-
摘要: 复杂动态系统运行过程中的在线安全性评估至关重要且富有挑战性. 构建有效的数据驱动模型需要大量有标注数据, 但这在实际中通常难以获得. 此外, 考虑到系统不同的运行工况, 安全性评估模型应该具有良好的泛化能力. 域自适应(Domain adaptation, DA)可以将模型从数据标注丰富的源域迁移到具有不同但相似数据分布的目标域. 然而, 源域中没有出现过的任务相关未知情景会降低模型的性能, 是目前尚未解决的挑战. 主动域自适应通过结合域自适应与主动学习技术, 为解决上述挑战提供了思路. 本文研究目标域存在任务相关未知情景的主动域自适应安全性评估问题, 提出一种基于改进能量模型的主动域自适应方法. 在所提方法中融合分布外检测器, 在此基础上主动选择目标域中具有代表性的无标注样本进行标注, 作为训练数据以提高域自适应模型的性能. 最后, 通过基于轴承数据的案例研究, 验证所提方法的有效性和适用性.Abstract: Online safety assessment of complex dynamic systems during operation is paramount and challenging. A large amount of labeled data is necessary to construct an effective data-driven model, which is difficult to obtain in practice. Furthermore, the safety assessment model should have a good generalization ability given the varying operation modes. Domain adaptation (DA) can transfer the model trained on a source domain with abundant labeled data to a target domain that has a different but similar data distribution. However, the task-related unknown scenarios that have not appeared in the source domain will degrade the model performance, which remains an unsolved challenge at present. Active domain adaptation provides a potential solution to the aforementioned challenge by combining domain adaptation with active learning techniques. This paper investigates the problem of active domain adaptation for safety assessment, specifically addressing task-related unknown scenarios within the target domain. An active domain adaptation method with the improved energy-based model is proposed, and the out-of-distribution detector is incorporated in the proposed method. On this basis, representative unlabeled samples from the target domain are actively selected for annotation, which are then used as training data to enhance the performance of the domain adaptation model. At last, a case based on the bearing data is studied to demonstrate the effectiveness and applicability of the proposed method.
-
图 1 基于改进能量模型的主动域自适应安全性评估框架
Fig. 1 The safety assessment framework of the active domain adaptation with the improved energy-based model
图 3 凯斯西储大学滚动轴承故障测试平台
Fig. 3 Rolling bearing fault test platform from Case Western Reserve University
图 5 不同方法的安全性评估整体平均准确率结果对比
Fig. 5 Comparison of overall average accuracy results for safety assessment across different methods
表 1 不同情景及其对应的安全等级标签
Table 1 Different scenarios and corresponding safety level label
情景 安全等级标签 正常 1 IR-7 1 B-7 1 OR-7 1 IR-14 2 B-14 2 OR-14 2 IR-21 3 B-21 3 OR-21 3 注: $x\text{-}y$表示发生在位置$x$、故障直径为$y$的故障情景. 例如, IR-7表示发生在内圈、故障直径为7 mils的故障情景. 
