摘要
针对虚假模态影响特征系统实现算法识别结果的问题,提出用奇异值分解结合模态能量水平来剔除特征系统实现算法识别结果中的虚假模态。利用奇异值分解(SVD)方法滤除信号中的部分噪声,减少噪声模态并提高识别结果精度,利用输出矩阵、状态矩阵的特征值和特征向量以及输入分配矩阵计算出识别结果中各阶模态能量矩阵,对其进行奇异值分解得到最大奇异值,将其作为各阶模态对输出能量贡献的衡量指标,称之为模态能量水平,然后由计算模态与噪声模态能量为零的特点剔除识别结果中的虚假模态。通过数值仿真和实例分析验证了方法的有效性。
Due to the influence of spurious modes on the eigensystem realization algorithm results,singular value decomposition（SVD） and model energy level are introduced to remove the spurious modes of eigensystem realization algorithm,reduce part of the noise modes and improve the accuracy by reducing measurement noise by SVD.The energy matrix of each mode can be calculated by the selection matrices,the eigenvalues and eigenvectors of the state matrix and the input distribution matrix.The largest singular value of the energy matrix obtained by SVD is a measure for the energy contribution of each mode,which is named mode energy level.Spurious modes resulting from noise or model redundancy are indicated according their mode energy level.A numerical example and an experimental example are presented to demonstrate the efficacy of the method.
出处
《重庆大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2012年第3期20-25,共6页
Journal of Chongqing University
基金
中央高校基本科研业务费资助项目(CDJZR10118801)
关键词
参数识别
奇异值分解
特征系统实现算法
虚假模态
稳定图
parameter identification
singular value decomposition（SVD）
the eigensystem realization algorithm
spurious modes
stabilization charts
