基于CNN特征提取及模型融合的飞机液压系统故障诊断

Fault Diagnosis of Aircraft Hydraulic System Based on CNN Feature Extraction and Model Fusion

为了对飞机液压系统进行有效故障诊断,采用CNN对飞机液压系统的压力信号进行特征提取。用提取到的特征输入线性模型、决策树、支持向量机、k邻近等算法对其进行故障诊断,并使用Stacking模型融合技术将多个模型融合。结果表明,相比于直接用CNN训练进行故障诊断,使用CNN提取出的特征进行训练能极大减少训练时间同时提高准确率。

To diagnose the fault of the aircraft hydraulic system effectively,CNN is used to extract the feature of the pressure signal of the aircraft hydraulic system.The extracted features are input into the linear model,decision tree,support vector machine,k-nearest neighbor algoG2rithm for tthe fault diagnosis.Stacking model fusion technique is used to fuse models.The results show that the features extracted from CNN can be used to reduce the training time as well as improve the accuracy,compared with using CNN training directly for the fault diagnosis.