基于 Stacking 的雷达伺服系统故障诊断

为了提高南京某所某型雷达伺服系统故障诊断准确率，考虑到传统故障诊断算法的局限性，提出一种基于 Stacking 集成算法的雷达伺服系统故障诊断方法。针对某所某型雷达伺服系统的历史监测数据，首先采用孤立森林算法识别异常样本|然后基于原始数据构造出新的特征，使用卡方检验进行特征选择，并使用SMOTE 算法解决样本不平衡问题|最后，通过建立一种新颖、准确的基于 XGBoost、随机森林和 BP 神经网络的Stacking 集成模型进行故障诊断。实验结果表明，该方法在测试集上的诊断准确率达到了 96.2%，比传统方法诊断准确率提高了 1.8%，证明该方法能够很好地完成雷达伺服系统故障诊断任务。

Fault Diagnosis of Radar Servo System Based on Stacking

In order to improve the accuracy of fault diagnosis for a certain type of radar servo system in a research institute in Nanjing，with the consideration of the limitations of traditional fault diagnosis algorithms，a fault diagnosis method of radar servo system basedon Stacking integrated algorithm is proposed. Aiming at the historical monitoring data of a certain type of radar servo system，we first use the isolated forest algorithm to identify abnormal samples. Then，new features are constructed based on the original data，feature selection is performed bt using Chi-square test，and sample imbalance is solved by using SMOTE algorithm. Finally，a novel and accurate Stacking integrated model based on XGBoost，random forest and BP neural network is established for fault diagnosis. The experimental results show that the diagnostic accuracy of the traditional method on the test set is up to 94.4%，and the diagnostic accuracy of the method proposed in this paper reaches 96.2%，which is 1.8% higher than the traditional method. The method proposed in this paper can well complete the fault diagnosis task of radar servo system.