摘要
为了在应用随机子空间识别方法时合理确定模型阶次,提出了一种基于函数拟合的模型定阶方法.首先,对投影矩阵进行奇异值分解,然后利用S型函数对归一化奇异值进行拟合,最后根据拟合函数估算模型阶次.基于弹簧振子系统的仿真算例对不同噪声水平下的模型定阶结果进行研究,并通过斜拉桥的模态分析来验证方法的应用效果.算例结果表明,噪声水平为0%、5%、10%、15%、20%时识别模型阶次是相同的,说明所提方法具有良好的稳定性.根据阶次识别结果得到的模态频率与实桥环境振动试验结果的相对误差在5%以内,从而验证了所提方法的有效性.
A model order determination method based on function fitting was proposed to reasonably determine the model order during the application of the stochastic subspace identification method.Firstly,the singular value decomposition was performed on the projection matrix.Then,the S-type function was adopted to fit the normalized singular value.Finally,the model order was estimated according to the fitting function.The simulation of a spring-vibrator system was used to study the results of the order determination at different noise levels.Modal analysis of a cable-stayed bridge was used to verify the application effect of the method.The simulation results indicate that the recognized model orders at the noise levels of 0%,5%,10%,15%and 20%are the same,showing the good stability of the proposed method.The relative errors between the modal frequencies based on the results of order determination and the ambient vibration test results of the real bridge are less than 5%,proving the effectiveness of the proposed method.
作者
廖聿宸
张坤
宗周红
吴睿
Liao Yuchen;Zhang Kun;Zong Zhouhong;Wu Rui(School of Civil Engineering,Southeast University,Nanjing 211189,China;CCCC Second Highway consultants Co.,Ltd.,Wuhan 430056,China)
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2020年第3期440-446,共7页
Journal of Southeast University：Natural Science Edition
基金
国家自然科学基金资助项目(51678141)
国家重点研发计划资助项目(2017YFC0703405).
关键词
结构健康监测
随机子空间识别
模型定阶
奇异值分解
S型函数
structure health monitoring
stochastic subspace identification
order determination
singular value decomposition
S-type function
