增强双流Transformer的柴油发动机剩余寿命预测模型  

Enhanced Two-Stream Transformer Model for Remaining Useful Life Prediction of Diesel Engines

作　　者：张曦 杨颖[1] 陈超君[2] 王春风[2] 杨磊[3] Zhang Xi;Yang Ying;Chen Chaojun;Wang Chunfeng;Yang Lei(School of Computer,Electronics and Information,Guangxi University,Nanning 530004;Process and Engineering Department,Guangxi Yuchai Machinery Co.,Ltd.,Yulin 537005;Guangxi Academy of Science,Nanning 530007)

机构地区：[1]广西大学计算机与电子信息学院,南宁530004 [2]广西玉柴机器股份有限公司工艺工程部,玉林537005 [3]广西科学院,南宁530007

出　　处：《汽车工程》2025年第2期292-300,325,共10页Automotive Engineering

基　　金：广西创新驱动发展专项(桂科AA20302002-1);广西科技基地与人才专项(桂科AD21076002)资助。

摘　　要：基于Transformer的模型在剩余使用寿命(remaining useful life,RUL)预测方面取得了显著的进展。然而,现有Transformer模型主要存在以下不足:模型在提取局部特征方面有所欠缺,且没有同时考虑输入特征的不同时间和不同空间的重要性。针对以上问题,提出一种增强的双流Transformer模型,通过局部特征提取模块和交互融合模块对模型进行增强。首先,通过局部特征提取模块分别在时间流和空间流提取局部特征,以弥补Transformer在局部特征提取方面的不足。然后,使用双流Transformer分别在时间和空间维度提取长期依赖,增强双流分支的互补学习。最后,构建交互融合模块,通过双线性融合方法捕获流级交互,进一步提升预测效果。使用多个模型在某柴油发动机制造商两个真实的数据集上进行实验,其结果表明评价指标RMSE和Score至少分别降低3.23%和5.89%。Transformer-based models have made significant progress in Remaining Useful Life(RUL)pre-diction.However,existing Transformer models have the following limitation of difficulty in local feature extraction and failure to consider the importance of varying temporal and spatial input features.To solve the problems,in this paper,an enhanced two-stream Transformer model is proposed,which is reinforced by the local feature extraction module and the interaction fusion module.Firstly,the local feature extraction module captures local features from both the temporal and spatial streams to compensate for the Transformer′s deficiency in local feature extraction.Then,the two-stream Transformer is used to extract long-term dependencies in the temporal and spatial dimen-sions,enhancing complementary learning between the two streams.Finally,the interaction fusion module is con-structed to capture stream-level interaction using bilinear fusion,further improving prediction performance.Experi-ments using multiple models on two real-world datasets from a diesel engine manufacturer demonstrate that the eval-uation metrics RMSE and Score are reduced by at least 3.23%and 5.89%,respectively.

关 键 词：剩余使用寿命预测 Transformer编码器 卷积神经网络 特征融合 滑动窗口 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TK428[自动化与计算机技术—控制科学与工程]