Keywords: Control and optimization, Process modeling, Ore preparation, flotation
Abstract: Industrial plant operators recognise the benefits of process control to reduce the variability of key process variables. However, the actual financial value for the operation often remains elusive. There are no clear answers to apparently simple questions like what standard deviation can be tolerated on a flotation feed rate or more broadly, what is the cost of variability? A similar vague understanding reigns among metallurgists about optimisation. Is it better to maximise the plant feed rate or the recovery? Stage objectives can hardly make sense without considering the global performance, but even then, defining the optimal plant target raises questions. Quantifying the effect of the grinding product attributes on the separation process performance indicators has also posed significant practical challenges. The technological developments of automated quantitative mineralogy of the last two and a half decades introduced new capabilities to solve this conundrum. As the mineral liberation distribution determines the ultimate grade and recovery curve, it seems natural to consider it for process control and optimisation applications. This paper examines the introduction of mineral liberation in the quest to generalise the solution of the process control and optimisation problem of separation plants. It reviews the advances of the last decade, focusing on model, simulation, control and optimisation developments. It also provides a prospective outlook of promising research work and technologies that will bring closer the materialisation of plant-wide or even mine-wide control.

Keywords: Mining operations, mineral processing, Control and optimization, Automation, instrumentation
Abstract: Effective material distribution among silos is essential in mineral processing to reduce wear and ensure continuous operation. This work proposes a control strategy to improve productivity and equipment integrity in screening by automatically deactivating lines when additional screening surface is not needed. The approach adjusts the number of active silos based on variables such as circulating load and silo levels, introducing the concept of available silos. Applied to an iron ore plant in Brazil, the strategy increased the average screening rate by 13.4% while deactivating lines 62.3% of the time, with no efficiency loss and potential equipment integrity gains.

Keywords: Control and optimization, Ore preparation, flotation, Production planning, distribution and logistics, plant management
Abstract: Knowing in advance a concentrator plant feed ore types can enable to adjust and optimize the feed ore blend according to the available fresh ores and stockpiles. In additions to that, operational configuration and key setpoints of a plant may be adjusted according to the known incoming feed ore requirements. For this, plant metallurgical digital twin simulation model aids in predicting performance of a concentrator for any given feed blend and plant operating modes. This paper presents an automated method for simultaneously adjusting an ore blend and plant configuration to maximize operating profit. The optimizer utilizes a concentrator plant first-principle digital twin model, based on known ore characteristics and equipment operation. The proposed optimizer can be used in day-to-day process operation to automatically suggest optimized results for both metallurgical and economical KPIs. A case simulation study of a copper concentrator with five ore types, several settings for processing circuit and equipment is presented. Here, for a fixed feed blend the plant configuration optimization resulted additional 1.7 % increment in operating profit, while simultaneously optimizing both feed blend and plant configuration showed even more remarkable potential for economic benefits.

Keywords: Artificial intelligence, machine learning systems, and human machine systems, Data mining and statistical analyses, Mining operations, mineral processing
Abstract: Ore crushing circuits are often characterized by distinct operating modes, knowledge of which contributes towards improved control and operation. Process modes were extracted in an unsupervised manner from an expanded, feature-rich (i.e., including auxiliary measurements) historical dataset by non-linear dimensionality reduction and clustering. Subsequently, a kernel support vector machine was trained to classify operating conditions according to previously identified modes using routine process measurements. Applying the classifier to unseen data indicated that the identified process modes were distinct and meaningful. Finally, Shapley additive explanations (SHAP) analysis was shown to be a valuable tool for understanding the classifier’s outputs.

Keywords: Control and optimization, Mining operations, mineral processing
Abstract: This article presents an application of the Smith predictor-based control strategy for controlling the feed rate of a primary crushing circuit in an iron ore beneficiation plant, focusing on compensating for system dead time. The methodology includes problem characterization, plant modeling, and application at Vale S.A.'s Fábrica Nova unit, Brazil. The results show a 4.3-minute reduction in the time to reach the feed setpoint and a 9.2% increase in the plant's productivity, without compromising process safety. The proposed strategy contributes to the process control literature by demonstrating the practical application of the Smith predictor for dead time compensation, providing a replicable alternative for other industrial systems.

