摘要
推力轴承作为水轮机组的主要部件之一,其健康运行对提升水轮机组的运行效率及可靠性、降低水轮机组的运维成本、避免重大事故的发生具有重要意义。因此,针对推力轴承开展故障诊断算法研究尤为重要。借助状态监测和机器学习技术,数据驱动的诊断算法成为目前研究热点。然而,现有方法的成功大多依赖训练数据(源域)与测试数据(目标域)同分布的假设,但水轮机组的多变工况使得这一假设难以成立。因此,本文提出基于动态加权对抗域适配的智能诊断方法,以提升SkyME9000面向实际变工况场景时的故障诊断性能。所提方法借助基于成对正交分类器的对抗学习框架开展域适配,通过成对推理结果的一致性和熵值度量目标域样本的适配度和可分性,并开展样本级的动态加权,实现故障诊断知识从源域向目标域的迁移。利用重庆大学推力轴承数据集,开展故障诊断试验。结果表明,所提方法能有效提升变工况下推力轴承的诊断性能。
As one of the critical components in hydro-turbine units,the healthy operation of thrust bearing is of para⁃mount importance for enhancing operational efficiency and reliability,reducing maintenance costs,and preventing major accidents.Therefore,research on fault diagnosis algorithms for thrust bearings is of vital importance.Leverag⁃ing condition monitoring and machine learning techniques,data-driven diagnostic algorithms have become a hot research topic.However,the success of existing methods largely relies on the assumption that training data(source domain)and test data(target domain)follow the same distribution.However,such an assumption is hard to meet due to the variable working conditions of hydroelectric units.To address this challenge,this paper proposes a dynamic weighted adversarial domain adaptation-based intelligent diagnostic method(DWADA),so as to enhance the diag⁃nostic performance of SkyME9000 under variable working conditions.The proposed method employs an adversarial learning framework with pairwise orthogonal classifiers to perform domain adaptation.By measuring the consistency of pairwise inferences and entropy,it evaluates the adaptability and separability of target domain samples,enabling dynamic sample-level weighting to facilitate the transfer of fault diagnosis knowledge from the source to the target domain.Experiments conducted on the Chongqing University thrust bearing dataset demonstrate that the proposed method significantly enhances fault diagnosis performance under variable working conditions.
作者
陈子旭
李亦凡
张卫君
陈小松
CHEN Zixu;LI Yifan;ZHANG Weijun;CHEN Xiaosong(China Institute of Water Resources and Hydropower Research,Beijing 100038,China;Beijing IWHR Technology Co.,Ltd,Beijing 100038,China)
出处
《中国水利水电科学研究院学报(中英文)》
北大核心
2026年第2期239-247,共9页
Journal of China Institute of Water Resources and Hydropower Research
基金
水科院基本自科研项目(AU0145B022021,AU110145B0012025)。
关键词
推力轴承
变工况
动态加权
域适配
SkyME9000
故障诊断
thrust bearing
variable working condition
dynamic weight
domain adaptation
SkyME9000
fault diagnosis
