中文摘要：

飞行控制系统作为无人机（unmanned aerial vehicle,UAV）的核心子系统,对其进行故障诊断可以大大提高无人机的安全性和可靠性。在无人机数学模型未知或者不确定的情况下,数据驱动的故障诊断方法比基于模型的方法更实用。考虑无人机飞行控制系统是典型的非线性动态系统,采用一种非线性主元分析方法对其进行故障诊断。利用数据建立无人机飞行控制系统正常状态下的动态核主元模型,通过T2和SPE两种统计量实现故障检测;故障发生后,利用重构贡献图的方法进行故障分离。仿真试验证明,该方法能对典型的无人机执行器和传感器故障进行有效监测和诊断。与动态主元分析相比,动态核主元分析方法对微小故障更为敏感。

英文摘要：

The flight control system （FCS） was the core subsystem of unmanned aerial vehicle （UAV）, performing FD for it could greatly improve the safety and reliability of UAV. When the mathematical model of UAV was unknown or uncertain, data-driven methods were more suitable than model-based FD methods. Considering that FCS of UAV was a typical nonlinear dynamic system, a nonlinear principal component analysis （PCA） method was used instead. A dynamic kernel principal component model under normal state was established for UAV, then fault detection was performed by T2 and SPE statistics; When a fault was detected, a method called reconstruction-based contribution was used for fault isolation. The simulation results showed that the proposed method could achieve better fault diagnosis effect for typical faults of actuators and sensors than linear DPCA model. Besides, DKPCA could achieve high sensitivity for small faults of UAV.