基于卷积神经网络的飞机液压系统故障诊断

Fault Diagnosis of Aircraft Hydraulic System Based on Convolutional Neural Network

中文关键词:  卷积神经网络  多传感器数据  故障诊断  飞机液压系统

英文关键词:convolutional neural network  multi-sensor data  fault diagnosis  aircraft hydraulic system

中文摘要:

      现代飞机系统的复杂度和集成度均大幅提升，使得其故障诊断存在难度高和准确率低等特点。采用一维卷积神经网络方法对军用飞机液压系统的故障诊断问题进行了研究，构建了满足多传感器数据分析要求的卷积神经网络模型。考虑到神经网络的输入来自不同的传感器数据序列，各数据序列之间的空间关系不明显，因此，即使网络输入是二维形式，而实际的卷积操作均在一维上进行。通过解决某飞机液压系统的故障诊断问题，证明将标准化后的多传感器数据序列及对应故障模式作为训练样本对卷积神经网络模型进行训练时，采用满足训练要求的网络对飞机液压系统进行故障诊断时具有较高的准确率。

英文摘要:

      The complexity and integration of modern military aircraft system have been greatly improved,making its fault diagnosis difficult and inaccurate.Aiming at the fault diagnosis of aircraft hydraulic system,it uses a fault diagnosis method based on one-dimensional convolutional neural network.This method builds a convolutional neural network model that can be used for multi-sensor data analysis.Since the input of the neural network comes from different sensor data sequences,the spatial relationship between the data sequences is not remarkable.Therefore,even if the input of the network is two-dimensional,the convolution operation is carried out in one dimension.The standardized multi-sensor data sequence and the corresponding failure mode can be used as training sample to train the convolutional neural network.The trained network has a high accuracy for the fault diagnosis of military aircraft system,which has been demonstrated by solving a fault diagnosis problem of an aircraft hydraulic system.