基于神经网络观测器的飞行器传感器故障检测

Sensor Fault Detection for Aircraft Based on Neural Network Observer

为了避免传感器故障对飞控系统的影响,实现传感器故障的快速检测与隔离,提出了一种基于神经网络观测器(NNOB)的传感器故障检测方法。在建立四旋翼飞行器姿态故障模型的基础上,利用非线性观测器得到的期望输出和传感器测量值设计基于神经网络(NN)的传感器故障观测器,利用扩展卡尔曼滤波器(EKF)更新神经网络的权值参数,通过Lyapunov理论证明权值参数更新的收敛性,最终构建出一种基于神经网络观测器的传感器故障检测系统。数值仿真实验结果表明,与现有神经网络故障检测方法相比,所提方法具有更高的故障检出率与更好的跟踪性能。

In order to avoid the influence of sensor faults on flight control system,and realize fast detection and isolation of sensor fault,a sensor fault detection method based on neural network observer(NNOB) is proposed.Based on the attitude fault model of quad-rotor aircraft,a sensor fault observer based on neural network(NN) is designed by using the expected output of the nonlinear observer and the sensor measurements.The weight parameters of the NN are updated by the extended Kalman filter(EKF).The convergence of the weight parameter updating is proved by Lyapunov theory.Finally,a sensor fault detection system based on NNOB is presented.The numerical simulation results show that the proposed method has better fault detection rate and tracking performance than the existing NN based fault detection methods.