基于极限学习机的高桩码头结构监测传感器优化布置研究  

Research on Optimal Arrangement of Monitoring Sensors for High⁃piled Wharf Structure Based on Extreme Learning Machine

作　　者：戴志培 李凯 苏静波[2] 季晓堂 刘旭 DAI Zhipei;LI Kai;SU Jingbo;JI Xiaotang;LIU Xu(No.3 Engineering Co.,Ltd.of CCCC Third Harbor Engineering Co.,Ltd.,Nanjing,Jiangsu 210011,China;Hohai University,Nanjing,Jiangsu 210098,China;CCCC Shanhai Harbour Engineering Desing&Research Institute Co.,Ltd.,Shanghai 200032,China)

机构地区：[1]中交三航局第三工程有限公司,江苏南京210011 [2]河海大学,江苏南京210098 [3]中交上海三航科学研究院有限公司,上海200032

出　　处：《施工技术（中英文）》2024年第16期140-145,共6页Construction Technology

基　　金：国家自然科学基金(51679081)。

摘　　要：高桩码头结构的全寿命周期监测是一个庞大的系统工程,监测点的优化布置尤为关键。结合有限元计算建立了ELM神经网络模型,同时采用K⁃flod法确定ELM模型超参数,借助Filter框架下的SBS策略评价监测点在码头状态判断中的重要性,使用Wrapper法中的全局最优搜索策略给出选定监测点下的优化布置方案。结果表明,码头状态预测测点重要性评价与全局优化结果具有相似性,该方法可对高桩码头结构的监测点进行优化布置。The whole life cycle monitoring of high⁃piled wharf structure is a huge system engineering,and the optimal arrangement of monitoring points is particularly critical.The ELM neural network model is established by finite element calculation,and the hyperparameters of ELM model are determined by K⁃flod method.The importance of monitoring points in the judgment of wharf state is evaluated by SBS strategy under Filter framework.The global optimal search strategy in Wrapper method is used to give the optimal layout scheme under selected monitoring points.The results show that the importance evaluation of the monitoring points of the wharf state prediction is similar to the global optimization results.This method can optimize the layout of the monitoring points of the high⁃pile wharf structure.

关 键 词：高桩码头 健康监测 神经网络 优化布置 

分 类 号：U656[交通运输工程—港口、海岸及近海工程]