A multi-source mixed-frequency information fusion framework based on spatial-temporal graph attention network for anomaly detection of catalyst loss in FCC regenerators

doi:10.1016/j.cjche.2025.02.025

Abstract

Abstract: Anomaly fluctuations in operating conditions, catalyst wear, crushing, and the deterioration of feedstock properties in fluid catalytic cracking (FCC) units can disrupt the normal circulating fluidization process of the catalyst. Although several effective models have been proposed in previous research to address anomaly detection in chemical processes, most fail to adequately capture the spatial-temporal dependencies of multi-source, mixed-frequency information. In this study, an innovative multi-source mixed-frequency information fusion framework based on a spatial-temporal graph attention network (MIF-STGAT) is proposed to investigate the causes of FCC regenerator catalyst loss anomalies for guide onsite operational management, enhancing the long-term stability of FCC unit operations. First, a reconstruction-based dual-encoder-decoder framework is developed to facilitate the acquisition of mixed-frequency features and information fusion during the FCC regenerator catalyst loss process. Subsequently, a graph attention network and a multilayer long short-term memory network with a differential structure are integrated into the reconstruction-based dual-encoder-shared-decoder framework to capture the dynamic fluctuations and critical features associated with anomalies. Experimental results from the Chinese FCC industrial process demonstrate that MIF-STGAT achieves excellent accuracy and interpretability for anomaly detection.

Key words: Chemical processes, Deep learning, Anomaly detection, Mixed-frequency, Non-stationary, Graph attention network

摘要： Anomaly fluctuations in operating conditions, catalyst wear, crushing, and the deterioration of feedstock properties in fluid catalytic cracking (FCC) units can disrupt the normal circulating fluidization process of the catalyst. Although several effective models have been proposed in previous research to address anomaly detection in chemical processes, most fail to adequately capture the spatial-temporal dependencies of multi-source, mixed-frequency information. In this study, an innovative multi-source mixed-frequency information fusion framework based on a spatial-temporal graph attention network (MIF-STGAT) is proposed to investigate the causes of FCC regenerator catalyst loss anomalies for guide onsite operational management, enhancing the long-term stability of FCC unit operations. First, a reconstruction-based dual-encoder-decoder framework is developed to facilitate the acquisition of mixed-frequency features and information fusion during the FCC regenerator catalyst loss process. Subsequently, a graph attention network and a multilayer long short-term memory network with a differential structure are integrated into the reconstruction-based dual-encoder-shared-decoder framework to capture the dynamic fluctuations and critical features associated with anomalies. Experimental results from the Chinese FCC industrial process demonstrate that MIF-STGAT achieves excellent accuracy and interpretability for anomaly detection.

关键词: Chemical processes, Deep learning, Anomaly detection, Mixed-frequency, Non-stationary, Graph attention network

Chunmeng Zhu, Nan Liu, Ludong Ji, Yunpeng Zhao, Xiaogang Shi, Xingying Lan. A multi-source mixed-frequency information fusion framework based on spatial-temporal graph attention network for anomaly detection of catalyst loss in FCC regenerators[J]. Chinese Journal of Chemical Engineering, 2025, 84(8): 47-59.

