强噪声干扰下基于双特征提取器的两阶段对抗迁移故障诊断方法  

Two-stage Adversarial Transfer Fault Diagnosis Method Based on Dual Feature Extractor under Strong Noise Interference

作　　者：吕铖跃 陈曦晖 杨一鸣 彭晓勇 丁坤[1] Lyu Chengyue;Chen Xihui;Yang Yiming;Peng Xiaoyong;Ding Kun(College of Mechanical and Electrical Engineering,Hohai University,Changzhou 213200,China)

机构地区：[1]河海大学机电工程学院,江苏常州213200

出　　处：《煤矿机械》2025年第5期167-170,共4页Coal Mine Machinery

基　　金：国家自然科学基金项目(51905147);江苏省自然科学基金面上项目(BK20201163);常州市应用基础研究计划项目(CJ20220208)。

摘　　要：齿轮箱是矿山机械中关键传动部件,工作时常受到外部噪声干扰导致数据分布差异增大,给故障诊断带来了困难。针对以上问题,提出强噪声干扰下基于双特征提取器的两阶段对抗迁移故障诊断方法。首先,为提升对无干扰源域数据和含噪声目标域数据的特征提取效果,增加目标域特征提取器与源域特征提取器构成双特征提取器结构,兼顾两域各自特点;然后,提出两阶段训练结构提升对抗训练性能,继承源域知识并进一步对齐两域分布,增加迁移诊断上限。基于自主设计的齿轮箱故障模拟实验平台,构建从无干扰振动数据到5个不同程度噪声干扰振动数据的迁移故障诊断任务,验证了该方法的有效性和优越性。As a key transmission component in mining machinery,the gearbox often working with external noise interference leads to increased discrepancy in data distribution,which brings difficulties in fault diagnosis.Aiming at the above problems,a two-stage adversarial transfer fault diagnosis method based on dual feature extractor under strong noise interference was proposed.Firstly,in order to improve the feature extraction effect on the interference-free and the noise containing data,the target domain feature extractor was added and constituted a dual feature extractor structure with the source domain feature extractor,which takes into account the respective features of the two domains;then,a two-stage training structure was proposed to improve the performance of the adversarial training,which inherits the knowledge of the source domain and further aligns the distribution of the two domains to increase the maximum limit of the transfer diagnosis.Based on the self-designed gearbox fault simulation experimental platform,the transfer fault diagnosis tasks from interference-free vibration data to five different degrees noise interference vibration data were constructed to verify the effectiveness and superiority of this method.

关 键 词：故障诊断 噪声干扰 域对抗网络 迁移学习 

分 类 号：TD407[矿业工程—矿山机电] TP18[自动化与计算机技术—控制理论与控制工程]