曲率模态和神经网络在损伤识别中的应用  

Application of Curvature Mode and Neural Network for Damage Identification

作　　者：杜永峰[1] 张冬兵[1] 

机构地区：[1]兰州理工大学防震减灾研究所,甘肃兰州730050

出　　处：《公路交通科技》2007年第11期77-80,共4页Journal of Highway and Transportation Research and Development

基　　金：甘肃省建设科技攻关资助项目(JK2002-7)

摘　　要：以连续梁为例,通过曲率模态及改进后的模态置信准则联合确定损伤位置,并使用神经网络判定各损伤处的损伤程度。数值模拟考虑了梁不同位置及不同程度损伤情况,研究了运用曲率模态和模态置信准则进行损伤和损伤位置的识别,分析了各自方法的优缺点,通过两种方法的联合运用,使损伤位置识别准确、合理,并且说明了1阶曲率模态的局部参数是损伤程度识别较优的网络输入量,最后运用神经网络识别损伤程度。Using continuous beam as the object, damage can be located through the analysis of curvature mode and improved modal assurance criterion, and then damage degree of all damage locations can be determined through neural network. Considering different locations and degrees in beam by numerical computation, virtues and shortcomings of curvature mode and improved modal assurance criterion can be gotten. Damage location can be detected more accurately and ~nably by using both two methods. It proves that first curvature mode is better local parameter as net-input vector. Structural damage degree is figured out precisely from the application of neural network at last.

关 键 词：桥梁工程 损伤识别 曲率模态 模态置信准则 BP网络 

分 类 号：U441.4[建筑科学—桥梁与隧道工程]