摘要
对于一类带有模型不确定性、噪声干扰的线性系统,研究了其故障检测中的阈值设计问题。建立了残差与模型不确定性、噪声、控制输入、故障之间的关系；提出了基于信号L2范数与系统H∞范数的鲁棒自适应阈值设计方法；针对其检测率低的问题,进一步基于概率鲁棒的思想,对模型不确定性与残差引入了截尾高斯分布这一“软边界”描述方式；在此基础上所设计的阈值兼顾了检测率和虚警率两方面的要求,考虑了模型不确定性与噪声的影响,能够跟随控制输入信号的变化,具有较好的鲁棒性和自适应性。
The design of threshold for robust fauh detect was studied for a linear system with model uncertainty and noise. The relationship between the residual and model uncertainty, noise, control input and fault was researched, and an adaptive threshold method based on 1.2-norm of signal and H∞ -norm of system was given. To increase the rate of fault detection, chopped Gauss distribution was introduced in this paper for the soft bound description of model uncertainty and residual. After considering the model uncertainty and noise, as well as the FDR and FAR, the threshold was designed and could be driven by control input. As a conclusion, the threshold designed has a better robustness and adaptive ability than former research.
出处
《宇航学报》
EI
CAS
CSCD
北大核心
2006年第1期36-40,共5页
Journal of Astronautics
关键词
故障检测
自适应阈值
H∞范数
截尾高斯分布
Fault detection
An adaptive threshold
H∞ Norm
Chopped gauss distribution
