摘要
提出了基于神经网络技术和受控结构振型数据的损伤定位分步识别方法。首先采用状态反馈控制的方法,有目的地对结构进行极点配置,得到设置的受控结构的动力特性数据。将结构划分为若干个子区域,以受控结构损伤前后振型差作为输入参数,利用概率神经网络确定结构损伤所在的子区域,然后用多损伤定位确信准则对结构损伤子区域中的具体损伤位置进行识别。数值仿真表明,利用概率神经网络能有效地确定结构损伤子区域,采用分步识别的策略能大大缩小具体损伤单元的识别范围,而利用受控结构的动力特性参数可提高识别指标对损伤的敏感度,进而提高损伤识别的准确性。
The paper proposes a two-step structural damage localization method based on a neural network and the mode shapes of structures under control. In the study, the state feedback control method is first employed to place the poles of the structure intentionally and the prescribed characteristic frequencies and mode shapes of the controlled structure are obtained accordingly. The structure is divided into several sub-domains and damaged sub-domain is identified using a probabilistic neural network, in which the mode differences between undamaged and damaged controlled structures are chosen as the input vectors. Then the multiple damage location assurance criterion is taken as the damage indicator to locate the specific damaged element. Numerical results show that the damaged sub-domain can be identified using the probabilistic neural network and the search domain for locating the specific damaged element is greatly reduced by the adoption of the proposed two-step identification strategy. The use of dynamic characteristic data of the structure under control can improve the sensitivity of the damage indicator and consequently the accuracy of damage identification.
出处
《工程力学》
EI
CSCD
北大核心
2006年第6期54-59,共6页
Engineering Mechanics
基金
教育部出国留学回国人员科研启动基金项目资助(教外司留[2001]498号)
国家自然科学基金项目资助(50479050)
关键词
工程结构
结构损伤
位置识别
概率神经网络
受控结构
动力特性
engineering structures
structural damage
localization
probabilistic neural network
controlled structure
dynamic characteristics
