pith. sign in

arxiv: 2602.03004 · v2 · pith:O64VDDRQnew · submitted 2026-02-03 · 💻 cs.LG · cs.AI

Graph Autoencoder for Process Monitoring

Xiangrui Zhang This is my paper

Pith reviewed 2026-05-21 14:47 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords causal graph learninggraph autoencoderprocess monitoringfault detectionspatial self-attentionGCLSTMTennessee Eastman process
0
0 comments X

The pith

A causal graph spatial-temporal autoencoder learns invariant structures from dynamic correlations to monitor industrial processes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper introduces the Causal Graph Spatial-Temporal Autoencoder (CGSTAE) that first applies spatial self-attention to capture changing correlations among process variables, then uses a three-step algorithm to extract an invariant causal graph by reversing the causal invariance principle. The resulting graph feeds into a graph convolutional LSTM encoder-decoder that reconstructs the time-series data in a sequence-to-sequence setup. Fault detection relies on two monitoring statistics computed in the learned feature space and the residual space. A sympathetic reader would care because conventional methods often lose interpretability when dealing with high-dimensional, time-varying industrial data, and an explicit causal graph could make both detection and diagnosis more transparent.

Core claim

The paper establishes that combining a spatial self-attention module for learning correlation graphs with a novel three-step causal graph structure learning algorithm, which leverages a reverse perspective of the causal invariance principle, allows derivation of a stable causal graph; when this graph is embedded in a GCLSTM-based autoencoder, the resulting CGSTAE supports effective process monitoring and fault detection through statistics in feature and residual spaces, as demonstrated on the Tennessee Eastman process and a real-world air separation process.

What carries the argument

The three-step causal graph structure learning algorithm that derives an invariant causal graph from varying correlation graphs produced by the spatial self-attention module.

If this is right

  • Fault detection becomes possible through separate statistics computed in the feature space and the residual space.
  • The learned causal graph supplies interpretable structure for understanding variable relationships during normal and faulty operation.
  • The approach applies successfully to both the Tennessee Eastman benchmark and a real air separation plant.
  • Reconstruction error in the sequence-to-sequence GCLSTM framework serves as a sensitive indicator of process deviations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the invariant causal graph holds across operating modes, the same structure could support root-cause isolation of detected faults rather than mere detection.
  • The method might transfer to other multivariate time-series domains such as sensor networks or financial monitoring where causal stability is assumed.
  • Online updating of the correlation graphs could allow the model to track slow drifts in process causality without full retraining.

Load-bearing premise

The three-step causal graph structure learning algorithm can reliably uncover an invariant causal graph from the varying correlation graphs generated by the spatial self-attention module.

What would settle it

A direct comparison showing that CGSTAE misses known faults in the Tennessee Eastman process dataset at rates no better than standard PCA or autoencoder baselines would falsify the claim of effective monitoring.

Figures

Figures reproduced from arXiv: 2602.03004 by XiangRui Zhang.

Figure 2
Figure 2. Figure 2: GCLSTM unit of the spatial-temporal encoder. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Flowchart of the CGSTAE-based process monitoring. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Process flow diagram of the TEP. The public dataset of the TEP can be download from the website 1 . We select all 52 variables for process monitoring. We use the 960 samples from normal operating conditions as the training set, and 21 fault operating conditions as the testing sets. Each testing set also has 960 samples, and the fault is introduced from the 161st sample. For data reorganization, we set the … view at source ↗
Figure 7
Figure 7. Figure 7: Graph structures of the TEP. (a) Prior causal graph (b) GAE-I (c) [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: SPE statistic of different methods for fault 11 of TEP. (a) AE [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Process flow diagram of the ASP argon distillation system. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Training data visualization. (a) AIA704 (b) AI705. [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: T2 statistic of different methods for ASP monitoring. (a) AE (b) LSTM-AE (c) GAE-I (d) GAE-II (e) DGSTAE (f) CGSTAE. The residual￾based fault detection methods KDGCN and KG-GCBiGCN do not have T2 statistic. threshold of 0.1 is applied to truncate the learned causal graph, resulting in a discrete causal graph. Then, an optimal subgraph is found on the discrete causal graph that includes all fault variables… view at source ↗
Figure 13
Figure 13. Figure 13: Graph structures of the ASP argon distillation system. (a) Prior [PITH_FULL_IMAGE:figures/full_fig_p011_13.png] view at source ↗
Figure 11
Figure 11. Figure 11: SPE statistic of different methods for ASP monitoring. (a) AE [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Fault diagnosis results of the nitrogen blockage fault. (a) Variable [PITH_FULL_IMAGE:figures/full_fig_p011_12.png] view at source ↗
Figure 14
Figure 14. Figure 14: Sensitivity analysis of the balancing hyperparameters. (a) [PITH_FULL_IMAGE:figures/full_fig_p012_14.png] view at source ↗
read the original abstract

