基于多分支网络结构搜索的工业过程故障诊断

Fault Diagnosis in Industrial Process Based on Multi-way Network Architecture Search

摘要

工业过程中常用神经网络进行故障诊断，然而神经网络的设计工作往往需要大量专家经验，并且随着样本量的增加，网络测试和超参数调整更加耗时耗力。针对该问题，本文提出了基于多分支网络结构搜索的故障诊断方法：首先，设计控制器自动生成以图卷积网络(GCN) 为基础单元的神经网络结构；其次，利用权重共享机制大大减少网络测试时间；此外，采用了 ε -贪婪策略，以避免控制器陷入局部最优；最后，在田纳西伊斯曼（TE）过程仿真上验证了方法的有效性，与其他经典方法相比，本文方法的平均准确率提升 3.85 个百分点以上，搜索速度提升了 71.2% 以上。

Abstract

In industrial processes, machine learning methods based on neural networks are commonly used for fault diagnosis. However, such methods suffer from poor transferability and usually require the design of distinct neural networks for different tasks. The optimization of network model structures and parameters mainly relies on repeated trial and error. This process not only depends on expert knowledge and experience but also consumes substantial time. As model depth increases, the difficulty of model design and the workload of manual parameter tuning multiply dramatically. Unlike manually designed neural networks, Neural Architecture Search (NAS) can automatically design and generate network structures, which reduces the reliance on expert experience and knowledge to a certain extent, simplifies manual optimization procedures, and cuts down the human and time costs incurred by structural design and parameter adjustment. Nevertheless, networks generated by conventional NAS methods usually have a single branch with a fixed structure. To address this issue, this paper proposes a fault diagnosis method based on multi-path neural architecture search. Firstly, a controller is designed to automatically generate neural network structures with graph convolutional networks as the basic unit. Secondly, a weight-sharing mechanism is adopted to substantially reduce network evaluation time. Furthermore, the ε-greedy strategy is introduced to prevent the controller from falling into local optima. Finally, the effectiveness of the proposed method is verified on the Tennessee Eastman (TE) process simulation platform. Compared with other classical methods, the proposed method achieves an average accuracy improvement of over 3.85 percentage points and a search speed improvement of more than 71.2%.

关键词

过程控制 / 故障诊断 / 神经网络 / 网络结构搜索 / 田纳西伊斯曼过程

Key words

process control / fault diagnosis / neural network / network architecture search / TE process

参考文献

[1] XU S. A survey of knowledge-based intelligent fault diagnosis techniques[C]//Journal of Physics: Conference Series. London, UK: IOP Publishing, 2019, 1187(3):032006.

[2] 俞文光, 娄舜杰, 王得磊, 等. 流程工业中工业大数据融合应用创新研究 [J].智能制造, 2024(2):21-27.

[3] 杨彪, 熊贇, 傅玲, 等. 工业数字化转型: 故障诊断方法研究进展 [J].大数据, 2024, 10(1):110-126.

[4] AMUTAK E A, FENDRI M, BETTI F, et al. Unlocking value from artificial intelligence in manufacturing[C]//World Economic Forum. Geneva, Switzerland: World Economic Forum, 2022.

[5] DOU W, LIU B, LIN C, et al. Architecture of virtual edge data center with intelligent metadata service of a geo-distributed file system[J].Journal of Systems Architecture, 2022, 128:102545.

[6] ZHENG X, CAI Z. Privacy-preserved data sharing towards multiple parties in industrial IoTs[J].IEEE Journal on Selected Areas in Communications, 2020, 38(5):968-979.

[7] AMRI O, FRI M, MSAFF M, et al. Industrial systems Fault's diagnosis approaches adapted to the context of the smart industry[C]//2023 3rd International Conference on Innovative Research in Applied Science, Engineering and Technology (IRASET). New York, USA: IEEE, 2023:1-5.

