Tactical–operational architecture for coordinating multi–criteria Pareto planning with distributed model predictive control of iron ore beneficiation processes
DOI:
https://doi.org/10.34185/1562-9945-5-166-2026-02Keywords:
multi-criteria optimisation, Pareto front, NSGA–II, distributed model-predictive control, tactical–operational coordination, iron ore beneficiation, end-to-end uncertainty assessment, block bootstrap, robustnessAbstract
Analysis of recent research and publications. The integration of real–time optimization with model predictive control and data–based hierarchies for beneficiation performance planning have been developed in a number of works; however, the exchange between levels is mostly one–way and does not provide for the propagation of separation–model uncertainty estimates from the operational level to the tactical one.
The purpose of the study is to develop a tactical–operational control architecture for iron ore beneficiation processes that combines static multi–criteria Pareto planning by NSGA–II/III with operational distributed model predictive control through a regulated two–way interface and provides end–to–end uncertainty computation from the separation–model level to the daily–planning level.
Presentation of the main material. The plant performance planning problem is formalized as a five–criteria Pareto problem (concentrate output, iron content, energy and environmental costs, capacity utilization) for four mineralogical–technological ore varieties over a 30–day horizon. A comparison of NSGA–II and NSGA–III showed that NSGA–II provides wider and more uniform coverage of the Pareto front (normalized hypervolume 0.330, 29% higher than NSGA–III). The separation–model coefficients were calibrated against industrial data from three magnetic separators using block bootstrap. A regulated coordination mechanism is proposed with a forward flow (Pareto set–points and admissible deviation zones computed from confidence intervals) and a backward flow (execution statistics and updated covariance matrices). The upper–level reaction is governed by a three–threshold scheme with a statistically justified upper bound of 30%. A robustness metric of Pareto–front points to separation–parameter identification error is developed.
Conclusions. A numerical experiment with a 30–day scenario confirmed the operability of the reaction regulation: after parameter drift and Pareto–front recomputation, the fraction of time outside the admissible zone returns to the normal range. The proposed architecture provides end–to–end uncertainty computation from the separation–model level, through daily–planning operating points, to operational distributed control.
References
1. Ding, J., Wang, H., Liu, C., & Chai, T. (2013). A multiobjective operational optimization approach for iron ore beneficiation process. In 2013 International Conference on Advanced Mechatronic Systems (ICAMechS) (pp. 582–587). IEEE. https://doi.org/10.1109/ICAMechS.2013.6681710
2. Ding, J., Yang, C., & Chai, T. (2017). Recent progress on data–based optimization for mineral processing plants. Engineering, 3(2), 183–187. https://doi.org/10.1016/J.ENG.2017.02.015
3. Ding, J., Yang, C., Xiao, Q., Chai, T., & Jin, Y. (2019). Dynamic evolutionary multiobjec-tive optimization for raw ore allocation in mineral processing. IEEE Transactions on Emerg-ing Topics in Computational Intelligence, 3(1), 36–48. https://doi.org/10.1109/TETCI.2018.2812897
4. Deb, K., Pratap, A., Agarwal, S., & Meyarivan, T. (2002). A fast and elitist multiobjective genetic algorithm: NSGA–II. IEEE Transactions on Evolutionary Computation, 6(2), 182–197. https://doi.org/10.1109/4235.996017
5. Deb, K., & Jain, H. (2014). An evolutionary many–objective optimization algorithm using reference–point–based nondominated sorting approach, Part I: Solving problems with box constraints. IEEE Transactions on Evolutionary Computation, 18(4), 577–601. https://doi.org/10.1109/TEVC.2013.2281535
6. Christofides, P. D., Liu, J., & Muñoz de la Peña, D. (2011). Networked and distributed predictive control: Methods and nonlinear process network applications. Springer. https://doi.org/10.1007/978–0–85729–582–8
7. Mercangöz, M., & Doyle, F. J., III. (2008). Real–time optimization of the pulp mill benchmark problem. Computers & Chemical Engineering, 32(4–5), 789–804. https://doi.org/10.1016/j.compchemeng.2007.03.004
8. Adetola, V., & Guay, M. (2010). Integration of real–time optimization and model predic-tive control. Journal of Process Control, 20(2), 125–133. https://doi.org/10.1016/j.jprocont.2009.09.001
9. Rawlings, J. B., Mayne, D. Q., & Diehl, M. M. (2017). Model predictive control: Theory, computation, and design (2nd ed.). Nob Hill Publishing.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 System technologies
This work is licensed under a Creative Commons Attribution 4.0 International License.