To improve the reliability and interpretability of industrial process monitoring, this article proposes a Causal Graph Spatial-Temporal Autoencoder (CGSTAE). The network architecture of CGSTAE combines two components: a correlation graph structure learning module based on spatial self-attention mechanism (SSAM) and a spatial-temporal encoder-decoder module utilizing graph convolutional long-short term memory (GCLSTM). The SSAM learns correlation graphs by capturing dynamic relationships between variables, while a novel three-step causal graph structure learning algorithm is introduced to derive a causal graph from these correlation graphs. The algorithm leverages a reverse perspective of causal invariance principle to uncover the invariant causal graph from varying correlations. The spatial-temporal encoder-decoder, built with GCLSTM units, reconstructs time-series process data within a sequence-to-sequence framework. The proposed CGSTAE enables effective process monitoring and fault detection through two statistics in the feature space and residual space. Finally, we validate the effectiveness of CGSTAE in process monitoring through the Tennessee Eastman process and a real-world air separation process.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes a Causal Graph Spatial-Temporal Autoencoder (CGSTAE) for industrial process monitoring. It combines a spatial self-attention mechanism (SSAM) to learn dynamic correlation graphs between process variables, a three-step causal graph structure learning algorithm that applies a reverse perspective of the causal invariance principle to extract an invariant causal graph from those varying correlations, and a GCLSTM-based spatial-temporal encoder-decoder for sequence-to-sequence reconstruction. Monitoring and fault detection rely on two statistics computed in feature space and residual space, with validation reported on the Tennessee Eastman process and a real-world air separation process.

Significance. If the three-step causal extraction procedure can be shown to recover stable, interpretable structures that measurably improve detection over non-causal baselines, the approach would offer a useful integration of causal graph learning with spatio-temporal autoencoders for process monitoring applications.

major comments (2)
  1. [Abstract (method description)] The headline claim that CGSTAE enables effective monitoring rests on the three-step causal graph algorithm correctly extracting an invariant causal graph from the dynamic correlation graphs produced by SSAM. The abstract invokes the reverse causal-invariance perspective but supplies neither a formal justification, synthetic graph-recovery experiments on known ground-truth structures, nor an ablation that isolates the contribution of the extracted causal graph versus a plain correlation graph or GCLSTM baseline.
  2. [Experimental validation] Validation on the Tennessee Eastman and air-separation processes is presented as demonstrating effectiveness, yet the reported results contain no error bars, no statistical significance tests, and no quantitative comparison showing that the causal-graph step improves fault-detection metrics over the non-causal GCLSTM component alone.
minor comments (1)
  1. [Monitoring statistics] Define the precise formulas and control limits for the two monitoring statistics in feature and residual space; the current description leaves their computation and threshold selection ambiguous.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments on our manuscript. We address each major comment below and outline the specific revisions we will make to improve clarity, rigor, and empirical support.

read point-by-point responses

  1. Referee: [Abstract (method description)] The headline claim that CGSTAE enables effective monitoring rests on the three-step causal graph algorithm correctly extracting an invariant causal graph from the dynamic correlation graphs produced by SSAM. The abstract invokes the reverse causal-invariance perspective but supplies neither a formal justification, synthetic graph-recovery experiments on known ground-truth structures, nor an ablation that isolates the contribution of the extracted causal graph versus a plain correlation graph or GCLSTM baseline.

