摘要
提出一种基于奇异值分解的unscented卡尔曼滤波(SVD-UKF)非线性滞回结构系统识别方法。SVD-UKF可被看成改进的unscented卡尔曼滤波(UKF)方法,相对UKF而言,SVD-UKF具有更好的鲁棒性和灵活性。此方法不仅避免象扩展卡尔曼滤波(EKF)为了计算Jacobians矩阵的所需的导数运算,并且可以克服常规UKF方法在计算协方差时经常遇到的病态条件的缺点。对非线性系统参数的识别和突然变化的识别的数值模拟结果显示了所提出方法的鲁棒性和灵活性。
A singular value decomposition based on derivative-free Kalman filter (SVD-UKF) method for the identification of nonlinear hysteretic structural systems is put forward, which considered as a improved algorithm of unscented Kalman filter (UKF), with improvements in robustness and flexibility over the UKF techniques. This method enables to avoid the derivation of Jacobians for Extended Kalman filter (EKF), and overcome the drawbacks of the UKF method often encountering ill-conditioned problems in covariance calculation. The simulation results for identifying the parameters and the abrupt changes of a nonlinear system demonstrate the robustness and flexibility of the proposed methodology.
出处
《应用力学学报》
EI
CAS
CSCD
北大核心
2008年第1期57-61,181,共5页
Chinese Journal of Applied Mechanics
基金
国家“十一五”科技支撑计划项目(2006BAJ13B03-4)
