运用BP-AdaBoost模型识别随机车载作用下大跨斜拉桥拉索损伤  

Damage Identification of Cables of Long-span Cable-stayed Bridges Using BP-AdaBoost Model under Random Vehicle Load

作　　者：谭冬梅[1] 谢华[1] 陈杰[1] 瞿伟廉[1] 查大奎 TAN Dong-mei;XIE Hua;CHEN Jie;QU Wei-lian;ZHA Da-kui(Hubei Key Lab of Roadway Bridge and Structure Engineering, Wuhan University of Technology,Wuhan 430070, China)

机构地区：[1]武汉理工大学道路桥梁与结构工程湖北省重点实验室,武汉430070

出　　处：《噪声与振动控制》2017年第2期163-167,共5页Noise and Vibration Control

基　　金：国家自然科学基金资助项目(51408452);湖北省重点实验室开放基金资助项目(DQJJ201509)

摘　　要：为了有效地进行大跨结构的损伤识别,提出随机车载作用下利用BP-AdaBoost(Back Propagation neural network,Adaptive Boosting)模型对大跨斜拉桥拉索进行损伤识别的方法。该方法首先依据交通调查数据,建立随机交通荷载模型,再运用提升框架,对结构损伤前后的振动测试信号进行提升小波包分解,将小波包信号分量能量累积变异值作为特征值,识别斜拉索损伤位置,然后以此建立BP-AdaBoost模型,利用AdaBoost算法和BP神经网络相结合的方法对大跨斜拉桥拉索的损伤程度进行识别,并研究噪声对该算法的影响。数值分析结果表明,该方法有较强的抗噪声干扰能力,在随机车载作用下,运用BP-AdaBoost模型能够有效识别大跨斜拉桥拉索损伤。The cable damage identification method of long-span cable-stayed bridges is proposed using BP-AdaBoost(Back Propagation neural network,Adaptive Boosting)model under random vehicle’s load.Firstly,the random traffic loadmodel is established according to the traffic survey data,and the vibration signal is decomposed using lifting wavelet packet(WP)analysis based on lifting scheme.Then,the corresponding characteristic vector is constructed by the energyaccumulating variation value of the lifting WP component energy.And this vector is used to identify the damage locationof cables of the cable-stayed bridge.Finally,the BP-AdaBoost model is established.Combining AdaBoost algorithm with BPneural network,the damage degree of the cable of the long-span cable-stayed bridge can be identified.The effect of noise onthe algorithm is studied.The numerical results show that the proposed method can be used to effectively identify the cabledamage of long-span cable-stayed bridges under random vehicle’s load.

关 键 词：振动与波 随机车载 BP-AdaBoost 损伤识别 拉索 提升小波包 

分 类 号：U441.3[建筑科学—桥梁与隧道工程]