大跨度空间网格结构的健康监测系统  

Health monitoring system with large-span space-netted structure

作　　者：徐菁[1] 曲丽敏[1] 卢翠萍[1] 

机构地区：[1]青岛理工大学土木工程学院,山东青岛266033

出　　处：《兰州理工大学学报》2016年第4期128-133,共6页Journal of Lanzhou University of Technology

基　　金：国家自然科学基金(51068091);山东省自然科学基金(ZR2013EEL013);山东省高等学校科技计划项目(J12LG09)

摘　　要：针对大跨度空间网格结构健康监测系统中的传感器布置以及损伤识别问题,以凯威特型单层球面网壳为例,首先,定义基于变形能的适应度函数,采用粒子群优化算法布置加速度传感器;其次,根据加速度响应信号,建立时间序列模型(AR模型),根据模型系数的变化,判断损伤存在与否;最后,以AR模型系数为输入,损伤位置为输出,建立BP神经网络,通过BP网络的训练和测试,判定损伤存在的位置.数值模拟结果表明:基于粒子群算法的传感器优化布置方法能够准确获取网壳结构中的关键信息点,有效节省传感器布置数目;基于时间序列分析和神经网络的损伤识别方法可以准确识别网壳结构的损伤及损伤位置.Aimed at the problem of sensor placement and system damage identification in health monito- ring system of large-span space-netted structure, and taking a Kiewitt single-layer spheric reticulated shell as example, a deformation energy-based fitness function was defined and the acceleration sensors were disposed based on particle swarm optimization algorithm, first. Then, based on the acceleration response signal, a time-sequence model （AR model） was established and, according to the change of the AR model coefficient, it was determined whether the damage occurred or not. Finally, taking the AR model coefficient as input and the position of the damage as output, the BP neural network was established. Through the BP network training and testing, the positions of the damage were determined. Numeric simulation result shows that the key information points in reticulated shell structure can be accurately obtained by means of optimal disposition of sensors with particle swarm algorithm and the number of sensor will be saved effectively, so that the damage and its position of reticulated shell structure can be accurately identified based on time-series analysis and damage identification method of BP neural network.

关 键 词：健康监测 粒子群算法 时间序列分析 神经网络 损伤识别 

分 类 号：TU311.4[建筑科学—结构工程] TU392.3