训练其能力值预测模型,并利用SHAP进行特征重要性分析.结果表明:双层桥墩的能力阈值不服从对数正态分布,且分布参数明显不同于以往研究的建议值;矢量式损伤极限状态能力模型能有效识别双层桥墩能力的分层现象,决定系数R2>0.95,具有良好的预测性;地震动激励角显著影响双层桥墩的抗震能力.相比于上层墩柱,双层桥墩严重损伤极限状态的目标值更容易受到下层墩柱特征参数的影响,可适当关注下层墩柱属性以增强双层桥墩整体的抗震性能.
A vectorial damage limit-state capacity model for double-layer piers based on RF
To determine the seismic capacity of double-layer piers,a vectorial damage limit-state capacity model was constructed based on the random forest(RF)algorithm.The seismic excitation angle was treated as a uniformly distributed random variable,and a large number of Pushover analyses were conducted to build a capacity sample database for double-layer piers,which was used to train a predictive model for their capacity values.SHAP was employed for feature importance analysis.The results indicate that the capacity threshold of double-layer piers does not follow a lognormal distribution,and the distribution parameters are significantly different from previously suggested values.The vectorial damage limit-state capacity model can effectively capture the stratification of capacity values of double-layer piers,with the coefficient of determination R2>0.95,demonstrating reliable prediction performance.The seismic excitation angle notably influences the seis-mic capacity of double-layer piers.Compared with upper columns,the target value for the severe damage limit-state of double-layer piers is more susceptible to the characteristics of lower columns.Therefore,attention should be given to the attributes of lower columns to enhance the overall seismic performance of double-layer piers.

bridge engineeringdouble-layer piersseismic excitation anglerandom forest(RF)interpret-ability

郭威佐、王克海

展开 >

东南大学交通学院,南京 211189

交通运输部公路科学研究院,北京 100088

桥梁工程 双层桥墩 地震动激励角 随机森林(RF) 可解释性

中国地震局工程力学研究所基本科研业务费专项资助项目内蒙古自治区交通运输厅交通建设科技资助项目山西省交通运输厅交通建设科技资助项目

2021D40NJ-2020-172020-2-01

2024

东南大学学报(自然科学版)
东南大学

东南大学学报(自然科学版)

CSTPCD北大核心
影响因子:0.989
ISSN:1001-0505
年,卷(期):2024.54(3)