    Authors: We agree that the abstract is concise and that additional justification and experiments would strengthen the presentation. The full manuscript (Section 3.2) details the three-step algorithm, which applies a reverse perspective of the causal invariance principle to identify structures that remain stable across the varying correlation graphs produced by SSAM. To address the comment directly, we will revise the abstract and add a short formal justification paragraph in the introduction. We will also include new synthetic experiments that recover known ground-truth causal graphs and an ablation study isolating the causal extraction step against plain correlation-graph and GCLSTM baselines. These additions will appear in the revised version. revision: yes

  2. Referee: [Experimental validation] Validation on the Tennessee Eastman and air-separation processes is presented as demonstrating effectiveness, yet the reported results contain no error bars, no statistical significance tests, and no quantitative comparison showing that the causal-graph step improves fault-detection metrics over the non-causal GCLSTM component alone.

    Authors: We acknowledge that the current experimental section would benefit from greater statistical rigor. The reported results on the Tennessee Eastman and air-separation datasets demonstrate improved monitoring performance, yet we agree that error bars, significance testing, and explicit comparisons are needed to quantify the contribution of the causal-graph component. In the revision we will rerun all experiments across multiple random seeds, report means with standard deviations, apply paired statistical tests, and add a dedicated ablation table that directly compares the full CGSTAE against a non-causal GCLSTM baseline and a correlation-only variant. These changes will be incorporated into the revised manuscript. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper proposes CGSTAE as a new architecture combining SSAM for dynamic correlation graphs and a novel three-step causal graph algorithm that applies a reverse view of the causal invariance principle to extract an invariant graph. This graph then feeds into a GCLSTM-based spatial-temporal autoencoder whose reconstruction yields feature-space and residual-space monitoring statistics. The derivation chain is self-contained: the algorithm is introduced as an original contribution rather than derived from prior fitted quantities or self-citations, and performance is assessed on external standard benchmarks (Tennessee Eastman and air-separation data) without reducing any claimed statistic or graph to a tautological renaming of the training inputs. No self-definitional, fitted-input-as-prediction, or load-bearing self-citation steps are present in the abstract or described method.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the unproven ability of the three-step algorithm to extract invariant causal structure and on the assumption that GCLSTM reconstruction errors reliably indicate faults.

axioms (1)
  • domain assumption Causal invariance principle holds for the process variables under study
    Invoked to justify the reverse-perspective extraction of a stable causal graph from varying correlations.

pith-pipeline@v0.9.0 · 5697 in / 1268 out tokens · 45968 ms · 2026-05-21T14:47:17.701438+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages

  1. [1]

    A survey of fault diagnosis and fault-tolerant techniques—part i: Fault diagnosis with model-based and signal-based approaches,

    Z. Gao, C. Cecati, and S. X. Ding, “A survey of fault diagnosis and fault-tolerant techniques—part i: Fault diagnosis with model-based and signal-based approaches,”IEEE Transactions on Industrial Electronics, vol. 62, no. 6, pp. 3757–3767, 2015

    work page 2015

  2. [2]

    Ealdl: Element-aware lifelong dictionary learning for multimode process monitoring,

    K. Huang, H. Zhu, D. Wu, C. Yang, and W. Gui, “Ealdl: Element-aware lifelong dictionary learning for multimode process monitoring,”IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 2, pp. 3744–3757, 2025

    work page 2025

  3. [3]

    Deep subdomain learning adaptation network: A sensor fault-tolerant soft sensor for industrial processes,

    X. Zhang, C. Song, J. Zhao, Z. Xu, and X. Deng, “Deep subdomain learning adaptation network: A sensor fault-tolerant soft sensor for industrial processes,”IEEE Transactions on Neural Networks and Learn- ing Systems, vol. 35, no. 7, pp. 9226–9237, 2024

    work page 2024

  4. [4]

    A new multivariate statistical process monitoring method using principal component analy- sis,

    M. Kano, S. Hasebe, I. Hashimoto, and H. Ohno, “A new multivariate statistical process monitoring method using principal component analy- sis,”Computers & Chemical Engineering, vol. 25, no. 7-8, pp. 1103– 1113, 2001

    work page 2001

  5. [5]

