摘要
针对正交异性桥面板钢箱梁桥,提出了一种基于机器视觉和BP神经网络的单车荷载识别方法.首先,基于机器视觉目标检测和图像处理技术,实现桥上行驶车辆时空参数的识别,包括车辆的横向位置、车速、车轴数和轴距等.然后,结合车辆时空参数和U肋纵桥向应变响应信号,建立BP神经网络模型来识别车辆轴重和总重.通过数值模拟验证了算法的有效性和抗噪性.最后,设计了一个钢箱梁桥模型试验进一步验证了提出的理论.试验结果表明:该方法的车辆横向位置、车速和轴距识别的最大误差分别为1.29%,-2.31%和1.79%;车辆模型总重识别最大误差为2.65%,轴重识别最大误差为4.81%,方法具有较好的精度和抗噪性能.
Aim at the orthotropic steel box beam bridge,a single vehicle load identification method based on machine vision and back propagation(BP)neural network was proposed.Firstly,machine vision target detection and image processing technology were used to identify spatiotemporal parameters when vehicle traveling on the bridge,which includes thetransverse position of the vehicle,vehicle speed,number of axles,and wheelbase.Combining with the spatiotemporal parameters and the longitudinal strain response of the U-rib,a BP neural network model was built to identify the axle and gross weight of the vehicle.The effectiveness and antinoise performanceof the algorithm were verified by numerical simulation.Finally,a model test of a steel box girder bridge was carried out to further verify the proposed method.The test results show that the maximum errors of vehicle lateral position,vehicle speed,and wheelbase recognition of this method are 1.29%,-2.31%,and 1.79% respectively,and that of gross weight and axle load are 2.65% and 4.81%respectively,which proves the proposed method has better accuracy and anti-noise performance.
作者
王超
杨青祥
齐天玉
伍贤智
WANG Chao;YANG Qingxiang;QI Tianyu;WU Xianzhi(School of Civil,Architecture and Environment,Hubei University of Technology,Wuhan 430068,China;China Railway Bridge Science Research Institute Co.Ltd.,Wuhan 430034,China)
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2023年第8期53-59,共7页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(51408250)
湖北工业大学研究生创新人才培养项目(校2022054)。
关键词
荷载识别
机器视觉
BP神经网络
时空信息
正交异性桥面板
load identification
machine vision
BP neural networks
spatiotemporal information
orthotropic deck
