约束UKF初始参数对Bouc-Wen模型参数识别的影响
|
摘要
为获得约束UKF初始参数对模型参数识别的影响规律,针对Bouc-Wen模型给出基于约束UKF在线参数识别方法,通过数值模拟分析初始状态估计均值与协方差、过程噪声协方差、观测噪声协方差等滤波器初始参数对模型参数识别精度与收敛速度的影响,提出相应的参数取值建议。结果表明:在无模型误差的情况下,约束UKF对初始参数的设置具有较好的鲁棒性;适当地增大初始状态估计协方差,减小过程噪声,采用真实系统观测噪声协方差以及减小初始参数值与真实值的偏差,可以有效提高参数识别收敛速度和精度。该研究为基于约束UKF的非线性结构模型在线参数识别方法提供了参考。
This paper is specifically devoted to investigating the law underlying the effects of initial parameters of the constrained UKF on model parameter identification. This targeted investigation consists of developing an online parameter identification method based on constrained UKF for Bouc-Wen model;performing numerical simulation to analyze the influence of initial parameters,such as initial state estimation covariance,process noise covariance,measurement noise covariance and initial state estimate mean,on precision and convergence speed of parameter identification; and thereby producing the suggestion related to initial parameters selection. The results show that the constrained UKF offers a better robustness to initial parameters in the absence of model error; improved convergence speed and precision of parameter identification can be effected by increasing the initial state estimation covariance,reducing process noise covariance,observing noise covariance using the real system and reducing the deviation of initial value from the real value of model parameter. The study may provide a reference for nonlinear model parameter identification based on the constrained UKF method.
This paper is specifically devoted to investigating the law underlying the effects of initial parameters of the constrained UKF on model parameter identification. This targeted investigation consists of developing an online parameter identification method based on constrained UKF for Bouc-Wen model;performing numerical simulation to analyze the influence of initial parameters,such as initial state estimation covariance,process noise covariance,measurement noise covariance and initial state estimate mean,on precision and convergence speed of parameter identification; and thereby producing the suggestion related to initial parameters selection. The results show that the constrained UKF offers a better robustness to initial parameters in the absence of model error; improved convergence speed and precision of parameter identification can be effected by increasing the initial state estimation covariance,reducing process noise covariance,observing noise covariance using the real system and reducing the deviation of initial value from the real value of model parameter. The study may provide a reference for nonlinear model parameter identification based on the constrained UKF method.
引文
[1]郭子雄,杨勇.恢复力模型研究现状及存在问题[J].世界地震工程,2004,20(4):47-51.
[2]WEN Y K.Method for random vibration of hysteretic systems[J].Journal of the Engineering Mechanics Division ASCE,1976,102(2):249-263.
[3]BABER T T,WEN Y K.Random vibration hysteretic,degrading systems[J].Journal of the Engineering Mechanics Division ASCE,1981,107(6):1069-1087.
[4]YANG J N,LIN S,HUANGL H,et al.An adaptive extended Kalman filter for structural damage identification[J].Structural Control and Health Monitoring,2006,13(4):849-867.
[5]JULIER S J,UHLMANN J K.Unscented filtering and nonlinear estimation[J].Proceendings of the IEEE,2004,92(3):401-422.
[6]WU M,SMYTH A.Real-time parameter estimation for degrading and pinching hysteretic models[J].International Journal of NonLinear Mechanics,2008,43(9):822-833.
[7]XIE Z B,FENG J C.Real-time nonlinear structural system identification via iterated unscented Kalman filter[J].Mechanical Systems and Signal Processing,2012,28:309-322.
[8]王涛,吴斌.基于UKF模型更新的混合试验方法[J].振动与冲击,2013,32(5):138-143.
[9]王涛,吴斌.基于约束UKF模型更新的混合试验方法[J].地震工程与工程振动,2013,33(5):100-109.
[10]WAN E,VAN DER MERWE R.Kalman filtering and neural networks[M].Wiley:[s.n.],2001:221-280.
[2]WEN Y K.Method for random vibration of hysteretic systems[J].Journal of the Engineering Mechanics Division ASCE,1976,102(2):249-263.
[3]BABER T T,WEN Y K.Random vibration hysteretic,degrading systems[J].Journal of the Engineering Mechanics Division ASCE,1981,107(6):1069-1087.
[4]YANG J N,LIN S,HUANGL H,et al.An adaptive extended Kalman filter for structural damage identification[J].Structural Control and Health Monitoring,2006,13(4):849-867.
[5]JULIER S J,UHLMANN J K.Unscented filtering and nonlinear estimation[J].Proceendings of the IEEE,2004,92(3):401-422.
[6]WU M,SMYTH A.Real-time parameter estimation for degrading and pinching hysteretic models[J].International Journal of NonLinear Mechanics,2008,43(9):822-833.
[7]XIE Z B,FENG J C.Real-time nonlinear structural system identification via iterated unscented Kalman filter[J].Mechanical Systems and Signal Processing,2012,28:309-322.
[8]王涛,吴斌.基于UKF模型更新的混合试验方法[J].振动与冲击,2013,32(5):138-143.
[9]王涛,吴斌.基于约束UKF模型更新的混合试验方法[J].地震工程与工程振动,2013,33(5):100-109.
[10]WAN E,VAN DER MERWE R.Kalman filtering and neural networks[M].Wiley:[s.n.],2001:221-280.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心