基于模糊神经网络的综合导航系统故障检测与诊断技术

文中主要针对综合导航系统对于缓变故障检测困难，以及故障难以定位等问题，提出一种基于状态χ^2。检验和ARTMAP神经网络技术的故障检测与诊断系统。用基于BP神经网络的状态χ^2检验法检测出故障数据后，以故障时刻的SCST值作为两个ARTMAP网络的输入模式，判断故障源、故障幅值和故障发生时间。并以SINS／GPS综合导航系统实际跑车数据为基础进行缓变故障仿真，验证该方法的有效性和检测精度。

Fault Diagnosis in Integration Navigation System Using BP Neural Networks And Fuzzy ARTMAP Neural Networks

A failure diagnosis scheme using the state chi-square test and the ARTMAP neural network techniques is presented for dynamic systems, in order to detect and diagnose soft failures. After a failure is detected by the state chi-square test, two ARTMAP neural networks are applied to recognize patterns formed by data of the state chi-square test in order to determine when the failure happens, how serious the failure is, and where the failure is located. The efficacy of the proposed scheme is demonstrated by simulation examples where soft failures in a SINS/GPS integrated navigation system are successfully diagnosed.