摘要
提出一种基于残差评价策略的故障检测方法,在仅有系统残差数据的基础上,不依赖于系统解析模型进行故障检测,为基于数据驱动方法进行复杂系统的故障检测提供了新的理论依据.为了提高系统故障检测的准确性,在分析故障检测漏报率和误报率的基础上,对故障检测阈值进行优化设计,并通过对车辆电源系统可能发生的励磁系统故障进行故障检测,验证了所提出方法的准确性和有效性.
This paper proposes a fault detection method based on the residual evaluation strategy. On the basis of the system residual error data only, do not rely on the system analytical model for fault detection, a new theoretical basis is provided for fault detection based on the data-driven method for the complex system. In order to improve the accuracy of the system fault detection, the fault detection threshold is optimally designed on the basis of fault miss alarm ratio and false alarm ratio. And by detecting the possible failure of the excitation system in the vehicle power system, the accuracy and effectiveness of the proposed method is verified.
出处
《控制与决策》
EI
CSCD
北大核心
2017年第7期1181-1188,共8页
Control and Decision
基金
国家自然科学基金项目(61364011)
甘肃省青年自然科学基金项目(1506RJYA108)
甘肃省先进控制重点实验室开放基金项目(XJK201505)
关键词
故障检测
数据驱动
K-L散度
残差评价
fault detection
data-driven
Kullback-Leibler divergence
residual evaluation
