基于FFT和神经网络的复模态参数识别  

Identification of complex modal parameters based on FFT and neural network

作　　者：刘海源[1] 陈建功[1] 张永兴[1] 

机构地区：[1]重庆大学土木工程学院,重庆400045

出　　处：《东南大学学报（自然科学版）》2009年第S2期217-221,共5页Journal of Southeast University：Natural Science Edition

基　　金：国家杰出青年科学基金资助项目(50625824);国家自然科学基金资助项目(50679097)

摘　　要：为了精确识别结构复模态参数,提出了一种基于快速傅里叶变换(FFT)和人工神经网络的模态识别方法.该方法首先对自由振动信号进行FFT预处理,得到粗略的各阶模态频率和相位.然后,根据模态的阶数设定神经元的个数,根据预处理后得到的频率和相位设定神经网络权值和基函数参数迭代的初始值.最后,通过对人工神经网络进行训练,达到利用自由振动信号进行时域模态识别的目的.仿真结果表明,该算法可消除频率法识别中因频谱泄露与噪声等产生的误差,提高模态识别的精度,因而是一种有效的时域识别方法.To precisely identify complex modal parameters of structural systems,a model identification algorithm is presented based on the fast Fourier transform(FFT) and the artificial neural network model.First,the free vibration signal is preprocessed by the FFT algorithm.The frequency and the phase of all modals are obtained.Secondly,the number of neural nodes is determined by the orders of the modals.The initial weights of neural network and the iterative initial parameters of the base function are assigned according to the frequency and the phase obtained by preprocessing.Finally,by training artificial neural network,the time-domain modal identification can be realized by using the free vibration signal.The simulation results indicate that this algorithm can eliminate the errors induced by frequency spectrum leakage and ground noise in frequency-domain methods and the accuracy of modal identification is improved.Therefore,it is an effective time-domain identification algorithm.

关 键 词：复模态 参数识别 快速傅里叶变换 神经网络 参数可调的基函数 

分 类 号：TU317[建筑科学—结构工程]