新型UKF在非线性系统执行器故障估计中的应用  

Application of new UKF to fault estimation of nonlinear system with actuator faults

作　　者：钱默抒[1,2] 姜斌[1] 杜董生[1] 杨浩[1] 

机构地区：[1]南京航空航天大学自动化学院,南京210016 [2]南京航空航天大学无人机研究院,南京210016

出　　处：《东南大学学报（自然科学版）》2011年第B09期120-124,共5页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61074080);南京航空航天大学博士论文创新与创优基金资助项目(BCXJ10-03);江苏省普通高校研究生科研创新计划资助项目(CXLX11_0199)

摘　　要：针对扩展卡尔曼滤波(extended Kalman filter,EKF)算法使用受限,粒子滤波算法动态跟踪能力差、易产生退化,单一无先导扩展卡尔曼滤波(unscented Kalman filter,UKF)算法滤波精度低等缺陷,根据极大后验(maximum a posterior,MAP)估计原理,设计了一种带限定记忆法的无先导卡尔曼故障估计滤波器.首先根据极大后验估计原理,推导出一种最优MAP-UKF乘性故障估计器;在此基础之上,又推导出次优的执行器故障估计滤波器,并对常值故障估计滤波器进行了无偏性证明;接着,将限定记忆法应用于MAP-UKF故障估计滤波器中.最后,将新型的UKF算法应用于倒立摆仿真,相比于EKF和单一UKF算法,所提出的故障估计算法收敛速度较快,滤波精度显著提高.Aiming at the constraints of the extended kalman filter （EKF）, the poor dynamic txacking and degradation ability of the particle filter algorithm, and the low precision of the single unscented kalman filter （UKF）, an unscented kalman fault estimation filter with limited memory method was designed based on maximum a posterior （MAP） estimation. Firstly, an optimal MAP-UKF multiplicative fault estimator is derived according to MAP estimation. Then, the sub-optimal actuator fault estimation filter algorithm is derived. Meanwhile, the unbiasedness of fault estimation filter is proved in this paper. Afterwards, the limited memory method is applied to the MAP-UKF fault estimation filter design. Finally, simulation results about the inverted pendulum show the better filtering performance of the designed MAP-UKF compared with the EKF and single UKF algorithms.

关 键 词：极大后验 扩展卡尔曼滤波 无先导卡尔曼滤波 故障估计 非线性系统 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]