随机子空间识别方法计算效率的改进

参数识别是结构健康监测领域研究中的重点。随机子空间法是近年来发展起来的一种线性系统辩识方法,可以有效地从环境激励的结构响应中获取模态参数。该方法的基本原理是将“将来”数据向“过去”数据进行垂直投影,进而根据该投影计算出可观测矩阵和一个Kalman滤波状态序列。而Kalman滤波序列是“过去”输出信号的线性组合,即“过去”输出和“将来”状态估计建立了关系。而从Hankel矩阵的组成来看,由于要使得该矩阵具有单无限的条件,所需的计算时间也比较长。据此对随机子空间方法进行了改进。改进的基本思想是采用一个测点的信号作为“过去”作为输出信号代替全部测点的信号。这样就减少了计算量。最后用一数值模拟算例进行了验证,结果良好。

Improvement in stochastic subspace identification

Parameter identification is currently one of the main research topics in the area of structural health monitoring.Stochastic subspace identification is a novel approach developed recent years.It can identify the modal parameters of linear structure from ambient vibration of structure.The key issue of the method is projecting "past" data onto "future" data.Then an observability matrix and the Kalman filter state sequence are obtained.Kalman filter state sequence is a linear composition of "past" output data.A correlation is set up between "past" output and "future" state.Because the Hankel matrix is a semi-infinity matrix,the workload is weighty.For this reason,improving stochastic subspace identification is presented.The key part is a replacement of all "past" by the data of a test point.This will reduce the workload of computation.The method is evaluated by a numerical simulation example.