基于DTCWT与深度残差网络的风电机组轴承故障诊断方法研究
Research on Fault Diagnosis Method for Wind Turbine Bearings Based on DTCWT and Deep Residual Networks
李冲 1郭莉莉 1胡杰 1雷少华 1王荣振 2夏冰2
作者信息
- 1. 徐州高新区安全应急装备产业技术研究院,江苏 徐州 221008
- 2. 中国矿业大学物联网(感知矿山)研究中心,江苏 徐州 221008
- 折叠
摘要
随着风力发电技术的迅速发展,风电机组的可靠性和性能成为风电项目高效运维持续关注的焦点.轴承是风电机组关键的机械部件,运行过程中容易出现磨损,因此,对轴承状态进行及时的诊断和维护至关重要.基于此,设计了一种基于双树复小波变换与深度残差网络相结合的风电机组轴承故障诊断模型DTRSANMD.该模型利用双树复小波变换对风电机组振动信号进行多尺度分解,将得到的特征信息输入加入了注意力机制的深度残差网中以获取有效的深度特征表示,并引入多核最大均值差异方法对变工况特征分布进行评价,有效地降低了源域和目标域之间的分布差异,提高了模型的泛化性能.最后,设计了风电机多源数据智能采集分析终端与系统,实现了风电机组机械部件的远程诊断分析,降低了维护成本.
Abstract
With the rapid development of wind power technology,the reliability and performance of wind turbines have become a focal point for efficient and continuous operation in wind power projects.Bearings,as critical mechanical components in wind turbines,are prone to wear during operation.Therefore,timely diagnosis and maintenance of bearing conditions are crucial.This paper presents a Wind Turbine Generator Bearing Fault Diagnosis Model called DTRSANMD based on the combination of dual-tree complex wavelet transform and deep residual networks.Firstly,the vibration signals of wind turbine generators are subjected to multiscale decomposition using dual-tree complex wavelet transform.Subsequently,the extracted feature information is input into a deep residual network incorporating attention mechanisms to obtain effective deep feature representations.Finally,the introduction of the multi-kernel maximum mean discrepancy method evaluates the feature distribution under variable operating conditions,effectively reducing distribution differences between the source and target domains,and thereby enhancing the model's generalization performance.In conclusion,a Wind Turbine Generator Multi-Source Data Intelligent Collection Analysis Terminal and System is designed,to enable remote diagnosis and analysis of mechanical components in wind turbines.This implementation contributes to the reduction of maintenance costs.
关键词
风电机组/故障诊断/深度残差网络/双树复小波变换/MK-MMDKey words
wind turbine unit/fault diagnosis/resnet/DTCWT/MK-MMD引用本文复制引用
基金项目
国家重点研发计划(2017YFC0804400)
国家重点研发计划(2017YFC0804401)
出版年
2024
NETL
NSTL
万方数据