Keywords: Control and optimization, Advanced process control, Mining operations, mineral processing
Abstract: In secondary grinding circuits, the product particle size is a key variable influencing downstream performance. However, online particle size analyzers are often too costly or impractical to implement, limiting the ability to reach and maintain production objectives. This paper investigates self-optimizing control (SOC) of the product particle size using readily available instrumentation. The method identifies linear combinations of process variables that remain close to their target values despite disturbances, using a null space approach applied to steady-state data extracted from a dynamic model of the grinding circuit. Simulation results show that SOC can significantly reduce product size fluctuations caused by ore hardness variations, ore feed rate variations, and ore particle size variations. Compared to the baseline strategy, which leads to deviations of up to 9.2% relative to the nominal product size, the best SOC configuration limits fluctuations to less than 2%, using simple control loops.

Keywords: Mining operations, mineral processing, Artificial intelligence, machine learning systems, and human machine systems, Measurement, sensors
Abstract: In mill operation, vibrations are generated by interactions between internal materials and the mill shell, which can directly reflect the milling condition and have been used for recognition of milling condition. However, existing methods exhibit limitations in time-frequency feature modeling and complex condition recognition, which limit the accuracy and practicality of condition monitoring. In this paper, a novel method is proposed for milling condition recognition based on vibration signals using a Transformer architecture. In the method, a time-frequency decoupled attention mechanism is proposed to extract intrinsic time and frequency features from vibration signals, and a feature aggregation module is introduced for effective integration of different features. Experimental results on a laboratory-collected mill vibration dataset demonstrate that the proposed method outperforms existing CNN-based and Transformer-based approaches in recognition performance and show the effectiveness of the method.

Keywords: Mining operations, mineral processing
Abstract: The extraction and depletion of natural resources, along with the production of waste, pose significant environmental challenges for the mining industry. These challenges are exacerbated by linear supply chains that involve extraction, manufacturing, and disposal. The Circular Economy (CE) aims to mitigate these impacts by encouraging a consumption and production model that minimizes waste and pollution while being both regenerative and restorative. This presentation outlines the evolution of CE in Mineral Processing and Extractive Metallurgy (MP&EM), highlighting its current status. Process Systems Engineering (PSE) is identified as a crucial element in driving the transition toward sustainability, utilizing its core competencies in process design, integration, modeling, optimization, control, and planning. In the author's view, CE within the context of MP&EM is still in its early stages, with limited proactive initiatives at the core of the processes and a largely reactive focus outside the production processes. The author positions PSE as a vital tool for the future of the mining industry, supporting the implementation of CE practices.

Keywords: Mining operations, mineral processing, State estimation and parameter identification, Measurement, sensors
Abstract: The charge in a semi-autogenous grinding mill consists of grinding media and slurry, of which the grinding media is a combination of large ore (rocks) and steel balls, and the slurry is a mixture of small ore (solids) and water. The constituents of the charge can be represented by a nonlinear model with states and parameters observable from measurements at the mill discharge. In this paper, a transformed version of the nonlinear model is used in an extended Kalman filter to estimate the volumetric filling of grinding media (rocks and balls), solids, and water with reasonable accuracy from measurement commonly available in industrial settings. This information is of valuable importance for the control of grinding mills.

Keywords: Surface treatment, galvanizing, protective coating, welding, Process modeling, Hot/cold rolling, forming, forging
Abstract: Understanding the wear behavior of the work rolls in a temper rolling or cold rolling process is crucial to model the surface profile transfer onto the rolled strip. Although roll wear has been extensively studied, its dynamics during the processing of zinc-coated materials are not yet fully understood. Zinc adhesion on the roll surface can significantly influence the measured height profiles. A novel empirical roll wear model predicts the surface height distribution as a function of the rolled length. The model accounts for the dominant material removal and also includes material transfer through smearing or zinc adhesion. The model preserves maximal statistical surface information, while being computable sufficiently quickly for realtime applications. A displacement interpolation method from optimal transport (OT) theory is adapted to approximate measured height distributions. Experimental validation, based on data from an industrial temper rolling process, demonstrates that the model accurately predicts both the evolution of the height distribution and the arithmetic mean roughness R_a.

