基于FEDformer-MTFA的轻工装备多源时序数据异常检测*

doi:10.16731/j.cnki.1671-3133.2026.06.012

摘要/Abstract

摘要： 在工业智能化与制造业数字化转型背景下,针对传统频域模型对温度骤升等局部突变不敏感、多源数据直接拼接导致时空关联性缺失等问题,提出了一种基于改进FEDformer(Frequency Enhanced Decomposition Trans-former)模型的轻工装备多源时序数据异常检测模型FEDformer-MTFA。该模型包含多尺度时频联合注意力机制,在FEDformer已有频域增强机制的基础上,通过融合不同时间尺度下的时域特征与多频带频域信息,并结合权重分配策略,进一步扩展了注意力的建模维度。该机制能够同时捕捉设备运行状态在多个时间粒度和频域分布上的动态特征,从而显著增强对微弱瞬态异常信号的敏感性与表征能力。同时,模型采用多任务检测头设计,并行预测重构误差与异常概率分布,通过可调节的超参数动态平衡两项任务的优化权重,有效缓解了单一优化目标导致的模型过拟合问题,提升了检测鲁棒性。此外,针对样本不平衡问题,模型引入联合加权损失函数,通过调整样本权重,使训练过程更加聚焦于潜在异常区域的学习,进一步优化了模型在复杂工况下的检测性能。实验结果表明,所提模型关键指标相较于FEDformer,准确率提升0.186 0,AUC值提升0.298 2。

关键词: 轻工装备智能运维, 多源时序数据分析, 工业异常检测, 时频联合注意力机制, 多任务学习

Abstract: Under the strategy of industrial intelligence and the digital transformation of manufacturing,traditional frequency-domain models often exhibit insensitivity to localized anomalies such as sudden temperature spikes,while direct concatenation of multi-source data leads to the loss of spatiotemporal correlations. To address these issues,a multi-source time-series anomaly detection model for light industrial equipment based on an improved Frequency Enhanced Decomposition Transformer (FED-former) model,termed FEDformer-MTFA,is proposed. The model incorporates a multi-scale time-frequency joint attention mechanism,which integrates time-domain features at different temporal scales with multi-band frequency-domain information,combined with an adaptive weight allocation strategy. This approach overcomes the limitations of single-scale frequency-domain attention in traditional models like FEDformer. The mechanism enables the simultaneous capture of dynamic characteristics of equipment operating states across multiple temporal granularities and frequency distributions,thereby significantly enhancing sensitivity and representational capability towards weak transient anomalous signals. Additionally,the model employs a multi-task detection head design,simultaneously predicting reconstruction errors and anomaly probability distributions. By dynamically balancing the optimization weights of the two tasks through adjustable hyperparameters,it effectively mitigates overfitting caused by a single optimization objective,thereby improving detection robustness. Furthermore,to tackle sample imbalance,the model incorporates a joint weighted loss function that dynamically adjusts sample weights,focusing training on potential anomaly regions and further optimizing detection performance under complex working conditions. Experimental results show that there are 0.186 0 and 0.298 2 improvements in accuracy and AUC,respectively,compared to FEDformer.

Key words: intelligent maintenance of light industrial equipment, multi-source time-series data analysis, industrial anomaly detection, time-frequency joint attention mechanism, multi-task learning

中图分类号:

TP18

卫嘉文, 吉卫喜. 基于FEDformer-MTFA的轻工装备多源时序数据异常检测^*[J]. 现代制造工程, 2026, 549(6): 110-127.

WEI Jiawen, JI Weixi. Anomaly detection in multi-source time-series data of light industrial equipment based on FEDformer-MTFA[J]. Modern Manufacturing Engineering, 2026, 549(6): 110-127.