下载: 导出CSV
表 2 不同方法在各种任务和不同未知情景下的安全性评估准确率$ ACC \uparrow (STD \downarrow) $比较
Table 2 The safety assessment accuracy $ ACC \uparrow (STD \downarrow )$ comparison of different methods under various tasks and different unknown scenarios
任务设定 未知情景 EDA EADA RAND 本文方法 (COPOD) 本文方法 (ECOD) 本文方法 (IForest) 本文方法 (KDE) 本文方法 (OCSVM) 本文方法 (LOF) IR 77.06 (1.73) 80.66 (3.90) 76.55 (2.54) 76.05 (1.32) 77.34 (1.02) 81.67 (1.95) 79.44 (4.04) 77.63 (4.42) 79.00 (4.50) 1hp $ \rightarrow $ 2hp B 83.62 (0.54) 81.96 (2.75) 83.62 (0.25) 83.04 (1.54) 81.10 (2.62) 82.47 (3.58) 81.82 (3.94) 81.39 (3.97) 82.76 (0.76) OR 82.40 (0.54) 82.32 (0.12) 80.30 (0.87) 80.88 (1.27) 81.17 (0.37) 81.67 (1.52) 81.02 (0.87) 79.94 (2.62) 82.11 (0.82) IR 74.86 (2.20) 72.13 (2.28) 75.36 (1.39) 74.50 (0.12) 74.21 (0.66) 75.36 (3.46) 74.64 (2.99) 77.37 (2.85) 74.14 (3.56) 1hp $ \rightarrow $ 3hp B 81.75 (4.38) 81.18 (3.48) 77.95 (1.11) 81.68 (1.41) 82.54 (1.35) 82.26 (0.90) 80.46 (1.86) 81.47 (3.26) 82.18 (0.90) OR 77.16 (1.71) 79.17 (0.82) 79.74 (2.75) 77.08 (3.99) 78.52 (3.14) 77.30 (0.69) 78.74 (1.59) 79.53 (3.17) 78.59 (1.59) IR 70.92 (1.32) 73.52 (1.19) 72.37 (2.05) 74.75 (1.44) 73.09 (1.02) 73.67 (1.62) 72.58 (1.30) 77.29 (2.06) 74.03 (0.57) 2hp $ \rightarrow $ 1hp B 84.49 (0.70) 83.98 (1.08) 83.62 (1.02) 84.20 (0.57) 84.85 (0.57) 83.91 (0.25) 83.91 (0.45) 84.20 (0.43) 84.49 (0.54) OR 79.15 (0.90) 79.22 (1.35) 79.65 (1.77) 79.15 (0.98) 79.22 (0.87) 79.43 (1.21) 79.37 (1.64) 78.35 (1.35) 79.29 (0.76) IR 75.22 (1.08) 73.35 (3.42) 73.99 (0.87) 75.07 (2.09) 76.08 (0.57) 76.29 (0.75) 77.01 (0.90) 75.86 (0.78) 74.93 (0.45) 2hp $ \rightarrow $ 3hp B 83.48 (1.73) 83.91 (1.08) 83.62 (0.22) 83.76 (0.12) 82.90 (1.32) 82.97 (3.02) 84.12 (0.69) 83.05 (1.40) 82.97 (1.20) OR 82.47 (1.65) 80.68 (1.79) 80.24 (2.24) 82.26 (0.66) 81.25 (0.78) 80.53 (1.74) 80.82 (1.71) 82.83 (0.82) 81.32 (1.53) IR 72.08 (2.13) 73.16 (0.99) 71.43 (0.94) 72.44 (0.62) 75.11 (2.81) 71.57 (0.66) 72.08 (2.34) 72.73 (0.43) 73.38 (0.99) 3hp $ \rightarrow $ 1hp B 83.04 (0.33) 84.63 (0.78) 83.98 (1.32) 84.13 (0.45) 84.56 (0.54) 85.93 (0.37) 85.35 (1.32) 84.34 (0.45) 84.70 (1.75) OR 72.29 (2.38) 71.00 (1.35) 69.77 (0.87) 73.02 (2.25) 71.28 (1.84) 70.09 (2.76) 74.96 (1.84) 74.24 (1.52) 70.63 (1.44) IR 76.70 (1.89) 76.26 (0.76) 77.34 (1.68) 76.84 (0.94) 76.41 (1.42) 78.14 (1.52) 76.41 (1.85) 75.61 (2.14) 77.20 (1.27) 3hp $ \rightarrow $ 2hp B 83.12 (1.52) 82.47 (1.72) 83.91 (1.54) 82.83 (0.25) 85.35 (1.19) 84.92 (1.74) 83.91 (1.52) 83.12 (1.21) 81.53 (0.82) OR 77.13 (4.25) 78.43 (3.50) 78.86 (0.45) 77.78 (1.09) 75.61 (2.67) 80.01 (2.53) 80.52 (1.69) 81.24 (0.66) 81.60 (0.43)
下载: 导出CSV
表 3 不同方法在不同未知情景下的安全性平均评估准确率$ \overline{ACC} \uparrow$比较
Table 3 The safety average assessment accuracy $ \overline{ACC} \uparrow$ comparison of different methods under different unknown scenarios
未知情景 EDA EADA RAND 本文方法(COPOD) 本文方法(ECOD) 本文方法(IForest) 本文方法(KDE) 本文方法(OCSVM) 本文方法(LOF) IR 74.47 74.85 74.51 74.94 75.37 76.12 75.36 75.25 75.45 B 83.25 83.02 82.78 83.27 83.55 83.74 83.26 82.93 83.11 OR 78.43 78.47 78.10 78.36 77.84 78.67 79.24 79.36 78.93