Keywords: Production planning, distribution and logistics, plant management, Iron making, steel making
Abstract: The production of steel is often considered a key indicator of a nation’s economic growth. The complex production environment of Thermo-Mechanical Treated (TMT) steel bar products, with multiple products of different diameters and grades, in the presence of sequence-dependent changeover times, requires efficient production planning and scheduling to ensure agile decision making and maximized profitability. In this work, we propose an integrated production planning and scheduling framework that, at the planning level, meets the intermediate product demands and maintains the inventory, while at the scheduling level ensures minimum changeover times and idle production periods to ensure maximum productivity. The proposed novel multi-grid discrete-time formulation and the resulting mixed-integer linear programming model demonstrate a significant reduction in the number of variables and constraints, increasing the computational efficiency. We further propose a rolling horizon implementation strategy to ensure an agile and alert decision making. We demonstrate the efficacy of the proposed optimization formulation on a real-world industrial TMT bar production facility in India.

Keywords: Process modeling, Mining operations, mineral processing, Fault diagnosis, process monitoring
Abstract: Dynamic flotation modelling is a valuable tool for designing process control and monitoring solutions for industrial application. However, flotation modelling is challenging, due to the complex interactions in the pulp and froth phases in a flotation cell, represented by hydrodynamic and kinetic relations. Dynamic flotation models published in literature are often not reproducible, due to a lack of sufficient detail (including parameter values and typical operating conditions), or do not consider specific relations of interest. In this work, a dynamic flotation model is developed for the purpose of rapid prototyping of industrial control and monitoring solutions, specifically such solutions that make use of pulp and froth sensors. These requirements determined the model complexity (e.g., including hydrodynamic and frother relations, while limiting the number of tunable parameters to allow easy calibration for new case studies). The goal is not a high-fidelity, high accuracy model, but a model that captures typical measurements, inputs, and interactions. A copper rougher case study with complete parameter and operating condition descriptions is provided, to promote reproducibility and further development of the model. Example applications of the dynamic flotation model to the design of advanced process control and frother advisor monitoring are also described.

Keywords: Ore preparation, flotation, State estimation and parameter identification, Advanced process control
Abstract: This paper introduces an adaptive mechanism, based on the modulating function method and Lyapunov–Krasovskii functional analysis, for the online estimation of time-varying delays in processes with complex multivariable dynamics within a min–max robust model predictive controller (RMPC) framework. Monte Carlo simulations were conducted to test the performance of a nominal RMPC under different time delay uncertainties, along with the RMPC with the proposed modulating function-based adaptive mechanism, showing that the latter is capable of reducing the uncertainty polytope of the controller, demonstrating less control action conservatism and a more robust transient response.

Keywords: Control and optimization, Mining operations, mineral processing, Process modeling
Abstract: Flotation is a common and important part of many mineral beneficiation processes, especially in gold applications; and therefore, its performance is essential to keep as well trimmed as possible. In this paper, an optimization method for the gold recovery by means of adjusting mass pulls of flotation tanks is introduced and tested with process simulations. The target of optimization was here to reduce the gold grade in final tailings while keeping the gold grade of both final concentrate and rougher concentrate in a certain range. Two optimization algorithms, differential evolution and interior-point method, were tested. The investigated hypothesis was “history-optimized control setpoints improve performance in future”, and it was shown to be correct, as indicated by the reduced gold grade in final tailings. Here, the methods demonstrated that the mass pull increase, especially in rougher flotation, was often favourable.

Keywords: Mining operations, mineral processing, Advanced process control, Automation, instrumentation
Abstract: This paper explores the potential of digital twins for flotation control and their integration into existing control hierarchies. While most literature delves directly into technical details such as model architectures and data-driven techniques, this study takes a step back to evaluate broader implementation value. It examines how digital twins can enhance advanced process control (APC) performance, from soft-sensor outputs and optimisation through recommendations (decision support), to direct interventions (autonomous control), fault detection and mitigation, and maintenance workflows. A business case based on a steady-state model is presented, using mass pull control as a practical example. The economic evaluation employs simplified models to avoid unnecessary complications, and results suggest an average benefit of approximately USD 3 million per annum. Implementation costs are assessed by drawing parallels with APC systems, and a net present value (NPV) analysis shows breakeven is achievable even under conservative scenarios where only 10% of the projected benefit is realised. Six key challenges are identified - including limited sensor coverage, data quality, model complexity, and the human/twin interface - that must be addressed for successful adoption. Ultimately, this work aims to foster a grounded discussion between researchers and practitioners on the real-world value and practicality of implementing digital twins in flotation control.