[8] 任浩, 屈剑锋, 柴毅, 等. 深度学习在故障诊断领域中的研究现状与挑战 [J].控制与决策, 2017, 32(8):1345-1358.

[9] LEITE D, ANDRADE E, RATIVA D, et al. Fault detection and diagnosis in industry 4.0: A review on challenges and opportunities[J].Sensors (Basel, Switzerland), 2024, 25(1):60.

[10] WANG Z, LIN T, YAO L, et al. A novel data-driven fault diagnosis method based on VMD-RCMFE-DDMA-BASS-VM model for rolling bearings[C]//2021 6th International Conference on Intelligent Computing and Signal Processing (ICSP). New York, USA: IEEE, 2021:171-176.

[11] 易瓅, 侍洪波, 宋冰, 等. 基于特征加权的化工过程中未见模式的故障诊断 [J].华东理工大学学报 (自然科学版), 2025, 51(5):693-703.

[12] 易永率, 赵海涛. 基于属性描述的多单元过程零样本故障诊断 [J].华东理工大学学报 (自然科学版), 2023, 49(6):845-853.

[13] MA L, SHI F, WU Z, et al. A survey of few-shot learning-based compound fault diagnosis methods for industrial processes[C]//2023 IEEE 6th International Conference on Industrial Cyber-Physical Systems (ICPS). New York, USA: IEEE, 2023:1-4.

[14] 陆宽. 基于数据驱动的化工过程故障诊断方法研究与应用 [D].广州: 广东工业大学, 2023.

[15] LI Z, HAN Q, YANG R, et al. Fault diagnosis for rotating machinery gearbox based on 1DCNN-RF[C]//2020 13th International Symposium on Computational Intelligence and Design (ISCID). New York, USA: IEEE, 2020:376-379.

[16] XIA Y, LI S. Missing data bearing fault diagnosis based on SVT-1DCNN[C]//2022 IEEE 5th International Conference on Electronic Information and Communication Technology (ICEICT). New York, USA: IEEE, 2022:695-700.

[17] WU Z, ZHAO J. Open-circuit fault diagnosis for grid-tied T-type inverters using an LSTM autoencoder[C]//2023 China Automation Congress (CAC). New York, USA: IEEE, 2023:3334-3339.

[18] ZHU D, SONG X, YANG J, et al. A bearing fault diagnosis method based on L1 regularization transfer learning and LSTM deep learning[C]//2021 IEEE International Conference on Information Communication and Software Engineering (ICICSE). New York, USA: IEEE, 2021:308-312.

[19] SUN Y, TIAN J, NIE X, et al. Multi-sensor based graph convolution fault diagnosis method[C]//2024 36th Chinese Control and Decision Conference (CCDC). New York, USA: IEEE, 2024:3941-3946.

[20] WANG S, JING B, PAN J, et al. Coupling fault diagnosis based on graph neural network[C]//2023 CAA Symposium on Fault Detection, Supervision and Safety for Technical Processes (SAFEPROCESS). New York, USA: IEEE, 2023:1-5.

[21] LIU Y, WANG T, ZHENG X, et al. A neural architecture search method with multi-dimensional correlation representation for fault diagnosis[C]//2023 IEEE Smart World Congress (SWC). New York, USA: IEEE, 2023:1-8.

[22] PAN W, ZHAO F, SHEN G, et al. Brain-inspired multi-scale evolutionary neural architecture search for deep spiking neural networks[J/OL].IEEE Transactions on Evolutionary Computation, (2024-11-27). https://doi.org/10.1109/TEVC.2024.3507812.

[23] PENG Y, SONG A, CIESIELSKI V, et al. PRE-NAS: Evolutionary neural architecture search with predictor[J].IEEE Transactions on Evolutionary Computation, 2022, 27(1):26-36.

[24] WEN Y W, PENG S H, TING C K. Two-stage evolutionary neural architecture search for transfer learning[J].IEEE Transactions on Evolutionary Computation, 2021, 25(5):928-940.

