Brain-inspired memory architecture for condition monitoring based on hippocampal-neocortical complementary learning

Abstract

This paper tackles the critical challenge of catastrophic forgetting and inefficient learning in artificial intelligence models processing continuous, non-stationary data streams. Inspired by neurobiological mechanisms, specifically the Complementary Learning Systems (CLSs) theory involving the hippocampus and neocortex, we propose a novel brain-inspired biomimetic memory system. The core innovation integrates dimensionality reduction techniques—covariance decomposition and low-dimensional mapping—for efficient feature extraction from long spatiotemporal-scale information flows, with a biomimetic learning/forgetting mechanism grounded in CLS principles. Evaluated on real-world power plant operational data, the proposed system demonstrates robust performance against noise and fluctuations, validating the effectiveness of the bionic learning/forgetting mechanism.

Keywords

Continual learning / Complementary learning systems / Principal component analysis / Condition monitoring

References

[1] S. Wold, Exponentially weighted moving principal components analysis and projections to latent structures, Chemom. Intell. Lab. Syst. 23 (1) (1994) 149-161.

[2] N. Cesa-Bianchi, A. Conconi, C. Gentile, On the generalization ability of on-line learning algorithms, IEEE Trans. Inf. Theory 50 (9) (2004) 2050-2057.

[3] I. Portnoy, K. Melendez, H. Pinzon, et al., An improved weighted recursive PCA algorithm for adaptive fault detection, Control Eng. Pract. 50 (2016) 69-83.

[4] S.J. Pan, Q. Yang, A survey on transfer learning, IEEE Trans. Knowl. Data Eng. 22 (10) (2009) 1345-1359.

[5] O. Moussadek, R. Salami, S. Calderara, DOLFIN: balancing stability and plasticity in federated continual learning, arXiv:2510.13567, 2025.

[6] J. Lin, L. Zettlemoyer, G. Ghosh, et al., Continual learning via sparse memory finetuning, arXiv:2510.15103, 2025.

[7] N. Paeedeh, M. Pratama, W.P. Ding, et al., Continual knowledge consolidation LORA for domain incremental learning, arXiv:2510.16077, 2025.

[8] R. Dong, P. Zhao, G. Wang, et al., EdgeSync: accelerating edge-model updates for data drift through adaptive continuous learning, arXiv:2510.21781, 2025.

[9] J.K. Mandivarapu, COLA: continual learning via autoencoder retrieval of adapters, arXiv:2510.21836, 2025.

[10] Z.Y. Yang, Industrial big data modeling and quality prediction based on different learning paradigms, Zhejiang University, Hangzhou, 2021.

[11] P. Ruvolo, E. Eaton, ELLA: an efficient lifelong learning algorithm, Proceedings of the 30th International Conference on Machine Learning (ICML 2013) 28 PMLR, 2013, pp. 507-515.

[12] S. Thrun, Lifelong learning algorithms, in: S. Thrun, L. Pratt (Eds.), Learning to Learn, Springer, Boston, 1998, pp. 181-209.

[13] A. Pentina, C. Lampert, A PAC-Bayesian bound for lifelong learning, Proceedings of the 31st International Conference on Machine Learning (ICML 2014) 32 PMLR, 2014, pp. 991-999.

[14] Z. Chen, B. Liu, Lifelong machine learning, Synth. Lect. Artif. Intell. Mach. Learn. 12 (3) (2018) 1-207.

[15] D.L. Silver, Q. Yang, L. Li, Lifelong machine learning systems: beyond learning algorithms, 2013 AAAI Spring Symposium, 2013, pp. 49-55.

[16] J.L. McClelland, B.L. McNaughton, R.C. O'Reilly, Why there are complementary learning systems in the hippocampus and neocortex: insights from the successes and failures of connectionist models of learning and memory, Psychol. Rev. 102 (3) (1995) 419-457.

[17] Z. Dong, H. Han, H. Li, et al., Long-term potentiation decay and memory loss are mediated by AMPAR endocytosis, J. Clin. Investig. 125 (1) (2015) 234-247.

[18] T.C. Sacktor, How does PKMζ maintain long-term memory? Nat. Rev. Neurosci. 12 (1) (2011) 9-15.

[19] M.Y. Ge, Regulation of retroactive interference by NMDA receptor subunit composition in the Hippocampal dentate gyrus, Huazhong University of Science and Technology, Wuhan, 2019.

[20] R. Dunia, S.J. Qin, A unified geometric approach to process and sensor fault identification and reconstruction: the unidimensional fault case, Comput. Chem. Eng. 22 (7-8) (1998) 927-943.

[21] H.H. Yue, S.J. Qin, Reconstruction-based fault identification using a combined index, Ind. Eng. Chem. Res. 40 (20) (2001) 4403-4414.

[22] Y. Dong, S.J. Qin, A novel dynamic PCA algorithm for dynamic data modeling and process monitoring, J. Process Control 67 (2018) 1-11.

[23] Y. Yu, M. Peng, H. Wang, et al., Long-term operation monitoring strategy for nuclear power plants based on continuous learning, Ann. Nucl. Energy. 178 (2022) 109323.