摘要
采用小波分析对获得的结构动力响应进行小波分解,根据各种响应信号对损伤的灵敏度选择损伤特征,从而识别结构出现损伤的时刻,以实现对其监控;分别对结构第一层位移响应信号和加速度响应信号做小波包分解得到各频段能量的特征向量,并分别作为特征参数输入到BP神经网络中实现损伤识别;比较了位移响应信号和加速度响应信号对损伤识别的灵敏性.模拟算例表明,小波分析和BP神经网络联合运用能准确地诊断结构损伤时刻、损伤位置和程度,具有一定的可行性.
The structure's dynamic response available is wavelet-decomposed with the wavelet analysis method, and the damage character is picked up by means of comparison to each other of the damage susceptibility of every response signal, so that the emerging moment (instant) of the damage is identified and the real-time monitoring of structure damage is realized. Further, by using wavelet packet decomposition of the displacement and acceleration respouse on the first story of the structure, the characteristic vector of the energy over all frequency band is obtained a'nd used as an input variable into BP neural network to identify the danage. The susceptibility of both the response signals of displacement and acceleration to the damage identification is compared to each other. It is shown by an illustrative simulation that coupled utilization of wavelet analysis and BP neural network can accurately make a diagnosis of the emerging instant, location, and degree of structure damage with certain feasibility.
出处
《兰州理工大学学报》
CAS
北大核心
2005年第5期121-124,共4页
Journal of Lanzhou University of Technology
基金
甘肃省建设科技攻关项目(JK2002-7)
关键词
损伤监测
小波分析
神经网络
结构健康监测
damage monitoring
wavelet analysis
neural network
structural “health” monitoring