摘要
针对超载超速车辆的长期非正常运营引起桥梁损伤破坏现象,考虑桥面不平度的影响,通过车桥耦合计算,应用神经网络方法对通过斜拉桥的重车荷载识别进行了初步探讨。首先基于相似公路上实测车流统计模型拟定车流荷载,通过车桥耦合响应分析,运用数值模拟数据建立网络的训练和精度检验样本,并加入5%的随机噪声,通过检验组(车重和车速)数据识别结果的误差分布来评价网络训练精度,车重与车速输出误差在5%以内,表明网络训练精度较高。然后,依据斜拉桥健康监测系统,采用斜拉桥实际运营状态下的监测响应(索力和应变)提取车辆荷载作用特征参数并组成网络的输入向量,运用满足精度要求的神经网络来识别车重和车速结果较好。其中,车速分布近似均值65 km/h的正态分布,车重大多分布在10~70 t的范围内,10 t以下小型车辆识别误差较高。车速车重总体呈现负相关,分布具有一定离散性,以上特征与实际情况基本吻合。表明将BP神经网络法与斜拉桥监测系统相结合的方式进行重车车速车重识别是可行的。
Aiming at the phenomenon of bridge injury and damage caused by long-term abnormal operation of overloaded and over speed vehicles and considering the influence of bridge surface irregularity, the identification of the load of heavy-duty vehicles passing through cable-stayed bridge is preliminary discussed by vehicle-bridge coupled calculation and the method of neural network. First, the traffic load is worked out based on the actual measured traffic flow statistics model on the similar highway. The network training and accuracy test samples of the network are established by numerical simulation data and vehicle-bridge coupling response analysis. Adding 5% random noise, the network training accuracy is evaluated by error distribution of identified data from examination group(vehicle weight and vehicle speed). The vehicle weight and speed output error is within 5%, indicating that the network training accuracy is higher. Then, based on the cable-stayed bridge health monitoring system, the monitoring response(cable force and strain) of the cable-stayed bridge under actual operation status is used to extract the characteristic parameters of vehicle load action and to form the input vector of the network. The results of identifying vehicle weight and speed by using neural network satisfying the accuracy are good. The vehicle speed distribution approximates a normal distribution with an average value of 65 km/h. Most of the vehicles weights are mostly distributed in the range of 10-70 t, and the identification error of small vehicles below 10 t is higher. The vehicle speed and weight are negatively correlated, and the distribution is discrete. These characteristics are basically consistent with the actual situation, indicating that the combination method of BP neural network and the cable-stayed bridge monitoring system is feasible to identify the speed and weight of heavy-duty vehicle.
作者
陶兴旺
孙宗光
陈一飞
TAO Xing-wang;SUN Zong-guang;CHEN Yi-fei(School of Transportation Engineering,Dalian Maritime University,Dalian Liaoning 116026,China)
出处
《公路交通科技》
CAS
CSCD
北大核心
2019年第12期87-93,共7页
Journal of Highway and Transportation Research and Development
基金
国家自然科学基金项目(51178070)
中央高校基本科研业务费专项资金项目(3132014326,3132016216)
关键词
桥梁工程
车辆荷载识别
神经网络
斜拉桥
监测响应
bride engineering
vehicle load identification
neural network
cable-stayed bridge
monitoring response