下载: 导出CSV
-
[1] Liu C, He X, Zhou D H, Huang B. Safety assessment for dynamic systems: A survey. Cybernetics and Intelligence, DOI: 10.26599/CAI.2024.9390001 [2] Wang M, Zhou D H, Chen M Y. Hybrid variable monitoring: An unsupervised process monitoring framework with binary and continuous variables. Automatica, 2023, 147: Article No. 110670 doi: 10.1016/j.automatica.2022.110670 [3] Zhang H J, Zhang C, Dong J, Peng K X. A new key performance indicator oriented industrial process monitoring and operating performance assessment method based on improved Hessian locally linear embedding. International Journal of Systems Science, 2022, 53(16): 3538−3555 doi: 10.1080/00207721.2022.2093420 [4] Song P Y, Zhao C H, Huang B. SFNet: A slow feature extraction network for parallel linear and nonlinear dynamic process monitoring. Neurocomputing, 2022, 488: 359−380 doi: 10.1016/j.neucom.2022.03.012 [5] 刘强, 卓洁, 郎自强, 秦泗钊. 数据驱动的工业过程运行监控与自优化研究展望. 自动化学报, 2018, 44(11): 1944−1956Liu Qiang, Zhuo Jie, Lang Zi-Qiang, Qin S. Joe. Perspectives on data-driven operation monitoring and self-optimization of industrial processes. Acta Automatica Sinica, 2018, 44(11): 1944−1956 [6] Zhang Z, He X. Active fault diagnosis for linear systems: Within a signal processing framework. IEEE Transactions on Instrumentation and Measurement, 2022, 71: Article No. 3505009 [7] Amini N, Zhu Q Q. Fault detection and diagnosis with a novel source-aware autoencoder and deep residual neural network. Neurocomputing, 2022, 488: 618−633 doi: 10.1016/j.neucom.2021.11.067 [8] Xu J M, Ke H B, Chen Z W, Fan X Y, Peng T, Yang C H. Oversmoothing relief graph convolutional network-based fault diagnosis method with application to the rectifier of high-speed trains. IEEE Transactions on Industrial Informatics, 2022, 19(1): 771−779 [9] Shakiba F M, Shojaee M, Azizi S M, Zhou M C. Real-time sensing and fault diagnosis for transmission lines. International Journal of Network Dynamics and Intelligence, 2022, 1(1): 36−47 [10] 彭开香, 马亮, 张凯. 复杂工业过程质量相关的故障检测与诊断技术综述. 自动化学报, 2017, 43(3): 349−365Peng Kai-Xiang, Ma Liang, Zhang Kai. Review of quality-related fault detection and diagnosis techniques for complex industrial processes. Acta Automatica Sinica, 2017, 43(3): 349−365 [11] Gao C, He X, Dong H L, Liu H J, Lyu G R. A survey on fault-tolerant consensus control of multi-agent systems: Trends, methodologies and prospects. International Journal of Systems Science, 2022, 53(13): 2800−2813 doi: 10.1080/00207721.2022.2056772 [12] Jia F L, Cao F F, He X. Adaptive fault-tolerant tracking control for uncertain nonlinear systems with unknown control directions and limited resolution. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2022, 53(3): 1813−1825 [13] Cai M, He X, Zhou D H. An active fault tolerance framework for uncertain nonlinear high-order fully-actuated systems. Automatic, 2023, 152: Article No. 110969 doi: 10.1016/j.automatica.2023.110969 [14] Liu X Q, Chen M Y, Sheng L, Zhou D H. Adaptive fault-tolerant control for nonlinear high-order fully-actuated systems. Neurocomputing, 2022, 495: 75−85 doi: 10.1016/j.neucom.2022.04.129 [15] 柴毅, 毛万标, 任浩, 屈剑锋, 尹宏鹏, 杨志敏, 等. 