参考文献

[1] 胡珉,白雪,徐伟,等. 多维时间序列异常检测算法综述[J]. 计算机应用,2020,40(6):1553-1564.
[2] RUFF L,JACOB R,ROBERT A,et al. A unifying review of deep and shallow anomaly detection[J]. Proceedings of the IEEE,2021,109(5):756-795.
[3] SCHAFFER A L,DOBBINS T A,PEARSON S A. Inter-rupted time series analysis using Autoregressive Integrated Moving Average (ARIMA) models:a guide for evaluating large-scale health interventions[J]. BMC Medical Research Methodology,2021,21:58-70.
[4] PANG G,SHEN C,CAO L,et al. A deep learning for anomaly detection:a review[J]. ACM Computing Surveys,2020,54(2):1-38
[5] YU Y,SI X,HU C. A review of recurrent neural networks:LSTM cells and network architectures[J]. Neural Computation,2019,31(5):1235-1270.
[6] LU W,CHENG Y,XIAO C,et al. Unsupervised sequential outlier detection with deep architectures[J]. IEEE Transactions on Image Processing,2017,26(9):4321-4330.
[7] LIU Y,XUA C,HUANGC B,et al. Landslide displacement prediction based on multi-source data fusion and sensitivity states[J]. Engineering Geology,2020,271:105608.
[8] GUNGOR E,OZMEN A. Distance and density based clustering algorithm using Gaussian kernel[J]. Expert Systems with Applications,2017,69:10-20.
[9] SU Y,ZHAO Y,NIU C,et al. Robust anomaly detection for multivariate time series through stochastic recurrent neural network[C]//Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York:Association for Computing Machinery,2019:2828-2837.
[10] AUDIBERT J,MICHIARDI P,GUYARD F,et al. USAD:Unsupervised Anomaly Detection on multivariate time series[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York:Association for Computing Machinery,2020:3395-3404.
[11] ZHANG J,PASCHALIDIS I C. Statistical anomaly detection via composite hypothesis testing for Markov models[J]. IEEE Transactions on Signal Processing,2018,66(3):589-602
[12] ZHONG J,WANG D,LI C. A nonparametric health index and its statistical threshold for machine condition monitoring[J]. Measurement,2021,167:108290.
[13] DENG A,HOOI B. Graph neural network-based anomaly detection in multivariate time series[C]//Proceedings of the 35th AAAI Conference on Artificial Intelligence. Palo Alto:AAAI Press,2021:4027-4035.
[14] CHEN F,YUAN Z, HUANG Y,Multi-source data fusion for aspect-level sentiment classification[J]. Knowledge-Based Systems,2020,187:104831.
[15] BLÁZQUEZ-GARC#xCD;A A,CONDE A.,MORI U,et al. A review on outlier/anomaly detection in time series data[J]. ACM Computing Surveys,2021,54(3):1-33.
[16] WU H,XU J,WANG J,et al. Autoformer:decomposition transformers with auto-correlation for long-term series forecasting[C]//Proceedings of the 35th International Conference on Neural Information Processing Systems. Red Hook:Curran Associates Inc.,2021:22419-22430.
[17] 范杏蕊,李元诚. 基于改进Autoformer模型的短期电力负荷预测[J]. 电力自动化设备,2024,44(4):171-177.
[18] HAN K,XIAO A,WU E,et al. Transformer in Transformer[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE,2021:15908-15918.
[19] KATHAROPOULOS A,VYAS A,PAPPAS N,et al. Transformers are RNNs:fast autoregressive transformers with linear attention[C]//Proceedings of the 37th International Conference on Machine Learning. Cambridge:PMLR,2020:5156-5165.
[20] LIU Y,HU T,ZHANG H,et al. Itransformer:inverted transformers are effective for time series forecasting[C]//Proceedings of the 12th International Conference on Learning Representations. Washington DC: JCLR, 2024:632.
[21] HAN K,WANG Y,CHEN H,et al. A survey on vision transformer[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2023,45(1):87-110
[22] ZHOU T,MA Z,WEN Q,et al. FEDformer:Frequency Enhanced Decomposed Transformer for long-term series forecasting[C]//Proceedings of the 39th International Conference on Machine Learning. Cambridge:PMLR,2022:27268-27286.
[23] ZHOU H,ZHANG S,PENG J,et al. Informer:beyond efficient Transformer for long sequence time-series forecasting[C]//Proceedings of the 35th AAAI Conference on Artificial Intelligence. Palo Alto:AAAI Press,2021:11106-11115.
[24] 杨彬,马廷淮,黄学坚,等. 基于时空特征融合与序列重构的时间序列异常检测[J]. 计算机科学与探索,2025,19(9):2384-2398.
[25] AUDIBERT J,MICHIARDI P,GUYARD F,et al. USAD:Unsupervised Anomaly Detection on multivariate time series[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York:Association for Computing Machinery,2020:3395-3404.
[26] IRFAN M,SHAHRESTANI S,ELKHODR M. Machine learning in neurological disorders:a multivariate LSTM and AdaBoost approach to Alzheimer′s disease time series analysis[J]. Health Care Science,2024,3(1):41-52.

