摘要
针对美国IASC-ASCE的结构健康监测科研组提出的基准结构进行结构自振频率识别研究.神经网络训练时使用的数据为有限元程序计算所得出,将有损伤结构在环境激励下某点的加速度响应,通过快速傅立叶变换得到的离散频率响应函数作为神经网络的输入;将损伤结构的自振频率作为神经网络的输出.通过对在不同噪声水平下训练的神经网络的识别结果进行分析比较,结果表明:应用人工神经网络进行结构自振频率识别是切实可行的.
This paper identified natural frequencies using artificial neural network. Theory of artificial neural network and a method of natural frequencies identification were introduced firstly. Then the numerical simulation for identification natural frequencies of benchmark structure that was set up by the IASC-ASCE Task Group on Health Monitoring was carried out. In the simulation, the training patterns and the testing patterns were computed by the FEA program. Discrete frequency response function was input of neural network and the natural frequencies were output of neural network. Finally, the testing results under several noise levels were compared. The result of study show that it is feasible able to identify the natural frequencies using artificial neural network.
出处
《辽宁工程技术大学学报（自然科学版）》
CAS
北大核心
2017年第2期186-190,共5页
Journal of Liaoning Technical University (Natural Science)
基金
辽宁省教育厅杰出青年学者成长计划项目(LJQ2013037)
关键词
基准结构
自振频率
人工神经网络
频率响应函数
噪声
benchmark structure
natural frequency
artificial neural network
frequency response function
noise
