摘要
改进了一种用于模态参数辨识的截断窗逼近幂迭代子空间方法(Truncated Window Approximated Power Iteration,TW-API)。该算法采用输入输出信号的截断窗互相关函数建立信号主子空间的递推关系,应用子空间投影来进行辨识。跟传统的TW-API方法相比,改进后的TW-API方法基于信号子空间理论,简化了数据矩阵递推的过程,显著地减少了算法的整体计算量和计算时间,同时仍能较好地保持辨识精度。在数值算例中,选用二连杆机械臂和在轨航天器模型这两个算例,分别进行了时变模态参数的辨识,并将结果与经典的投影估计子空间跟踪方法(Projection Approximation Subspace Tracking,PAST)和逼近幂迭代(Approximated Power Iteration,API)递推方法进行比较,4种辨识方法的仿真结果表明本文提出的新算法能有效辨识系统的时变模态参数。
An improved method based on Truncated Window Approximated Power Iteration（TW-API）Tracking for modal parameter identification is proposed in this paper.The truncation window cross-correlation function of input and output signals is adopted to establish the recursion relation of signal dominant subspace,and subspace projection is then used to identify the modal parameters.Compared with the traditional TW-API method,the improved one,which is based on the theory of signal subspace,simplifies the recursion procedure of data matrix,significantly reduces the computing cost and computation time,and meanwhile guarantees the calculation accuracy.In numerical simulation,a structure of two-link robotic manipulator and a model of on-orbit spacecraft are used to identify the time-varying modal parameters.Moreover,the computation results of these two methods are also compared with the classical Projection Approximation Subspace Tracking（PAST）and Approximated Power Iteration（API）recursive methods.The results of the four identification methodologies demonstrate that the proposed new algorithm in this paper can effectively identify the time-varying modal parameters of the system.
出处
《振动工程学报》
EI
CSCD
北大核心
2015年第5期721-729,共9页
Journal of Vibration Engineering
基金
国家自然科学基金资助项目(11072044
11372056
11002032)
高等学校博士学科点专项科研基金资助项目(20110041130001)
关键词
模态参数辨识
时变系统
递推子空间方法
航天器
modal parameter identification
time-varying system
recursive subspace method
spacecraft
