桥梁极值应力的改进高斯混合粒子滤波器动态预测

Improved Gaussian Mixed Particle Filter Dynamic Prediction of Bridge Monitored Extreme Stress

摘要:

为合理地动态预测在役桥梁的极值应力信息，应用桥梁健康监测(BHM)系统的长期日常监测极值应力数据，建立非线性动态模型，引入扩展卡尔曼滤波器（EKF）与高斯混合粒子滤波器（GMPF）相结合的改进高斯混合粒子滤波器(IGMPF)预测算法，对监测极值应力的一步向前预测分布参数及其状态变量的后验分布参数进行预测分析, 并进行了实例验证.IGMPF不仅可以得到实测极值应力状态的合理重要性函数，还可以解决传统预测方法的短期性和精度不高的问题，为实际BHM系统的动力响应预测提供了理论基础.

关键词:监测极值应力数据; 非线性动态模型; 扩展卡尔曼滤波器; 高斯混合粒子滤波器; 改进高斯混合粒子滤波器

Abstract:

To reasonably and dynamically predict the extreme stress information of inservice bridge, in this paper, the nonlinear dynamic models were built including monitoring equation and state equation with the longterm everyday monitored extreme stress data of bridge health monitoring (BHM) system. Then the improved Gaussian mixed particle filter (IGMPF) prediction algorithm was introduced which was obtained by using extended Kalman filter (EKF) and GMPF. IGMPF can predict onestep forward prediction distribution parameters of monitored extreme stress and the posteriori distribution parameters of extreme stress state variable. Finally, an actual example was provided to illustrate the application and feasibility of the IGMPF algorithm built. The IGMPF prediction algorithm can not only obtain the reasonable importance functions of monitored extreme stress states, but also solve the problems of shortterm prediction and low precision of the traditional prediction methods. It provides a theoretical foundation for dynamic response prediction of the actual BHM.

Key words:monitored extreme stress data; nonlinear dynamic model; extended Kalman filter; Gaussian mixed particle filter; improved Gaussian mixed particle filter