References

[1] X.T. Bi, R.S. Qin, D.Y. Wu, S.D. Zheng, J.S. Zhao, One step forward for smart chemical process fault detection and diagnosis, Comput. Chem. Eng. 164 (2022) 107884.
[2] S. Chauhan, G. Vashishtha, R. Zimroz, Analysing recent breakthroughs in fault diagnosis through sensor: a comprehensive overview, Comput. Model. Eng. Sci. 141 (3) (2024) 1983-2020.
[3] Q. Liao, Y.F. Wang, H.R. Zhang, Recent advances in intelligent oil and gas engineering, Chem. Eng. Res. Des. 188 (2022) 1011-1012.
[4] M. Suvarna, K.S. Yap, W.T. Yang, J. Li, Y.T. Ng, X.N. Wang, Cyber-physical production systems for data-driven, decentralized, and secure manufacturing: a perspective, Engineering 7 (9) (2021) 1212-1223.
[5] Z.Y. He, H.D. Shao, J.S. Cheng, Y. Yang, J.W. Xiang, Kernel flexible and displaceable convex hull based tensor machine for gearbox fault intelligent diagnosis with multi-source signals, Measurement 163 (2020) 107965.
[6] J. Tang, J.F. Qiao, Z. Liu, N. Sheng, W. Yu, G. Yu, Dual-layer optimized selective information fusion using multi-source multi-component mechanical signals for mill load parameters forecasting, Mech. Syst. Signal Process. 135 (2020) 106371.
[7] J. Tang, J.F. Qiao, Z.W. Wu, T.Y. Chai, J. Zhang, W. Yu, Vibration and acoustic frequency spectra for industrial process modeling using selective fusion multi-condition samples and multi-source features, Mech. Syst. Signal Process. 99 (2018) 142-168.
[8] Y.J. Zhang, J.L. Wang, Do high-frequency stock market data help forecast crude oil prices? Evidence from the MIDAS models, Energy Econ. 78 (2019) 192-201.
[9] X. Zhao, M. Han, L.L. Ding, W.L. Kang, Usefulness of economic and energy data at different frequencies for carbon price forecasting in the EU ETS, Appl. Energy 216 (2018) 132-141.
[10] E. Andreou, E. Ghysels, A. Kourtellos, Regression models with mixed sampling frequencies, J. Econom. 158 (2) (2010) 246-261.
[11] A.G. Rameshrao, E. Koley, S. Ghosh, Reliability enhancement of hybrid microgrid protection against communication data loss and converter faults using cubic-spline interpolation, Savitzky Golay filtering and GRU network, Comput. Electr. Eng. 116 (2024) 109144.
[12] H.G. Han, M.T. Sun, H.Y. Han, X.L. Wu, J.F. Qiao, Univariate imputation method for recovering missing data in wastewater treatment process, Chin. J. Chem. Eng. 53 (2023) 201-210.
[13] T.Y. Gao, J.L. Yang, Q. Tang, A multi-source domain information fusion network for rotating machinery fault diagnosis under variable operating conditions, Inf. Fusion 106 (2024) 102278.
[14] Y.M. An, Y.G. Dang, J.J. Wang, H.M. Zhou, S.T. Mai, Mixed-frequency data sampling grey system model: forecasting annual CO₂ emissions in China with quarterly and monthly economic-energy indicators, Appl. Energy 370 (2024) 123531.
[15] J.H. Lin, G. Michailidis, A multi-task encoder-dual-decoder framework for mixed frequency data prediction, Int. J. Forecast. 40 (3) (2024) 942-957.
[16] L. Allen, H.P. Lu, J. Cordiner, Knowledge-enhanced spatiotemporal analysis for anomaly detection in process manufacturing, Comput. Ind. 161 (2024) 104111.
[17] A. Bhakte, V. Pakkiriswamy, R. Srinivasan, An explainable artificial intelligence based approach for interpretation of fault classification results from deep neural networks, Chem. Eng. Sci. 250 (2022) 117373.
[18] H.H. Gao, C. Wei, W.J. Huang, X.J. Gao, Multimode process monitoring based on hierarchical mode identification and stacked denoising autoencoder, Chem. Eng. Sci. 253 (2022) 117556.
[19] L. Liu, H.T. Zhao, Z.W. Hu, Graph dynamic autoencoder for fault detection, Chem. Eng. Sci. 254 (2022) 117637.
[20] H. Wang, S.H. Wang, W.F. Sun, J.W. Xiang, Multi-sensor signal fusion for tool wear condition monitoring using denoising transformer auto-encoder Resnet, J. Manuf. Process. 124 (2024) 1054-1064.
[21] S. Zhang, T. Qiu, Semi-supervised LSTM ladder autoencoder for chemical process fault diagnosis and localization, Chem. Eng. Sci. 251 (2022) 117467.
