基于神经网络的航空传感器故障检测

用离线训练的神经网络进行导航传感器故障检测。首先,用已获得的正常飞行数据通过离线训练的方法训练神经网络并构造估计器的结构,然后用已选择好结构并训练好的神经网络作为估计器对传感器的读数进行一步预测。若预测值与传感器实际值之间的差值仅为递推误差和传感器输出噪声,则认为传感器工作正常,若相应的残差分量显著增大,则认为传感器故障。因此设计了相应的检测策略进行故障检测,以达到既避免不必要的报警、切换,又准确、及时的监测、报警。通过仿真试验验证,结果证明该方法可行。

Aeronautical Sensor Fault Detection Based on Neural Networks

The neural networks are used to perform sensor fault detection by off-line training.In the proposed approach,at the first,neural networks are trained by normal aero data and construct estimator.And then,the selected and trained neural networks are used as estimator to predict the next sensor readings.If the discrepancy between the estimation and sensor reading is only the recursive error,the sensor is in working order,else the sensor is in failure.Therefore the definite strategy is designed to prevent the needless alarm and switching.The simulation result shows the proposed approach is feasible.