基于深度学习的飞机关键部件故障诊断研究
详细信息 查看官网全文
- 英文论文题名：Research on Aircraft Key Parts Fault Diagnosis Using Deep Learning
- 论文作者：邵海东 ; 姜洪开 ; 王福安
- 英文论文作者：Shao Haidong ; Jiang Hongkai ; Wang Fuan ; School of Aeronautics ; Northwestern Polytechnical University
- 年：2016
- 作者机构：西北工业大学航空学院;
- 论文关键词：深度学习 ; 飞机 ; 关键部件 ; 故障诊断 ; 滚动轴承
- 英文论文关键词：deep learning ; aircraft ; key parts ; fault diagnosis ; rolling bearing
- 会议召开时间：2016-11-25
- 会议录名称：2016年航空科学与技术全国博士生学术论坛摘要集
- 语种：中文
- 分类号：V267
- 学会代码：XBHH
- 会议名称：2016年航空科学与技术全国博士生学术论坛
- 会议地点：中国陕西西安
- 主办单位：西北工业大学研究生院、中国高校航空学院院长联席会
- 学会名称：西北工业大学航空学院
- 页数：2
- 文件大小：567k
- 原文格式：D
- 会议级别：全国
摘要
飞机是集各种高新技术于一体的大型复杂系统。在飞行过程中,由于高温高速的恶劣工作环境,飞机的关键部件(如叶片、轴、轴承等)会不可避免地会发生故障,这些故障往往会影响飞机的整体性能,甚至造成机毁人亡。飞机工作环境复杂多变,且飞机系统组成非常复杂,部件之间互相关联且紧密耦合,因此其故障往往具有多样性、隐蔽性、不确定性以及因果关系复杂等新特点,这给传统的飞机关键部件故障诊断方法带来了极大的挑战。因此,本文提出了深度学习新方法来有效解决这个实际问题。深度学习通过构建含多隐层的学习模型,实现逐层的特征变换,从而自适应地捕获隐藏于故障数据内部的有用信息,最终提升诊断的准确性。将深度学习方法用于滚动轴承的故障诊断,其识别精度远高于传统的人工神经网络和支持向量机。
As a large and complex system integrated with various advanced technologies, aircraft will inevitably fail in its key parts(axle, bearing, blade and so on) owing to high temperature and speed during the flight, which will affect the overall performance of the aircraft and lead to serious casualties. Aircraft key parts faults show the new characteristics such as diversity, concealment, uncertainty and so on because of the uncertain working conditions and complex system components. Hence, it is a great challenge to diagnose the various faults and assess the health status using traditional intelligent methods. In this paper, a novel intelligent method based on deep learning is proposed for aircraft key parts fault diagnosis. Through deep learning, deep neural networks with deep architectures, instead of shallow ones, can be established to adaptively capture the useful information from input data and approximate complex non-linear functions in fault diagnosis issues. The proposed method is applied to analyze the experimental rolling bearing signals. The resultsshow that the proposed method demonstrates evident advantages over traditional intelligent methods such as artificial neural network(ANN) and support vector machine(SVM).
As a large and complex system integrated with various advanced technologies, aircraft will inevitably fail in its key parts(axle, bearing, blade and so on) owing to high temperature and speed during the flight, which will affect the overall performance of the aircraft and lead to serious casualties. Aircraft key parts faults show the new characteristics such as diversity, concealment, uncertainty and so on because of the uncertain working conditions and complex system components. Hence, it is a great challenge to diagnose the various faults and assess the health status using traditional intelligent methods. In this paper, a novel intelligent method based on deep learning is proposed for aircraft key parts fault diagnosis. Through deep learning, deep neural networks with deep architectures, instead of shallow ones, can be established to adaptively capture the useful information from input data and approximate complex non-linear functions in fault diagnosis issues. The proposed method is applied to analyze the experimental rolling bearing signals. The resultsshow that the proposed method demonstrates evident advantages over traditional intelligent methods such as artificial neural network(ANN) and support vector machine(SVM).
引文