一种基于神经网络的传感器故障诊断方法

提出了基于两级神经网络结构的多传感器斜置组件的故障诊断方法，以消除传感器安装误差、刻度系数误差以及常值偏差对故障检测与隔离准确性的影响。与基于参数估计的奇偶向量补偿方法相比，这种方法不需要各项误差的动态模型和噪声的统计特性。

Fault Diagnosis of Sensors Based on Neural Networks

A neural network based on the scheme for fault detection and isolation (FDI) of a redundant strapdown inertial measurement unit is described.The proposed method uses two feedforward networks.The first is trained to compute the parity vector and eliminate the effects of input axis misalignment,scale factor error and biases.The second is designed to detect and identify the failures with the computed parity vector.Comparing with other compensation algorithms based on parameter estimation,the proposed technique does not need to have dynamic model of error states and statistics of noise.The simulation results show that the technique can significantly improve the FDI performance,especially during vehicle maneuvers.