航天发射系统运行安全性评估研究进展与挑战. 自动化学报, 2019, 45(10): 1829−1845Chai Yi, Mao Wan-Biao, Ren Hao, Qu Jian-Feng, Yin Hong-Peng, Yang Zhi-Min, et al. Research on operational safety assessment for spacecraft launch system: Progress and challenges. Acta Automatica Sinica, 2019, 45(10): 1829−1845 [16] Serbanescu D, Ulmeanu A P. Selected Topics in Probabilistic Safety Assessment: Methodology and Practice in Nuclear Power Plants. Switzerland: Springer Nature, 2020. 1−9 [17] Aldemir T. A survey of dynamic methodologies for probabilistic safety assessment of nuclear power plants. Annals of Nuclear Energy, 2013, 52: 113−124 doi: 10.1016/j.anucene.2012.08.001 [18] 赵福均, 周志杰, 胡昌华, 常雷雷, 王力. 基于证据推理的动态系统安全性在线评估方法. 自动化学报, 2017, 43(11): 1950−1961Zhao Fu-Jun, Zhou Zhi-Jie, Hu Chang-Hua, Chang Lei-Lei, Wang Li. Online safety assessment method based on evidential reasoning for dynamic systems. Acta Automatica Sinica, 2017, 43(11): 1950−1961 [19] Liu Z Y, Deng Y, Zhang Y, Ding Z J, He X. Safety assessment of dynamic systems: An evidential group interaction-based fusion design. IEEE Transactions on Instrumentation and Measurement, 2021, 70: Article No. 3523014 [20] Liu C, Zhang Y, He X. Expert-augmented data-driven safety level assessment scheme with incremental learning. In: Proceedings of the 12th CAA Symposium on Fault Detection, Supervision, and Safety for Technical Processes (SAFEPROCESS). Chengdu, China: IEEE, 2021. 1−6 [21] Wilson G, Cook D J. A survey of unsupervised deep domain adaptation. ACM Transactions on Intelligent Systems and Technology, 2020, 11(5): Article No. 51 [22] Zhao Z B, Zhang Q Y, Yu X L, Sun C, Wang S B, Yan R Q, et al. Applications of unsupervised deep transfer learning to intelligent fault diagnosis: A survey and comparative study. IEEE Transactions on Instrumentation and Measurement, 2021, 70: Article No. 3525828 [23] An Y Y, Zhang K, Chai Y, Liu Q, Huang X H. Domain adaptation network base on contrastive learning for bearings fault diagnosis under variable working conditions. Expert Systems With Applications, 2023, 212: Article No. 118802 doi: 10.1016/j.eswa.2022.118802 [24] Chen Z Y, Liao Y X, Li J P, Huang R Y, Xu L, Jin G, et al. A multi-source weighted deep transfer network for open-set fault diagnosis of rotary machinery. IEEE Transactions on Cybernetics, 2022, 53(3): 1982−1993 [25] Yang B, Xu S C, Lei Y G, Lee C G, Stewart E, Roberts C. Multi-source transfer learning network to complement knowledge for intelligent diagnosis of machines with unseen faults. Mechanical Systems and Signal Processing, 2022, 162: Article No. 108095 doi: 10.1016/j.ymssp.2021.108095 [26] Zhu J, Huang C G, Shen C Q, Shen Y J. Cross-domain open-set machinery fault diagnosis based on adversarial network with multiple auxiliary classifiers. IEEE Transactions on Industrial Informatics, 2022, 18(11): 