Abstract: In this paper, a design framework of intermittent fault (IF) diagnosis and fault tolerant control (FTC) is developed for a class of linear discrete-time systems disturbed by noise. Firstly, a two-layer Kalman filter scheme is employed to realize IF isolation and estimation accurately. The first layer Kalman filter can detect and isolate the actuator intermittent fault by constructing the weighted square sum of the filter error as the system residual. The second optimal Kalman filter achieves a joint optimal estimation of system state, fault-free constraint state and intermittent fault by reconstructing the filter gain into a fault-free constrained gain and an intermittent fault gain, and solving the problem of the minimum variance unbiased estimation under the intermittent fault gain constraint. Furthermore, the optimal estimation, linear quadratic Gaussian (LQG) control and fault decoupling are used to construct the active fault-tolerant controller, so that the system can still guarantee the system performance in case of intermittent fault. Finally, the effectiveness of the proposed method is verified by the simulation of DC motor control system.

Key words: Intermittent faults, optimal fault estimation, two-fold Kalman filter, LQG, fault tolerant control 

摘要： 针对一类发生间歇故障且受噪声干扰的线性离散系统，发展了一种故障诊断与主动容错的设计方案。首先，为了及时检测并准确诊断间歇故障，设计了双层Kalman滤波器，第一层Kalman滤波器通过构造滤波误差加权平方和作为系统残差，实现对执行器间歇故障的检测与隔离，第二层最优Kalman滤波器通过将滤波器增益重构为无故障约束增益与间歇故障增益，并求解间歇故障增益约束下的最小方差无偏估计问题，实现了系统状态、无故障约束状态和间歇性故障的联合最优估计。进一步地，利用最优估计、线性二次型高斯(LQG)控制和故障解耦方法构造主动容错控制器，使系统在发生间歇故障时仍能保障系统性能。最后，通过直流电机控制系统的仿真验证了所提方法的有效性。

关键词: 间歇故障, 最优故障估计, 双层Kalman滤波, 线性二次型高斯(LQG), 容错控制