基于深度学习的航空发动机不平衡故障部位识别

针对基于机匣测点的航空发动机不平衡故障部位识别问题,提出了基于深度卷积神经网络的航空发动机不平衡故障部位诊断方法。针对某典型双转子航空发动机,建立整机耦合动力学模型,并利用数值积分算法实现不平衡故障数值仿真;在从发动机压气机端到涡轮端的高、低压转子上选择4个不平衡故障部位作为诊断对象,通过仿真分析得到发动机典型转速下的转子不同部位不平衡故障的仿真样本;计算4个机匣测点信号的规范化频谱,通过对大量仿真数据的处理得到反映不同不平衡故障部位的故障样本集;利用仿真得到的大量不平衡故障样本,训练深度卷积神经网络,利用深度卷积神经网络的优良特征学习能力实现航空发动机不平衡故障的不同部位进行识别,数值试验结果表明该方法对航空发动机不平衡故障部位的识别准确率达到95%。

Aero-engine unbalanced fault location identification method based on deep learning

For the problem of aero-engine unbalanced fault location diagnosis based on casing test points, a method of aero-engine unbalanced fault location diagnosis based on deep convolution neural network was presented. The coupling dynamic model of a typical dual-rotor aero-engine was established, and the numerical integration method was used to realize the numerical simulation of unbalanced fault. Four unbalanced fault positions were selected from the high and low pressure rotors of the compressor end to the turbine end as the diagnostic object. A large number of unbalanced fault samples obtained by simulation were used to train the deep convolution neural network, and the excellent feature learning ability of the deep convolution neural network was used to realize the identification of different positions of the aeroengine unbalanced fault. The numerical experimental results fully showed the accuracy of the method to identify the unbalanced fault locations of aero-engine reached to 95%.