8077−8086 doi: 10.1109/TII.2021.3138558 [27] Zhang W, Li X, Ma H, Luo Z, Li X. Open-set domain adaptation in machinery fault diagnostics using instance-level weighted adversarial learning. IEEE Transactions on Industrial Informatics, 2021, 17(11): 7445−7455 doi: 10.1109/TII.2021.3054651 [28] Zhao C, Shen W. Adaptive open set domain generalization network: Learning to diagnose unknown faults under unknown working conditions. Reliability Engineering & System Safety, 2022, 226: Article No. 108672 [29] Liu C, Zhang Y, Ding Z J, He X. Active incremental learning for health state assessment of dynamic systems with unknown scenarios. IEEE Transactions on Industrial Informatics, 2022, 19(2): 1863−1873 [30] Fu B, Cao Z J, Wang J M, Long M S. Transferable query selection for active domain adaptation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, USA: IEEE, 2021. 7268−7277 [31] Prabhu V, Chandrasekaran A, Saenko K, Hoffman J. Active domain adaptation via clustering uncertainty-weighted embeddings. In: Proceedings of the 18th IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, Canada: IEEE, 2021. 8485−8494 [32] Xie B H, Yuan L H, Li S, Liu C H, Cheng X J, Wang G R. Active learning for domain adaptation: An energy-based approach. In: Proceedings of the 36th AAAI Conference on Artificial Intelligence. Vancouver, Canada: AAAI, 2022. 8708−8716 [33] LeCun Y, Chopra S, Hadsell R, Ranzato M, Huang F J. A tutorial on energy-based learning [Online], available: https://yann.lecun.com/exdb/publis/pdf/lecun-06.pdf, September 8, 2024 [34] Liu W T, Wang X Y, Owens J D, Li Y X. Energy-based out-of-distribution detection. arXiv preprint arXiv: 2010.03759, 2021. [35] Li Z, Zhao Y, Botta N, Ionescu C, Hu X Y. COPOD: Copula-based outlier detection. In: Proceedings of the IEEE International Conference on Data Mining. Sorrento, Italy: IEEE, 2020. 1118−1123 [36] Li Z, Zhao Y, Hu X Y, Botta N, Ionescu C, Chen G. Ecod: Unsupervised outlier detection using empirical cumulative distribution functions. IEEE Transactions on Knowledge and Data Engineering, 2022, 35(12): 12181−12193 [37] Liu F T, Ting K M, Zhou Z H. Isolation forest. In: Proceedings of the 8th IEEE International Conference on Data Mining. Pisa, Italy: IEEE, 2008. 413−422 [38] Desforges M J, Jacob P J, Cooper J E. Applications of probability density estimation to the detection of abnormal conditions in engineering. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, 1998, 212(8): 687−703 doi: 10.1243/0954406981521448 [39] Scholkopf B, Williamson R C, Smola A, Shawe-Taylor J, Platt J. Support vector method for novelty detection. In: Proceedings of the 12th International Conference on Neural Information Processing Systems. Denver, USA: MIT Press, 1999. 582−588 [40] Breunig M M, Kriegel H P, Ng R T, Sander J. LOF: Identifying density-based local outliers. In: Proceedings of the ACM SIGMOD International Conference on Management of Data. Dallas, USA: ACM, 2000. 93−104 -
图(5) / 表(3)
计量
- 文章访问数: 321
- HTML全文浏览量: 112
- PDF下载量: 108
- 被引次数: 0
