PLC-Embedded Virtual Pressure Sensor with Continual Learning for Mining Tailings Pumping

• Luís V. Souza UFOP / ITV
• Wagner F. Timoteo Vale S.A.
• Guilherme Euler Vale S.A.
• Lucas M. Ribeiro UFOP
• Eduardo S. Luz UFOP / ITV

DOI: https://doi.org/10.5753/kdmile.2026.32067

Resumo

Reliable pressure measurement in mining tailings pipelines is essential for safe and efficient operation, but physical sensors are prone to failures. This work proposes a virtual pressure sensor based on a compact LSTM network embedded in a PLC. Non-stationary data and hardware constraints introduce covariate shift and catastrophic forgetting in continuous learning. Three training strategies are compared: static dataset (A1), sequential learning without mitigation (A2), and learning with experience replay and targeted oversampling (A3). Results show that A2 suffers severe degradation (R² = 0.48), while A3 achieves the best performance (RMSE = 0.36 bar, R² = 0.94) without increasing inference cost. The model is validated in a SIMATIC PCS 7 environment, confirming real-time feasibility with 517 parameters.

Palavras-chave: catastrophic forgetting, continual learning, LSTM, PLC, soft sensor, tailings pumping

Referências

Chalkley, M., Conard, B., Lakshmanan, V., and Wheeland, K. Tailings and Effluent Management. In Proceedings of the International Symposium on Tailings and Effluent Management. Halifax, Canada, 1989.

de Souza Neto, F. F., Silva, G. A. L., Silva, C. D., and Luz, E. J. S. Aprendizado Online e Contínuo para Previsão de Temperatura em Transportadores de Correia: Uma Abordagem Adaptativa com LSTM. In Anais do Simpósio Brasileiro de Automação Inteligente (SBAI). Fortaleza, CE, Brasil, 2025.

Goodfellow, I., Bengio, Y., and Courville, A. Deep Learning. MIT Press, Cambridge, 2016.

Kirkpatrick, J., Pascanu, R., Rabinowitz, N., Veness, J., Desjardins, G., Rusu, A. A., Milan, K., Quan, J., Ramalho, T., Grabska-Barwinska, A., Hassabis, D., Clopath, C., Kumaran, D., and Hadsell, R. Overcoming Catastrophic Forgetting in Neural Networks. Proceedings of the National Academy of Sciences 114 (13): 3521–3526, 2017.

McCloskey, M. and Cohen, N. J. Catastrophic Interference in Connectionist Networks: The Sequential Learning Problem. Psychology of Learning and Motivation, vol. 24. Academic Press, New York, 1989.

Parisi, G. I., Kemker, R., Part, J. L., Kanan, C., and Wermter, S. Continual Lifelong Learning with Neural Networks: A Review. Neural Networks vol. 113, pp. 54–71, 2019.

Rolnick, D., Ahuja, A., Schwarz, J., Lillicrap, T. P., and Wayne, G. Experience Replay for Continual Learning. In Advances in Neural Information Processing Systems (NeurIPS). Vol. 32. Vancouver, Canada, 2019.

Shen, F., Wu, J., Ye, L., and Wang, B. Memory-Adaptive Supervised LSTM Networks for Deep Soft Sensor Development of Industrial Processes. IEEE Sensors Journal 24 (13): 21641–21654, 2024.

Wills, B. A. and Finch, J. Wills’ Mineral Processing Technology: An Introduction to the Practical Aspects of Ore Treatment and Mineral Recovery. Butterworth-Heinemann, Oxford, 2015.

Zhang, M., Xu, B., Jie, J., Hou, B., and Zhou, L. A Novel Bidirectional Long Short-Term Memory Network with Weighted Attention Mechanism for Industrial Soft Sensor Development. IEEE Sensors Journal 24 (11): 18546–18555, 2024.

Zhou, J., Wang, X., Yang, C., and Xiong, W. A Novel Soft Sensor Modeling Approach Based on Difference-LSTM for Complex Industrial Process. IEEE Transactions on Industrial Informatics 18 (5): 2955–2964, 2022.