数据驱动与协方差驱动随机子空间法差异化分析  

Performance comparison for data-driven and covariance-driven stochastic subspace identification method

作　　者：辛峻峰 王树青[1] 刘福顺[1] 

机构地区：[1]中国海洋大学工程学院,山东青岛266100

出　　处：《振动与冲击》2013年第9期1-4,20,共5页Journal of Vibration and Shock

基　　金：国家自然科学基金项目(51079134;50909088;51279188;51009124);国家自然基金重大国际合作项目(51010009)

摘　　要：针对能有效从环境激励结构振动响应中获取模态参数的随机子空间法,传统观点认为无论在理论上或应用中数据驱动随机子空间法与协方差驱动随机子空间法在模态参数识别过程中表现一致,实际应用中表现不一致问题,理论上探讨两种方法出现差异的原因,并进行相应的数值模拟。研究结果表明:基于QR分解的数据驱动随机子空间法无论计算精度或对较弱势模态的识别能力均明显优于协方差驱动随机子空间法。The stochastic subspace method is a linear system identification method developed in recent years, by which modal parameters can be effectively estimated from the response of structure under ambient excitation. The data-driven and covariance-driven stochastic subspace identification methods traditionally were thought to be consistent with each other both in theory and in application for modal identification. However, there exists in practice the difference between the two methods. Therefore, the reasons of the performance difference were analyzed and numerically studied. The results demonstrate that the data-driven stochastic subspace identification method outperforms the covariance driven subspace identification method not only on the accuracy of identified parameters but also on the capacity of identifying weaker modes.

关 键 词：模态识别 随机子空间法 数据驱动 协方差驱动 

分 类 号：O32[理学—一般力学与力学基础]