[22] S. Plakias, Y.S. Boutalis, A novel information processing method based on an ensemble of Auto-Encoders for unsupervised fault detection, Comput. Ind. 142 (2022) 103743.
[23] Y. Jing, X.L. Ge, B.T. Liu, Toward understandable semi-supervised learning fault diagnosis of chemical processes based on long short-term memory ladder autoencoder (LSTM-LAE) and self-attention (SA), Comput. Chem. Eng. 189 (2024) 108817.
[24] C.Y. Ding, S.L. Sun, J. Zhao, MST-GAT: a multimodal spatial-temporal graph attention network for time series anomaly detection, Inf. Fusion 89 (2023) 527-536.
[25] M.W. Jia, D.Y. Xu, T. Yang, Y. Yao, Y. Liu, Graph-guided masked autoencoder for process anomaly detection, Process. Saf. Environ. Prot. 186 (2024) 1345-1357.
[26] J.J. Luo, Z.H. Jin, H.P. Jin, Q. Li, X. Ji, Y.Y. Dai, Causal temporal graph attention network for fault diagnosis of chemical processes, Chin. J. Chem. Eng. 70 (2024) 20-32.
[27] J. Seo, Y. Noh, Y.J. Kang, J. Lim, S. Ahn, I. Song, K.C. Kim, Graph neural networks for anomaly detection and diagnosis in hydrogen extraction systems, Eng. Appl. Artif. Intell. 135 (2024) 108846.
[28] Y.F. Shi, B. Wang, Y.W. Yu, X.F. Tang, C. Huang, J.Y. Dong, Robust anomaly detection for multivariate time series through temporal GCNs and attention-based VAE, Knowl. Based Syst. 275 (2023) 110725.
[29] S.L. Sun, H. Ding, Z.D. Zhao, W.F. Xu, D. Wang, SCG-GFFE: a Self-Constructed graph fault feature extractor based on graph Auto-encoder algorithm for unlabeled single-variable vibration signals of harmonic reducer, Adv. Eng. Inform. 62 (2024) 102579.
[30] U. Goswami, H. Kodamana, M. Ramteke, Fault detection using graph neural differential auto-encoders (GNDAE), Comput. Chem. Eng. 189 (2024) 108775.
[31] J.L. Zheng, X. Chen, C.H. Zhao, Coarse-to-fine condition identification for wide-range non-stationary processes driven by coupled condition indicators, Contr. Eng. Pract. 128 (2022) 105328.
[32] J. Huang, X.Y. Sun, X. Yang, K.X. Peng, Fault detection for chemical processes based on non-stationarity sensitive cointegration analysis, ISA Trans. 129 (Pt B) (2022) 321-333.
[33] C.H. Zhao, Perspectives on nonstationary process monitoring in the era of industrial artificial intelligence, J. Process. Contr. 116 (2022) 255-272.
[34] L. Lin, L.Z. Zu, S. Fu, Y.K. Liu, S.H. Zhang, S.W. Suo, C.S. Tong, Integrating adversarial training strategies into deep autoencoders: a novel aeroengine anomaly detection framework, Eng. Appl. Artif. Intell. 136 (2024) 108856.
[35] L.J. Wang, H.H. Chen, L.Q. Yang, J.L. Li, Y. Li, X.N. Wang, Single-atom catalysts property prediction via Supervised and Self-Supervised pre-training models, Chem. Eng. J. 487 (2024) 150626.
[36] Y. Liu, K. Huang, B.J. Ma, K. Wei, Y. Li, C. Yang, W. Gui, Quality-related fault detection for dynamic process based on quality-driven long short-term memory network and autoencoder, Neural Netw. 181 (2025) 106819.
[37] Y.T. Li, K. Wu, J. Liu, Self-paced ARIMA for robust time series prediction, Knowl. Based Syst. 269 (2023) 110489.
[38] H. Liu, Y.C. Sun, W.H. Ding, H.L. Wu, H.Y. Zhang, Enhancing non-stationary feature learning for remaining useful life prediction of aero-engine under multiple operating conditions, Measurement 227 (2024) 114242.
[39] D.P.Kingma, J. Ba, Adam: A Method for Stochastic Optimization, In:International Conference on Learning Representations (ICLR) 2015,San Diego, California, USA,2015.
[40] J.Y. Luo, X.Y. Kong, C.H. Hu, H.Z. Li, Key-performance-indicators-related fault subspace extraction for the reconstruction-based fault diagnosis, Measurement 186 (2021) 110119.
[41] J.H. Yu, X. Gao, B.F. Li, F. Zhai, J.S. Lu, B. Xue, S.Y. Fu, C. Xiao, A filter-augmented auto-encoder with learnable normalization for robust multivariate time series anomaly detection, Neural Netw. 170 (2024) 478-493.
[42] Y. Wang, H. Wu, J. Zhang, Z. Gao, J. Wang, P.S. Yu, M. Long, M. Long, PredRNN: a recurrent neural network for spatiotemporal predictive learning, IEEE Trans. Pattern Anal. Mach. Intell. 45 (2) (2023) 2208-2225.