Abstract:

In order to reduce the noise effect in frequency-domain subspace algorithm for mode identification,a new weighted frequency subspace algorithm was put forward herein.The noise variance in frequency domain was estimated with a few experiments.The weight matrix was made up of the weight coefficient from the estimated variance of the input and output.This algorithm considered the noise of the input and output and allocated the weight with variance in frequency domain.The simulation was carried out on a system with two DOF  to validate the effect.The experiments on a variable section were put up.The results from the simulation and experiments show that the improved algorithm can exactly get the modal parameters under the conditions with  strong noise.

Key words: subspace, weight, noise, parameter identification

摘要：

为解决频域子空间模态参数辨识方法受噪声影响的问题,提出一种加权频域子空间方法。运用该方法对多次实验得到的频域数据的方差进行估计,把输入、输出频域点估计的方差倒数作为加权系数,构成加权系数矩阵。考虑了输入及输出噪声,根据噪声的频域方差分配权重。为验证改进方法的有效性,在一个二自由度系统上进行了仿真验证。以某变截面轴的模态分析为例,进行了实验验证。仿真及实验结果表明,提出的改进加权算法能够在较强噪声时准确分辨模态参数。

关键词: 子空间, 加权, 噪声, 参数辨识