Graph Neural Network (GNN) has been widely adopted for non-Euclidean industrial data modeling. However, traditional methods relying on global adjacency matrices often fail to capture localized topological features of individual samples. Moreover, their offline operation neglects valuable information from newly collected data, limiting diagnostic accuracy. To address these challenges, this paper proposes an incremental graph learning framework for fault diagnosis based on horizontal visibility adjacency matrices (HVAMs). First, the visibility criterion is applied to preserve intrinsic fluctuation patterns of raw signals, generating HVAMs that precisely encode local topological structures. Second, a Q-learning-based incremental learning framework is designed, integrating category-aware elastic loss weighting to dynamically select high-value incremental samples. This strategy balances class distributions and optimizes training sequences. Finally, a collaborative update mechanism is established, combining gradient sensitivity propagation with topology-preserving regularization, which enhances model adaptability to online scenarios. Experiments on the TE benchmark and coal mill datasets demonstrate that our method significantly improves incremental learning performance with streaming data, achieving higher diagnostic accuracy. The results validate its robustness in complex industrial environments.
CHENZ Y, WUJ, DENGC, et al. Deep attention relation network: A zero-shot learning method for bearing fault diagnosis under unknown domains[J]. IEEE Transactions on Reliability, 2023, 72(1): 79-89. DOI: 10.1109/TR.2022.3177930 .
[2]
FRANKP M. Fault diagnosis in dynamic systems using analytical and knowledge-based redundancy—A survey and some new results[J]. Automatica (Journal of IFAC), 1990, 26(3): 459-474. DOI: 10.1016/0005-1098(90)90018-D .
TIANY, ZHUZ, TIANJ D, et al. Parameters optimization of electric vehicles dynamic wireless power transfer system based on LCC-S compensation topology[J]. Journal of Mechanical Engineering, 2021, 57(14): 150-159 (Ch). DOI: 10.3901/jme.2021.14.150 .
[5]
WUH, ZHAOJ S. Deep convolutional neural network model based chemical process fault diagnosis[J]. Computers & Chemical Engineering, 2018, 115: 185-197. DOI: 10.1016/j.compchemeng.2018.04.009 .
GAMAF, BRUNAJ, RIBEIROA. Stability properties of graph neural networks[J]. IEEE Transactions on Signal Processing, 2020, 68: 5680-5695. DOI: 10.1109/TSP.2020.3026980 .
[8]
CHENQ, DONGX J, TUG W, et al. TFN: An interpretable neural network with time-frequency transform embedded for intelligent fault diagnosis[J]. Mechanical Systems and Signal Processing, 2024, 207: 110952. DOI: 10.1016/j.ymssp.2023.110952 .
[9]
HUZ X, JIANGP. An imbalance modified deep neural network with dynamical incremental learning for chemical fault diagnosis[J]. IEEE Transactions on Industrial Electronics, 2019, 66(1): 540-550. DOI: 10.1109/TIE.2018.2798633 .
XIAOY, WANGQ F, YANGZ, et al. Research on incipient warning and diagnosis method of sudden unbalance fault for rotating machinery[J]. Journal of Mechanical Engineering, 2023, 59(11): 308-318 (Ch). DOI: 10.3901/jme.2023.11.308 .
DOWNSJ J, VOGELE F. A plant-wide industrial process control problem[J]. Computers & chemical engineering, 1993, 17(3): 245-255. DOI: doi.org/10.1016/0098-1354(93)80018-I .
[16]
LUQUEB, LACASAL, BALLESTEROSF, et al. Horizontal visibility graphs: Exact results for random time series[J]. Physical Review E—Statistical, Nonlinear, and Soft Matter Physics, 2009, 80(4): 046103. DOI: https://doi.org/10.1103/PhysRevE.80.046103 .
BRODYS, ALONU, YAHAVE. How attentive are graph attention networks [C]//Proceedings of International Conference on Learning Representations. 2022: 21659-21685. DOI: arxiv.org/abs/2105.14491 .
[20]
DEFFERRARDM, BRESSONX, VANDERGHEYNSTP. Convolutional neural networks on graphs with fast localized spectral filtering[C]//Proceedings of the 30th International Conference on Neural Information Processing Systems. 2016: 3844-3852. DOI: 10.5555/3157382.3157527 .
[21]
MORRISC, RITZERTM, FEY M, et al. Weisfeiler and Leman go neural: Higher-order graph neural networks[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33(1): 4602-4609. DOI: 10.1609/aaai.v33i01.33014602 .