    Sparse canonical variate analysis approach for process monitoring,

    Q. Lu, B. Jiang, R. B. Gopaluni, P. D. Loewen, and R. D. Braatz, “Sparse canonical variate analysis approach for process monitoring,”Journal of Process Control, vol. 71, pp. 90–102, 2018

    work page 2018

  6. [6]

    Deep learning of latent variable models for indus- trial process monitoring,

    X. Kong and Z. Ge, “Deep learning of latent variable models for indus- trial process monitoring,”IEEE Transactions on Industrial Informatics, vol. 18, no. 10, pp. 6778–6788, 2021

    work page 2021

  7. [7]

    Spatial-temporal causality modeling for industrial processes with a knowledge-data guided reinforcement learning,

    X. Zhang, C. Song, J. Zhao, Z. Xu, and X. Deng, “Spatial-temporal causality modeling for industrial processes with a knowledge-data guided reinforcement learning,”IEEE Transactions on Industrial Infor- matics, vol. 20, no. 4, pp. 5634–5646, 2024

    work page 2024

  8. [8]

    A comprehensive survey on graph neural networks,

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and S. Y . Philip, “A comprehensive survey on graph neural networks,”IEEE Transactions on Neural Networks and Learning Systems, vol. 32, no. 1, pp. 4–24, 2020

    work page 2020

  9. [9]

    Review on graph neural networks for process soft sensor development, fault diagnosis, and process monitoring,

    M. Jia, Y . Yao, and Y . Liu, “Review on graph neural networks for process soft sensor development, fault diagnosis, and process monitoring,” Industrial & Engineering Chemistry Research, vol. 64, no. 17, pp. 8543– 8564, 2025

    work page 2025

  10. [10]

    Gslb: The graph structure learning benchmark,

    Z. Li, X. Sun, Y . Luo, Y . Zhu, D. Chen, Y . Luo, X. Zhou, Q. Liu, S. Wu, L. Wang, and J. Yu, “Gslb: The graph structure learning benchmark,” inAdvances in Neural Information Processing Systems, vol. 36, 2023, pp. 30 306–30 318

    work page 2023

  11. [11]

    Information- based gradient enhanced causal learning graph neural network for fault diagnosis of complex industrial processes,

    R. Liu, Y . Xie, D. Lin, W. Zhang, and S. X. Ding, “Information- based gradient enhanced causal learning graph neural network for fault diagnosis of complex industrial processes,”Reliability Engineering & System Safety, vol. 252, p. 110468, 2024

    work page 2024

  12. [12]

    Bayesian-based causal structure inference with a domain knowledge prior for stable and interpretable soft sensing,

    X. Zhang, C. Song, B. Huang, and J. Zhao, “Bayesian-based causal structure inference with a domain knowledge prior for stable and interpretable soft sensing,”IEEE Transactions on Cybernetics, vol. 54, no. 10, pp. 6081–6094, 2024

    work page 2024

  13. [13]

    On feature learning in the presence of spurious correlations,

    P. Izmailov, P. Kirichenko, N. Gruver, and A. G. Wilson, “On feature learning in the presence of spurious correlations,”Advances in Neural Information Processing Systems, vol. 35, pp. 38 516–38 532, 2022

    work page 2022

  14. [14]

    Stable soft sensor modeling based on causality analysis,

    F. Yu, Q. Xiong, L. Cao, and F. Yang, “Stable soft sensor modeling based on causality analysis,”Control Engineering Practice, vol. 122, p. 105109, 2022

    work page 2022

  15. [15]

    Causal discovery based on observational data and process knowledge in industrial processes,

    L. Cao, J. Su, Y . Wang, Y . Cao, L. C. Siang, J. Li, J. N. Saddler, and B. Gopaluni, “Causal discovery based on observational data and process knowledge in industrial processes,”Industrial & Engineering Chemistry Research, vol. 61, no. 38, pp. 14 272–14 283, 2022

    work page 2022

  16. [16]

