域适应残差网络与多通道融合的齿轮箱故障诊断  

Gearbox Fault Diagnosis Based on Domain Adaptive Residual Network and Multi-channel Fusion

作　　者：张飞 万安平 ZHANG Fei;WAN Anping(School of Mechanical and Electrical Engineering,Anhui University of Science and Technology,HuainanAnhui 232000,China;Department of Electrical and Mechanical Engineering,Hangzhou City University,Hangzhou 310015,China)

机构地区：[1]安徽理工大学机电工程学院,安徽淮南232000 [2]浙大城市学院机电系,浙江杭州310015

出　　处：《兰州工业学院学报》2025年第2期73-78,共6页Journal of Lanzhou Institute of Technology

基　　金：国家自然科学基金(52372420)。

摘　　要：针对齿轮箱故障诊断中存在的单一传感器数据有限、可靠性低和泛化能力不足等问题,提出一种结合域适应残差网络与多通道融合的方法。首先,采用粒子群算法对变分模态分解(VMD)进行参数寻优;其次,对齿轮箱多通道原始数据进行VMD分解,进行初步特征提取,并引入注意力机制对各个通道的特征进行加权;然后,利用改进的残差网络进行深度特征提取来增强模型对数据的表征能力;最后,采用最大均值差异(MMD)和交叉熵分类损失作为优化目标,使用softmax分类器输出故障类别的预测概率,通过微调目标域模型,实现了齿轮箱在多工况下的样本分类任务。与加权多通道融合深度迁移学习算法相比,在4种不同工况下的诊断精度分别提高了3.38%、4.17%、3.13%和1.25%。Aiming at the problems of limited data,low reliability and insufficient generalization ability of single sensor in gearbox fault diagnosis,a method combining domain adaptive residual network and multi-channel fusion is proposed.Firstly,particle swarm optimization is used to optimize parameters of variational mode decomposition(VMD).Secondly,VMD decomposition of the multi-channel raw data of the gear box is carried out to extract preliminary features,and attention mechanism is introduced to weight the features of each channel.Then,the improved residual network is used to extract deep features to enhance the ability of the model to represent the data.Finally,the maximum mean difference(MMD)and cross entropy classification loss are taken as optimization objectives,and the predicted probability of the fault category is output by softmax classifier.By fine-tuning the target domain model,the task of sample classification of the gearbox under multiple working conditions is realized.Compared with the weighted multi-channel fusion deep transfer learning algorithm,the diagnostic accuracy is improved by 3.38%,4.17%,3.13%and 1.25%respectively under four different working conditions.

关 键 词：域适应 齿轮箱 故障诊断 迁移学习 多通道融合 

分 类 号：TH133.3[机械工程—机械制造及自动化]