实时工作模态参数数据驱动随机子空间识别

Improved data-driven stochastic subspace identification of online operational modal parameters

基于数据驱动随机子空间识别(SSI)算法,引入Hankel矩阵行块数i和列数j经验确定方法和模态传递范数(Modal Transfer Norm),采用奇异熵进行系统自动定阶方法,使识别过程自动化,实现工程模态参数的实时识别。应用Choleshy分解求解下三角矩阵R,缩减了计算量,特别是连续两个时刻的窗口数据大量重叠时,大大减小了计算时间,从而满足实时识别时间上的要求。三跨连续梁桥数值算例和一座实桥测试数据结果表明,每次识别均在移动步长时间内完成,速度快、精度高,可有效运用于大型结构模态参数的实时识别。

Based on the data-driven stochastic subspace identification (SSI) algorithm, an improved on-line SSI procedure was presented to automatically extract the operational modal parameters of structures. The Hankel matrix’s dimension parameters i and j were determined by experiential approach and the mode transfer norm (MTN) was used to improve the stabilization diagrams. A singular entropy-based approach was implemented to determine the order of system so that the identification procedure becomes more automatic. To reduce the calculation work, the Choleshy factorization technique was used to handle the lower tri-matrix R. As a result, the calculation efficiency is much improved when the data were overlapped for two different time windows. The applicability and efficiency of proposed method are validated via both simulated 3-spans continuous beam and full-size bridge that was tested under field operational conditions. It is demonstrated that the approach has the advantages of less calculation time consumption and higher precision. The proposed method can be used to perform online operational modal parameter identification of large-scale structures.