    Physics-guided graph learn- ing soft sensor for chemical processes,

    Y . Liu, M. Jia, D. Xu, T. Yang, and Y . Yao, “Physics-guided graph learn- ing soft sensor for chemical processes,”Chemometrics and Intelligent Laboratory Systems, vol. 249, p. 105131, 2024

    work page 2024

  17. [17]

    Intrinsic causality embedded concurrent quality and process monitoring strategy,

    W. Yu, C. Zhao, B. Huang, and M. Xie, “Intrinsic causality embedded concurrent quality and process monitoring strategy,”IEEE Transactions on Industrial Electronics, vol. 71, no. 11, pp. 15 111–15 121, 2024

    work page 2024

  18. [18]

    Neural network weight comparison for industrial causality discovering and its soft sensing application,

    Y . He, X. Kong, L. Yao, and Z. Ge, “Neural network weight comparison for industrial causality discovering and its soft sensing application,” IEEE Transactions on Industrial Informatics, vol. 19, no. 8, pp. 8817– 8828, 2022

    work page 2022

  19. [19]

    Peters, D

    J. Peters, D. Janzing, and B. Sch ¨olkopf,Elements of causal inference: foundations and learning algorithms. The MIT Press, 2017

    work page 2017

  20. [20]

    Sequence to sequence learning with neural networks,

    I. Sutskever, O. Vinyals, and Q. V . Le, “Sequence to sequence learning with neural networks,” inAdvances in Neural Information Processing Systems, vol. 27. Curran Associates, Inc., 2014

    work page 2014

  21. [21]

    Variational inference over graph: Knowl- edge representation for deep process data analytics,

    Z. Chen, Z. Song, and Z. Ge, “Variational inference over graph: Knowl- edge representation for deep process data analytics,”IEEE Transactions on Knowledge and Data Engineering, vol. 36, no. 6, pp. 2730–2744, 2024

    work page 2024

  22. [22]

    Knowledge-enhanced dis- tributed graph autoencoder for multiunit industrial plant-wide process monitoring,

    W. Wu, C. Song, J. Zhao, and G. Wang, “Knowledge-enhanced dis- tributed graph autoencoder for multiunit industrial plant-wide process monitoring,”IEEE Transactions on Industrial Informatics, vol. 20, no. 2, pp. 1871–1883, 2023

    work page 2023

  23. [23]

    Graph convolutional network- based method for fault diagnosis using a hybrid of measurement and prior knowledge,

    Z. Chen, J. Xu, T. Peng, and C. Yang, “Graph convolutional network- based method for fault diagnosis using a hybrid of measurement and prior knowledge,”IEEE transactions on Cybernetics, vol. 52, no. 9, pp. 9157–9169, 2021

    work page 2021

  24. [24]

    Spatial-temporal associations representation and application for process monitoring using graph convolution neural network,

    H. Ren, X. Liang, C. Yang, Z. Chen, and W. Gui, “Spatial-temporal associations representation and application for process monitoring using graph convolution neural network,”Process Safety and Environmental Protection, vol. 180, pp. 35–47, 2023

    work page 2023

  25. [25]

    Interaction-aware graph neural networks for fault diagnosis of complex industrial processes,

    D. Chen, R. Liu, Q. Hu, and S. X. Ding, “Interaction-aware graph neural networks for fault diagnosis of complex industrial processes,” IEEE Transactions on Neural Networks and Learning Systems, vol. 34, no. 9, pp. 6015–6028, 2021

    work page 2021

  26. [26]

    Graph convolutional network soft sensor for process quality prediction,

    M. Jia, D. Xu, T. Yang, Y . Liu, and Y . Yao, “Graph convolutional network soft sensor for process quality prediction,”Journal of Process Control, vol. 123, pp. 12–25, 2023

    work page 2023

  27. [27]

    Causal generative model for root- cause diagnosis and fault propagation analysis in industrial processes,

    Y . He, L. Yao, Z. Ge, and Z. Song, “Causal generative model for root- cause diagnosis and fault propagation analysis in industrial processes,” IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1–11, 2023. ZHANGet al.: CAUSAL GRAPH SPATIAL-TEMPORAL AUTOENCODER FOR RELIABLE AND INTERPRETABLE PROCESS MONITORING 13

    work page 2023

  28. [28]

