基于机器学习的中小跨径公路梁桥抗震设计评价方法研究  

Study on Seismic Design Evaluation Methods for Highway Medium-span and Small-span Girder Bridges Based on Machine Learning

作　　者：王克海[1,2] 鲁冠亚 张盼盼 WANG Ke-hai;LU Guan-ya;ZHANG Pan-pan(Research Institute of Highwary,Ministry of Transport,Beijing 100088,China;School of Transportation,Southeast University,Nanjing Jiangsu 210096,China)

机构地区：[1]交通运输部公路科学研究院,北京100088 [2]东南大学交通学院,江苏南京210096

出　　处：《公路交通科技》2019年第2期74-84,共11页Journal of Highway and Transportation Research and Development

基　　金：科技部国际科技合作项目(2009DFA82480);交通运输部西部交通建设科技项目(2009318223094);交通运输部公路工程行业标准项目(JTG-C-201012);中央级公益性科研院所基本科研业务费项目(2016-9018)

摘　　要：为了探究机器学习方法在桥梁抗震中实现的基本思路,简要回顾了桥梁抗震分析理论与技术的发展和现状,综述了机器学习的应用领域,特别是在土木工程领域中的应用现状。介绍了机器学习的概念,总结了其关键因素和搭载平台,通过简单实例说明了机器学习的常用方法和代表算法。首先,总结了我国公路中小跨径桥梁多采用简支梁和连续梁的梁桥形式,同时统计了汶川地震中该类桥型的震害现象,包括桥墩、支座和挡块、桥台根据震害现象的破坏等级划分。其次,总结了国内外学者针对支座和挡块、桥墩、桥台开展的一系列抗震性能试验研究,回归出用于抗震分析的本构关系,进而得到桥梁各构件(包括基础)的抗震设计因素。最后,阐述了机器学习方法辅助该类桥梁抗震设计和评价中实现的总体思路,说明面向桥梁抗震任务的机器学习主要有两方面的工作:第一项是收集数量可观的桥梁设计资料,建立数据集;第二项是数据挖掘,包括对原始数据的处理,调试或开发合理的机器学习算法模型,并简要探讨了现有基于性能的概率抗震设计和评价方法对我国桥梁进行分析时存在的不足。在此基础上,指出下一步工作,展望桥梁抗震分析技术的未来发展方向——基于人工智能的桥梁抗震研究,并倡导学科之间的相互融合及专业人员之间的相互交流。In order to explore the basic thinking of machine learning method in bridge seismic resistance,the development and actuality of bridge seismic analysis theory and technology are briefly reviewed,and the application field of machine learning is summarized,especially in the field of civil engineering.The concept of machine learning is introduced,its key factors and current software platforms are summed up,and the common methods and representative algorithms of machine learning are illustrated by simple examples.First, it is summarized that the highway bridges with medium and small spans in China mainly use the types of simply supported girder bridge and continuous girder bridge,meanwhile,the earthquake damage phenomenon of these bridges in the Wenehuan earthquake are counted,including the damage level division of piers, bearings,shear keys and abutments according to earthquake damage phenomenon.Second,a series of seismic performance tests by international and domestic academics for bearings,shear keys,piers and abutments are summarized,which are used to sum up the constitutive relationships for seismic analysis,and then the seismic design factors of bridge components (including foundation)are obtained.Finally,the overall analysis methodology based on machine learning to assist seismic design and evaluation of such bridges is introduced,which explained that the machine learning for the bridge seismic resistance tasks mainly has 2 aspects:the first is to collect the considerable amount of bridge design data and set up the dataset;the second is data mining,including processing of raw data,debugging or developing a reasonable machine learning algorithm model.The shortcomings of the existing performance-based probabilistic seismic design and evaluation method for analyzing Chinese bridges are discussed briefly.On this basis,the work of the next step is pointed out,and the major concerns in bridge seismic analysis technology in the future are discussed, i.e.,research on bridge,seismic resistance based on artificial intel

关 键 词：桥梁工程 抗震设计评价 综述 中小跨径梁桥 机器学习 

分 类 号：U442.5[建筑科学—桥梁与隧道工程]