基于改进强跟踪滤波的广义系统传感器故障诊断及隔离  

Sensor fault diagnosis and isolation for descriptor systems based on modified strong tracking filtering algorithm

作　　者：梁天添 王茂[1] LIANG Tiantian, WANG Mao(Center for Control Theory and Guidance Technology, Harbin Institute of Technology, Harbin 150001, China)

机构地区：[1]哈尔滨工业大学空间控制与惯性技术研究中心,哈尔滨150001

出　　处：《中国惯性技术学报》2018年第4期554-560,共7页Journal of Chinese Inertial Technology

基　　金：国家自然科学基金(61174037);国家自然科学基金创新群体项目(61021002)

摘　　要：在广义系统故障诊断过程中,若系统动态模型中存在不确定性,传统的无迹卡尔曼滤波算法将失去其传感器故障估计精度。为解决该问题,提出一种改进的强跟踪卡尔曼滤波算法以实现广义连续-离散系统的传感器故障诊断及隔离。首先,提出基于多重渐消因子的强跟踪滤波算法以实现动态模型存在不确定性广义连续-离散系统的故障诊断;然后提出一种结合多模型自适应估计的强跟踪卡尔曼滤波(STUKFMMAE)算法以实现传感器故障的有效隔离。最后,针对基于广义连续-离散系统的惯性传感器故障模型提出仿真算例。仿真数据表明,传统无迹卡尔曼滤波对于传感器故障估计误差为0.002左右,而提出的基于多重渐消因子的强跟踪滤波算法对于传感器故障估计误差最大值为未超过4×10^(-4),且STUKFMMAE相较于UKFMMAE算法具有更好的隔离效果。仿真结果验证了设计方案的有效性。In view that the conventional unscented Kalman filter（UKF） may lose its accuracy in sensor fault estimation if there exist uncertainties in the dynamic model of the descriptor system, a MSTUKF（modified strong tracking unscented Kalman filter） algorithm is proposed for sensor fault diagnosis and isolation for the sampled-data descriptor model. First, a strong tracking UKF（STUKF） algorithm based on the multiple fading factors is proposed to realize the sensor fault diagnosis for the sampled-data descriptor systems with uncertainties. Then, a sensor fault isolation method combining the STUKF algorithm with the multiple-model adaptive estimation（MMAE） is proposed. Finally, simulation examples for inertial sensor fault models based on the sampled-data descriptor systems are proposed. Simulation results show that the estimation errors of sensor fault by conventional UKF algorithm and by the proposed STUKF algorithm are 0.002 and 4×10^（-4）, respectively, and the STUKFMMAE has better fault isolation effect than the UKFMMAE algorithm. The simulation results verify the effectiveness of the proposed method.

关 键 词：广义系统 连续-离散系统 故障诊断及隔离 多模型自适应估计 强跟踪卡尔曼滤波 

分 类 号：V448.2[航空宇航科学与技术—飞行器设计]