神经网络集成在发动机故障诊断中的应用研究

提出一种基于A daBoost的集成神经网络故障诊断方法,利用多层前向神经网络作为故障弱分类器,通过简单地训练若干个单一神经网络并将其预测结果进行合成,实现了对航空发动机多类故障的诊断。针对一个涡轮喷气发动机气路部件的仿真实验表明,这种方法提高了最终故障分类器的泛化能力,便于工程应用。 

Aero-Engine Fault Diagnosis Using Neural Network Ensemble Method

This paper developed a multiclass fault diagnosis method for aero-engine,which used three-layer perceptrons(feed-forward neural networks) as weak classifiers and then they were combined together to create an aggregate hypothesis with AdaBoost algorithm.A final fault classifier is the neural network ensemble which was formed by the simply trained three-layer perceptrons and this made it possible to design practical neural network fault classifiers easily.A simulation experiment for the gas path components of a turbojet engine was conducted to demonstrate the effectiveness of the method.Through the experiment,24 groups of fault-test data of the turbojet engine were all correctly classified into 5 classes.The experimental results show that the generalization ability of the final fault classifier is improved effectively.