大转数波动条件下机理约束图权重增强网络的航空发动机附件机匣故障诊断方法
Fault Diagnosis for Aero-engine Accessory Gearbox by Mechanism Constraint Graph Weight Enhancement Networks under the Large Revolution Fluctuations
余晓霞 1汤宝平 2魏静 2张志刚1
作者信息
- 1. 重庆理工大学机械工程学院 重庆 400054
- 2. 重庆大学机械传动全国重点实验室 重庆 400044
- 折叠
摘要
针对现有深度学习模型在故障特征提取过程中容易受到大转数波动工况影响,导致航空发动机附件机匣故障诊断精度较低的问题,提出了一种机理约束图权重增强网络(Mechanism constraint graph weight enhancement networks,MCGWENet)的航空发动机附件机匣故障诊断方法.首先通过度量振动信号时域和频域特征的欧式距离约束图的邻接矩阵,将物理机理嵌入到所构建的图结构中;然后将振动信号的小波包分解结果作为节点特征;并结合所设计的邻接矩阵构造出可用于航空发动机附件机匣故障诊断的图;最后,通过所设计的图权重增强层抑制大转数波动对故障特征提取的影响,提升所提模型的故障诊断精度.试验结果表明,所提模型能够有效识别机匣故障,可用于航空发动机附件机匣健康管理.
Abstract
Aiming at the problem that the existing deep learning models are easily affected by the large revolution fluctuation in the process of fault feature extraction,which leads to the low accuracy of aero-engine accessory gearbox fault diagnosis,a mechanism constraint graph weight enhancement networks(MCGWENet)is proposed for accessory gearbox fault diagnosis.First,the adjacency matrix of the graph is constrained by measuring the euclidean distance of the time and frequency domain characteristics of the vibration signal,and the physical mechanism is embedded into the constructed graph structure.Then the results of wavelet packet decomposition of vibration signals are used as node features;and the designed adjacency matrix is combined to construct a graph that can be used for the fault diagnosis of aero-engine accessory gearbox.Finally,the designed graph weight enhancement layer is used to suppress the influence of large revolution fluctuation conditions on fault feature extraction and improve the fault diagnosis accuracy of the proposed model.The experimental results show that the proposed method can effectively identify faults and can be used for health management of aero-engine accessory gearbox.
关键词
航空发动机附件机匣/机理约束/图权重增强网络/大转数波动/故障诊断Key words
aero-engine accessory gearbox/mechanism constraint/graph weight enhancement network/large revolution fluctuation/fault diagnosis引用本文复制引用
基金项目
国家重点研发计划(2020YFB1709800)
国家自然科学基金(51775065)
重庆市高校创新研究群体(CXQT21027)
装备预研教育部联合基金()
重庆英才计划包干制(cstc2021ycjhbgzxm0261)
重庆理工大学科研启动基金(0119230961)
重庆理工大学国家自然科学基金培育(0119230874)
出版年
2024
NETL
NSTL
万方数据