基于F-R-M方法的桥梁地震风险评估研究  

Seismic risk assessment of bridges based on F-R-M method

作　　者：周敉[1] 封伟[1] 谢功元[2] 张玥[3] 贺拴海[1] 

机构地区：[1]长安大学旧桥检测与加固技术交通行业重点实验室,陕西西安710064 [2]湖北省交通运输厅工程质量监督局,湖北武汉430014 [3]西安科技大学建筑与土木工程学院,陕西西安710054

出　　处：《广西大学学报（自然科学版）》2017年第1期299-308,共10页Journal of Guangxi University（Natural Science Edition）

基　　金：交通运输部应用基础研究项目(2015319812010);陕西省社会发展科技攻关项目(2015SF265);国家山区公路工程技术研究中心开放基金(gsgzj-2012-07);陕西省自然科学基础研究计划项目(2013JQ7031);中央高校基本科研业务费资助项目(3108211511070);陕西省交通科技项目(15-19K)

摘　　要：为解决现有有限元分析方法对求解地震作用下桥梁的失效概率计算量大、耗时长的问题,建立了一种神经网络仿真与概率有限元相结合的桥梁地震风险评估方法(F-R-M法)。首先通过多条地震波对桥梁进行时程分析,并得到相应的桥梁结构的地震响应;然后建立RBF神经网络,并将量化后的地震波能量、场地类别、峰值个数、结构弯矩作为输入参数,对神经网络进行训练和检验,训练成功后可采用Monte Carlo方法生成大量随机数,通过神经网络仿真得到桥梁结构随机地震响应数据,并计算出结构的失效概率;最后以某三跨连续刚构桥为例,采用有限元软件建立了全桥纤维单元模型,选取了24条地震波对其进行了动力增量(IDA)分析,并采用F-R-M法对该桥的地震风险概率进行了评估。实例分析表明,F-R-M法计算效率高、计算误差小,神经网络RBF将地震响应和地震输入之间的非线性映射关系很好地仿真模拟,研究结果可为桥梁抗震设计和地震风险评估提供参考。Present finite element method for failure probability calculation of bridges under the action of earthquake is dissolved and time- consuming. In order to solve this problem,a new method that combines neural network simulation and probabilistic finite element method （ F-R-M ） for seismic risk evaluation of bridges is establ ished. First ly, the multiple seismic waves from bridge history are analyzed, and corresponding seismic response of the bridge is obtained. Then RBF neural network is establ ished. Taking seismic wave en e rg y, numerical field ty p e , seismic wave peak number and pier moment as input parameters, the network of RBF are trained and te s ted. After the success of the training, the Monte Carlo method is adopted to generate a large number of random numbers, random seismic response of the bridge is given by the simulation of neural network and the failure probability of the bridge can be calculated. A three span continuous rigid frame bridge was taken as an example. The numerical model of the bridge was established by using the finite element software. 24 seismic waves were selected to analyze the dynamic increment （ID A ） , and the seismic risk probability of the bridge was evaluated by the F-R-M method. The case analysis shows that the F-R-M method has high efficiency and small calculation error. The RBF neural network is a good simulation for the highly nonlinear relationship between seismic input and seismic r e sp on se, so the method can provide a reference for bridge seismic design and decision making.

关 键 词：桥梁 随机有限元 神经网络 地震 风险评估 

分 类 号：U447[建筑科学—桥梁与隧道工程]