    Causal inference by using invariant prediction: identification and confidence intervals,

    J. Peters, P. B ¨uhlmann, and N. Meinshausen, “Causal inference by using invariant prediction: identification and confidence intervals,”Journal of the Royal Statistical Society Series B: Statistical Methodology, vol. 78, no. 5, pp. 947–1012, 2016

    work page 2016

  29. [29]

    Label-free multivariate time series anomaly detection,

    Q. Zhou, S. He, H. Liu, J. Chen, and W. Meng, “Label-free multivariate time series anomaly detection,”IEEE Transactions on Knowledge and Data Engineering, vol. 36, no. 7, pp. 3166–3179, 2024

    work page 2024

  30. [30]

    Knowledge automation through graph mining, convolution, and explanation framework: A soft sensor practice,

    Z. Chen and Z. Ge, “Knowledge automation through graph mining, convolution, and explanation framework: A soft sensor practice,”IEEE Transactions on Industrial Informatics, vol. 18, no. 9, pp. 6068–6078, 2022

    work page 2022

  31. [31]

    Process monitoring using recurrent kalman variational auto-encoder for general complex dynamic processes,

    Z. Zhang, J. Zhu, S. Zhang, and F. Gao, “Process monitoring using recurrent kalman variational auto-encoder for general complex dynamic processes,”Engineering Applications of Artificial Intelligence, vol. 123, p. 106424, 2023

    work page 2023

  32. [32]

    Sensor fault detection and diagnosis using graph convolutional network combining process knowledge and process data,

    L. Guo, H. Shi, S. Tan, B. Song, and Y . Tao, “Sensor fault detection and diagnosis using graph convolutional network combining process knowledge and process data,”IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1–10, 2023

    work page 2023

  33. [33]

    Knowledge graph embedding with graph convolutional network and bidirectional gated recurrent unit for fault diagnosis of industrial processes,

    J. Dong, C. Chen, C. Zhang, J. Ma, and K. Peng, “Knowledge graph embedding with graph convolutional network and bidirectional gated recurrent unit for fault diagnosis of industrial processes,”IEEE Sensors Journal, vol. 25, no. 5, pp. 8611–8620, 2025

    work page 2025

  34. [34]

    A comparison study of basic data-driven fault diagnosis and process monitoring meth- ods on the benchmark tennessee eastman process,

    S. Yin, S. X. Ding, A. Haghani, H. Hao, and P. Zhang, “A comparison study of basic data-driven fault diagnosis and process monitoring meth- ods on the benchmark tennessee eastman process,”Journal of Process Control, vol. 22, no. 9, pp. 1567–1581, 2012

    work page 2012

  35. [35]

    Data-knowledge-driven distributed monitoring for large-scale processes based on digraph,

    W. Wu, C. Song, J. Liu, and J. Zhao, “Data-knowledge-driven distributed monitoring for large-scale processes based on digraph,”Journal of Process Control, vol. 109, pp. 60–73, 2022

    work page 2022

  36. [36]

    Hierarchical fault propagation path recognition method based on knowledge-driven graph attention autoencoder with bilayer pooling for large-scale industrial system,

    Y . Liu, Z. Xu, J. Zhao, C. Song, and D. Wang, “Hierarchical fault propagation path recognition method based on knowledge-driven graph attention autoencoder with bilayer pooling for large-scale industrial system,”Advanced Engineering Informatics, vol. 63, p. 102930, 2025

    work page 2025

  37. [37]

    A soft sensor for multirate quality variables based on mc-cnn,

    B. Song, Y . Zhou, H. Shi, Y . Tao, and S. Tan, “A soft sensor for multirate quality variables based on mc-cnn,”IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 8, pp. 13 927– 13 938, 2025

    work page 2025

  38. [38]

    Flexible clockwork recurrent neural network for multirate industrial soft sensor,

    S. Chang, X. Chen, and C. Zhao, “Flexible clockwork recurrent neural network for multirate industrial soft sensor,”Journal of Process Control, vol. 119, pp. 86–100, 2022

    work page 2022