变工况条件下三相异步电机匝间短路故障诊断
Diagnosis of interturn short circuit fault in three-phase asynchronous motor under variable operating conditions
李剑君 1李昂 1王勇飞 1冯治国 1牛天宇2
作者信息
- 1. 国能大渡河检修安装有限公司 成都 610041
- 2. 四川大学电气工程学院 成都 610065
- 折叠
摘要
针对三相异步电机匝间短路故障在不同工况下数据分布不一致带来的泛化识别准确率下降的问题.提出了一种基于残差-自注意力网络的迁移学习方法,通过在残差网络中嵌入自注意力机制实现特征强化并利用源域数据进行模型训练,然后利用迁移学习的微调策略使得模型能更好地适应目标域的特征分布,以此来增加模型在目标域数据中的适应性能力.此外,通过设计对比实验探究了引入微调训练以及在模型中嵌入自注意力机制对于模型诊断性能的影响.实验结果表明,所提方法在3种负载条件下迁移的平均准确率为87.5%,相较于一般的残差网络准确率提高了4.5%,同时召回率和F1 分数分别提高了约10%和6%.
Abstract
The inconsistent distribution of data in different operating conditions poses a challenge in diagnosing inter-turn short circuit faults in three-phase motors.In this paper,a transfer learning method based on residual-self-attention network is proposed.By embedding self-attention mechanism in the residual network,feature enhancement is achieved.The model is trained using source domain data and then fine-tuned using transfer learning strategies to better adapt to the feature distribution of the target domain.Furthermore,a comparative experiment is designed to investigate the impact of fine-tuning training and embedding self-attention mechanism on the diagnostic performance of the model.The experimental results show that the average accuracy of the proposed method for migration under three different load conditions is 87.5%.Compared to the general residual network accuracy,it has increased by 4.5%.At the same time,the recall rate and F1 score have increased by approximately 10%and 6%respectively.
关键词
三相异步电机/故障诊断/匝间短路/变工况/迁移学习Key words
three-phase asynchronous motor/fault diagnosis/inter-turn short circuit/variable operating conditions/transfer learning引用本文复制引用
出版年
2024
NETL
NSTL
维普
万方数据