基于主成分分析和奇偶向量的动态检测方法

Dynamic detection method based on principal component analysis and parity vector

摘要:

导航系统中冗余IMU传统故障检测方法由于数学模型过于复杂，计算量大，存在较大延时，难以实现实时故障检测，而主成分分析法仅仅应用于静态情况下的故障检测与隔离，针对主成分分析法无法在动态情况下对冗余IMU进行故障检测的缺点，提出了一种基于奇偶空间法改进主成分分析的故障检测算法，该方法利用奇偶向量隔离车辆的动态变量，以消除动态变量对故障检测的影响，再用PCA方法检测数据以实现对车辆传感器信息的实时检测，通过将原始数据集转置到特征平面来形成图案，实现了IMU传感器正常与故障模式的准确分离，提高了冗余IMU故障检测的结果精确性和可靠性。实验结果表明，该方法能够较好检测动态状态下冗余IMU的故障，提高了主成分分析的故障检测性能，可有效消除导航系统运动的负面影响。

Abstract:

The traditional fault detection method of redundant IMU in navigation system is difficult to realize real-time fault detection due to its too complex mathematical model, large calculation and large delay. However, PCA is only applied to fault detection and isolation in static situation. Aiming at the disadvantage that PCA is not able to detect redundant IMU in dynamic situation, a fault detection method based on parity space is proposed This method uses even and odd vectors to isolate the dynamic variables of vehicles, so as to eliminate the influence of dynamic variables on fault detection. Then PCA method is used to detect the data to realize the real-time detection of vehicle sensor information. By transposing the original data set to the feature plane to form a pattern, the normal and fault modes of IMU sensors are realized accurately Separation improves the accuracy and reliability of fault detection results of redundant IMU. The experimental results show that this method can detect the faults of redundant IMU in dynamic state, improve the performance of PCA, and effectively eliminate the negative effects of navigation system motion.