中国烟草学报2025，Vol.31Issue(3)：70-77,8.DOI:10.16472/j.chinatobacco.2024.T0054
基于改进Transformer的烟机设备健康状态诊断方法
Health status diagnosis method for tobacco machine equipment based on improved transformer
摘要
Abstract
[Objective]This study proposes a rapid and effective method for diagnosing the health status of tobacco machine equipment,aiming to enhance the timeliness and accuracy of maintenance.[Methods]By collecting multimodal parameter data of the entire set of tobacco machine equipment,a model based on the improved Transformer algorithm was designed and trained.The model fully utilizes the improved Transformer architecture,effectively capturing long-distance dependencies between data through the self-attention mechanism,enhancing the ability to extract features from multi-dimensional and nonlinear signals.Simultaneously,the model achieves precise identification of key parameters by dynamically adjusting the weights of each monitoring point and introduces a weighted cross-entropy loss function to address data imbalance issues.[Results]After training and validation,the model achieved a diagnosis accuracy of 98.98％,a recall rate of 99.23％,and a fault prediction success rate of 94.11％,demonstrating outstanding performance in practical applications and significantly outperforming traditional diagnostic methods.[Conclusion]This method can accurately determine the health status of tobacco machine equipment and has significant practical application potential.关键词
烟机设备/健康状态/模型诊断/TransformerKey words
cigarette equipment/health status/model diagnosis/transformer引用本文复制引用
王茂方,陈圣,金义龙..基于改进Transformer的烟机设备健康状态诊断方法[J].中国烟草学报,2025,31(3):70-77,8.基金项目
红云红河烟草(集团)有限责任公司科技项目"卷烟设备健康状态监测系统的研究"(HYHH2021XX02) （集团）
