基于深度学习的桥梁健康监测数据有效性分析  

Validity Analysis of Bridge Health Monitoring Data Based on Deep Learning

作　　者：梁宗保[1] 柴洁 纳守勇 马天立 唐玉 LIANG Zongbao;CHAI Jie;NA Shouyong;MA Tianli;TANG Yu(School of Information Science and Engineering,Chongqing Jiaotong University,Chongqing 400074,China;Chongqing Zhong-Wan Expressway Co.,Ltd.,Chongqing 401147,China)

机构地区：[1]重庆交通大学信息科学与工程学院,重庆400074 [2]重庆忠万高速公路有限公司,重庆401147

出　　处：《重庆交通大学学报（自然科学版）》2021年第3期78-83,共6页Journal of Chongqing Jiaotong University(Natural Science)

摘　　要：作为桥梁结构健康监测系统的基石,监测数据的有效性分析是十分重要,然而现今大多数分析方法都依赖统计学理论,需要大量的领域知识,不适用于大规模数据集。提出了一种灰色关联度与深度学习相结合的方法,通过灰色关联分析对数据进行预处理,自动给定数据标签并进行标签正确性验证,结合深度学习模型DNN、DBN对数据有效性进行分析。实验表明:所提方法将监测数据有效性分析准确率提升至94.47%,具有较好的预测性能,解决了传统人工分析存在的低效率、低准确度的问题,适用于大型桥梁结构健康监测系统。As the cornerstone of bridge structural health monitoring system,the validity of monitoring data is very important.However,most of the current methods rely on statistical theory and require a lot of domain knowledge,which is not suitable for large-scale datasets.A method combining grey relation degree and deep learning was proposed.Data was preprocessed by grey relation analysis,and data labels were automatically given and validated.The validity of monitoring data was analyzed by deep learning models such as DNN and DBN.Experiments show that the proposed method improves the accuracy of monitoring data validity analysis to 94.47%,and has good prediction performance.It solves the problems of low efficiency and low accuracy existing in traditional manual analysis,and is suitable for structure health monitoring system of large-scale bridge.

关 键 词：桥梁工程 桥梁结构健康监测 数据有效性 灰色关联度分析 DNN DBN 

分 类 号：U449.7[建筑科学—桥梁与隧道工程] TP183[交通运输工程—道路与铁道工程]