线性离散系统间歇故障的鲁棒检测方法

Robust Detection of Intermittent Faults of Linear Discrete-Time Stochastic Systems

针对一类存在未知扰动、过程噪声和测量噪声的离散线性随机动态系统,研究了其鲁棒间歇故障检测问题.设计了降维未知输入观测器的输出,通过引入滑动时间窗口构造了截断残差信号,使得该残差信号对未知扰动解耦而对间歇故障敏感;针对间歇故障的发生时刻和消失时刻,分别提出了2个假设检验,在给定误报率和漏报率的情况下,分别定量确定了间歇故障发生时刻和消失时刻的检测阈值以满足检测性能要求,并在概率框架下分析了间歇故障的可检测性.同时,在某卫星模型上对所提出的方法进行了仿真验证.结果表明,所提出的方法能够快速、准确地检测出间歇故障的发生时刻和消失时刻.

In this paper, a robust detection method for intermittent faults （IFs） was proposed for a class of linear discrete-time stochastic systems, subject to unknown disturbances, process noises and measurement noises. A reduced-order unknown input observer was designed and a novel truncated residual was con- structed by introducing a sliding time window, which is decoupled from the unknown disturbances, but sensitive to the IF. Two hypothesis tests were put forward to detect all the appearing time and the disap- pearing time of the IF, respectively. In addition, for the given false alarm rate and the missing detection rate, the detection thresholds for the appearing time and the disappearing time of the IF were rigorously designed for desired performances, respectively, and the detectability of the IF was also analyzed in a prob- abilistic sense. The simulation results on a satellite model were presented to verify the proposed method. The results illustrate that both the appearing time and the disappearing time of the IF can be detected fast and validly in spite of unknown disturbances.