基于FEDformer-MTFA的轻工装备多源时序数据异常检测^*

Anomaly detection in multi-source time-series data of light industrial equipment based on FEDformer-MTFA

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	樊一, 刘斯麒, 沈洌政, 朱海平. 基于深度强化学习的模块化集成建造车间实时调度方法研究^*[J]. 现代制造工程, 2026, 548(5): 18-30.
[2]	卢进南, 何益佳. 融合STFT-SWT和2DCNN-SVM的小样本轴承故障分类诊断^*[J]. 现代制造工程, 2026, 546(3): 146-154.
[3]	余风军, 周晓平. 基于改进A^*-DWA算法的移动机器人避障技术[J]. 现代制造工程, 2026, 545(2): 30-34.
[4]	于子涵, 朱俊江, 李子枭. 基于GA-EF-XGBoost的铣削表面粗糙度预测^*[J]. 现代制造工程, 2026, 545(2): 111-116.
[5]	葛师语, 王玉芳, 张毅, 华晓麟. 考虑夹具的双资源约束柔性作业车间调度研究^*[J]. 现代制造工程, 2026, 544(1): 1-14.
[6]	张卫星, 宋树权, 于霜, 何春伟. 基于双流自适应网络的电机轴承故障诊断^*[J]. 现代制造工程, 2026, 544(1): 123-134.
[7]	俞峰, 徐勇, 范锐敏, 韩冰. 基于改进麻雀搜索算法的多目标柔性车间调度问题研究[J]. 现代制造工程, 2025, 543(12): 27-34.
[8]	郑国梁, 张朝阳, 吉卫喜, 于俊杰. 基于DQN的改进NSGA-Ⅱ求解多目标柔性作业车间调度问题[J]. 现代制造工程, 2025, 540(9): 1-11.
[9]	刘洋, 李申, 刘飞, 宋小刚. 基于DRMPSO的绳驱冗余机械臂逆运动学求解与运动规划[J]. 现代制造工程, 2025, 540(9): 53-59.
[10]	宋宇宁. 基于CloFormer-Bi-LSTM的液压泵故障诊断方法[J]. 现代制造工程, 2025, 539(8): 141-150.
[11]	夏良才, 陈仕军. 基于Dueling-DQN的协同进化算法求解绿色分布式异构柔性作业车间调度问题^*[J]. 现代制造工程, 2025, 538(7): 8-19.
[12]	张宁宁, 万卫兵, 张梦晓, 赵宇明. 面向多目标动态作业车间调度的强化学习决策算法研究^*[J]. 现代制造工程, 2025, 538(7): 20-30.
[13]	刘晋飞, 刘乙涵, 陈明, 黄华. 基于改进IGA的多品种变批量智能车间调度^*[J]. 现代制造工程, 2025, 535(4): 1-10.
[14]	杨丹, 舒先涛, 余震, 鲁光涛, 纪松霖, 王家兵. 深度强化学习求解动态柔性作业车间调度问题^*[J]. 现代制造工程, 2025, 533(2): 10-16.
[15]	周思聪, 向峰, 李红军, 左颖. 基于自编码器GAN数据增强的工业小目标缺陷检测^*[J]. 现代制造工程, 2025, 533(2): 101-108.