摘要
用神经网络BP算法直接对所测结构频响函数进行分析,用来识别结构的状态。给出了一个tansig/purelin应用的例子,比较了两种网络,发现newff比newcf能够获得更好的识别效果。结果也表明,与基于特征值分析的方法相比,数据处理的方法更加简单,可以进行状态的分类识别。
This paper presents a Back-Propagation network method(BP network) applying structure frequency response function as the network inputs.The method is proposed to identify a structure's state information.A example is given in detail,in which a two-layer tansig/purelin network is used.Two kinds of networks are compared,the newff network tends to give a more reasonable answer than the newcf network.And the results show that the method is more simple,correct and effective than other methods such as the eigenvalue analysis.
出处
《振动.测试与诊断》
EI
CSCD
2007年第1期45-47,共3页
Journal of Vibration,Measurement & Diagnosis
基金
国防科技预先研究项目
关键词
频响函数
神经网络
BP网络
结构健康监测
frequency response function neural network BP network structure health monitoring