[25] XUE Y, TONG W, NERI F, et al. Evolutionary architecture search for generative adversarial networks based on weight sharing[J].IEEE Transactions on Evolutionary Computation, 2023, 28(3):653-667.

[26] LIAO J, SHEN M. RL-based scheduling and placement for deep learning jobs on large-scale GPU clusters[C]//2024 International Conference on Networking, Architecture and Storage (NAS). New York, USA: IEEE, 2024:1-4.

[27] FAN Y, TANG X, ZHOU G, et al. EfficientAutoGAN: Predicting the rewards in reinforcement-based neural architecture search for generative adversarial networks[J].IEEE Transactions on Cognitive and Developmental Systems, 2020, 14(1):234-245.

[28] SHI J, FAN Y, ZHOU G, et al. Distributed gan: Toward a faster reinforcement-learning-based architecture search[J].IEEE Transactions on Artificial Intelligence, 2021, 3(3):391-401.

[29] LYU B, WEN S, SHI K, et al. Multiobjective reinforcement learning-based neural architecture search for efficient portrait parsing[J].IEEE Transactions on Cybernetics, 2021, 53(2):1158-1169.

[30] ALI S, WANI M A. Gradient-based neural architecture search: A comprehensive evaluation[J].Machine Learning and Knowledge Extraction, 2023, 5(3):1176-1194.

[31] CAI H, ZHU L, HAN S. Proxylessnas: Direct neural architecture search on target task and hardware[C]//International Conference on Learning Representations. New Orleans, Louisiana, USA: OpenReview.net, 2019.

[32] SHU Y, DAI Z, WU Z, et al. Unifying and boosting gradient-based training-free neural architecture search[J].Advances in Neural Information Processing Systems, 2022, 35:33001-33015.

[33] SHI X, ZHOU P, CHEN W, et al. Efficient gradient-based neural architecture search for end-to-end ASR[C]//Companion Publication of the 2021 International Conference on Multimodal Interaction. New York, USA: Association for Computing Machinery, 2021:91-96.

[34] FLOREANO D, DÜRR P, MATTIUSSI C. Neuroevolution: From architectures to learning[J].Evolutionary Intelligence, 2008, 1(1):47-62.

[35] ZOPH B, LE Q V. Neural architecture search with reinforcement learning[C]//5th International Conference on Learning Representations. Toulon, France: OpenReview.net, 2016.

[36] BAKER B, GUPTA O, NAIK N, et al. Designing neural network architectures using reinforcement learning[EB/OL]. (2016-11-01). https://arXiv preprint arXiv: 1611.02167, 2016.

[37] LIU H, SIMONYAN K, YANG Y. Darts: Differentiable architecture search[C]//7th International Conference on Learning Representations. New Orleans, Louisiana, USA: OpenReview.net, 2019.

[38] WANG R, JIANG H, LI X, et al. A reinforcement neural architecture search method for rolling bearing fault diagnosis[J].Measurement, 2020, 154:107417.

[39] LI X, ZHENG J, LI M, et al. One-shot neural architecture search for fault diagnosis using vibration signals[J].Expert Systems with Applications, 2022, 190:116027.

[40] CAO J, MA J, HUANG D, et al. Finding the optimal multilayer network structure through reinforcement learning in fault diagnosis[J].Measurement, 2022, 188:110377.

[41] LI H, WU G, ZHENG W S. Combined depth space based architecture search for person re-identification[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York, USA: IEEE, 2021:6729-6738.

[42] FANG Y, LU H, XIAO D, et al. Attention-boosted path exploration neural architecture search for fault diagnosis of rotating machinery[J].IEEE Sensors Journal, 2024, 25(4):6634-6645.

[43] QIN Y, WANG X, CUI P, et al. Gqnas: Graph q network for neural architecture search[C]//2021 IEEE International Conference on Data Mining (ICDM). New York, USA: IEEE, 2021:1288-1293.

基金资助

国家自然科学基金项目(62473154，62473155，62473156)