基于CART算法的桥梁损伤动力数据分析方法  

Dynamic data analysis method of bridge damage based on CART algorithm

作　　者：项长生[1,2] 刘海龙 赵驰 苏天涛 XIANG Chang-sheng;LIU Hai-long;ZHAO Chi;SU Tian-tao(School of Civil Engineering,Lanzhou University of Technology,Lanzhou 730050,Gansu,China;Western Center of Disaster Mitigation in Civil Engineering,Ministry of Education,Lanzhou University of Technology,Lanzhou 730050,Gansu,China)

机构地区：[1]兰州理工大学土木工程学院,甘肃兰州730050 [2]兰州理工大学西部土木工程防灾减灾教育部工程研究中心,甘肃兰州730050

出　　处：《长安大学学报（自然科学版）》2023年第4期50-59,共10页Journal of Chang’an University(Natural Science Edition)

基　　金：国家自然科学基金项目(51868045)。

摘　　要：针对传统损伤检测方法难以准确识别桥梁结构损伤程度的不足,利用分类回归树(classfication and regression tree,CART)算法在数据挖掘方面的优势,通过计算基尼系数选取合适的特征对数据样本进行分类,提出一种损伤检测方法对桥梁的损伤动力信息进行学习分类。首先利用附加质量方法构建结构动力响应数据集,计算附加质量的模态应变能指标ξ,对结构损伤进行定位;然后将ξ作为决策树的特征,输入到CART算法中进行训练,对损伤程度进行分类和识别,并对该方法进行抗噪性验证,最后通过简支梁和连续梁算例进行验证分析。研究结果表明:基于附加质量的损伤识别指标能准确定位损伤,且CART算法能够有效识别桥梁结构的损伤程度,在2%、5%噪声水平下,2种算例的损伤程度识别准确率分别达到99%、95%和95%、90%以上,具有较高准确率和较强鲁棒性;该方法为桥梁结构损伤程度识别提供了一种新的参考。Aimed at the defect that traditional damage detection methods were difficult to accurately identify the damage degree of bridge structure,a damage detection method was proposed,which can learn and classify the damage dynamic information of bridge by calculating Gini coefficient to sort data samples through selecting appropriate features,great advantages of CART decision tree algorithm were taken in data mining.Firstly,the add-mass method was used to construct structural dynamic response data set,and modal strain energy index ξ of add-mass was calculated to locate structural damage.Then,ξ was taken as the feature of decision tree and input into the CART algorithm for training,and then the damage degree was classified and identified,meanwhile,anti-noise of this method was verified.Finally,the simple supported beam and continuous beam were used for verification analysis.The results show that the damage identification indexξbased on added mass can accurately locate the damage,and the CART classifier can effectively identify the damage degree of the bridge structure.The damage classification accuracy of the two cases can reach 99%,95% and 95%,90% with 2% and 5% noise levels,respectively,which means this method achieves a high accuracy and strong robustness,and can provide a new reference for identifying damage degree of bridge structures.4 tabs,19 figs,23 refs.

关 键 词：桥梁工程 损伤识别 CART算法 附加质量法 模态应变能 

分 类 号：U411[交通运输工程—道路与铁道工程]