Keywords: Advanced process control
Abstract: Since its introduction in the 1940’s, about 80 years ago, conventional advanced process control (APC) has largely been overlooked by the academic community, yet it is still thriving in industrial practice, even after 40 years with model-based multivariable control (MPC). So it is safe to predict that conventional APC (including PID control) will not be replaced by MPC, but will remain in the toolbox along with MPC.

Conventional APC includes the standard control elements that industry commonly uses to enhance control when simple single-loop PID controllers do not give acceptable control performance. Examples of such control elements are cascade control, selectors, split range control, input or valve position control (VPC), multiple controllers (and MVs) for the same CV, and nonlinear calculation blocks. The goal of this talk is to take a systematic view on how to design a conventional APC system. The starting point is usually optimal steady-state economic operation. The process may have many manipulated variables (MVs) for control (typically valves), but usually most of these are used to control “active” constraints, which are the constraints which optimally should be kept at their limits at steady state. For the remaining unconstrained degrees of freedom, we should look for self-optimizing variables, which are measured variables for which the optimal values depend weakly on the disturbances.

However, during operation one may reach new (active) constraints, either on MVs or CVs, which may be easily observed from measurements of the potential constraints. Since the number of degrees of freedom doesn’t change, we will need to give up the control of another variable, which will either by another constraint (on CV or MV) or an unconstrained CV (self-optimizing variable). The key is then to know which variable give up, and in many cases we may determine this based on physical insight, and implement it using conventional APC elements.

In summary, optimal economic operation may in many cases be achieved by use of simple conventional APC elements, but there is a lack of understanding, both in industry and academia, on how such control systems should be designed. The presentation offers a new beginning in terms of providing a systematic approach.

Keywords: Energy, environment, health, safety, Iron making, steel making
Abstract: Steel as the base material number 1 for global industry is essential for the energy transition but at the same time responsible for around 7% of global greenhouse gas emissions. Based on the example of the European steel industry, which is committed to reducing its direct and indirect CO2 emissions and achieving an 80-95% reduction in CO2 emissions by 2050 compared to 1990 levels, the opportunities and requirements of this transformation will be explained. The overall transformation could be achieved through hydrogen-based steelmaking, adaptation of fossil fuel-based steelmaking through process integration, and capture and use of carbon waste for the production of chemicals and increased recycling of steel scrap and steel by-products.

However, the transformation is dependent on the energy transition to renewable energies, which has yet to be completed, and a leading market for this green steel. However, the hydrogen-based steelmaking route imposes stricter requirements on future feedstocks and additional challenges in transforming primary steelmaking while retaining the existing secondary metallurgy and continuous casters in the steel plants in terms of productivity and ensuring that the metallurgical requirements of the product portfolio are met.

Over the course of its long history, particularly in the last 150 years, iron and steel production has undergone many changes and further developments in processing and, with the appropriate political support, will also manage the transformation to carbon-neutral steelmaking.

Keywords: Ore preparation, flotation, Mining operations, mineral processing, Control and optimization
Abstract: Ore heterogeneity (e.g., head grade, hardness, density and lithology) represents an important preoccupation for any given mine. The actual consequence of this variability is not always clear on the bottom line, especially when the average value remains constant over time. This paper examines the effect of random variations of two feed particle characteristics, namely the size index and grindability, on the net value of a grinding/flotation plant product. The simulated plant consists of rougher flotation cells processing the ore ground by a 2- stage semiautogenous/ball mill circuit. Normal Gaussian distributions model the centered heterogeneity over time for three level of standard deviations (low, moderate and high) generating 32 possible scenarios. The high variability case reduces the product net value compared to the low variability one, equivalent to 1.5 M/year loss. The additional revenues come from the ability to process more ore on average, while achieving the desired product size index in narrow variability situations. Conversely, high standard deviations lead to a slightly coarser ground product. The proposed simulation setting provides a quantitative basis to assess the financial benefits of regulatory control, and ore blending systems.

Keywords: Ore preparation, flotation, Artificial intelligence, machine learning systems, and human machine systems
Abstract: Ores are becoming extremely variable with mineralogy and hardness disturbing the integrated crushing, grinding, flotation, and thickening processes. The current grinding and flotation sensors provide large amounts of data for process optimization. To augment the operational knowledge for proactive actions for improving the performance of the grinding and flotation circuits, we need to add the right process knowledge context and operational modes. Using the latest tools and cloud computing enables the creation of new workflows and collaboration between mining, concentrator plants, and the enterprise, including services providers. The novel approach of using machine-learning techniques coupled with dynamic process models in grinding, such as Dynamill™ and Dynaflote™, a new operational integrated grinding model is realized and implemented. Today, subject matter experts (SMEs) can increase productivity by developing predictive models to classify the operating conditions owing to large variations in ores, catching the hidden production, energy, and water losses by ore type and unmeasured disturbances. Remote support and coaching have become established methods to obtain high benefits in remote mining operations all over the world. A Net Metal Production Rate (NMPR) model is presented to determine the optimal mill throughput, with the right P80, % solids Cyclone Feed Rate and Metal Flotation Recovery estimates in real time. As such, the right manipulated controlled variables can be known to maximize the NMPR. The development of advanced predictive capabilities (enhanced Particle Size Distribution Shape (PSDS) soft sensors, and operational derived variables such as air holdup, energy intensity, and flotation bank profile) in flotation recovery is now possible.

Keywords: Control and optimization, Mining operations, mineral processing, Advanced process control
Abstract: Advanced control is significant to improve the process performance for grinding and classification process. Due to frequent variation of ore property, many advanced control methods cannot work effectively in a long period. To overcome this problem, an adaptive fuzzy logic control method based on the trend of grinding sound is proposed for grinding process. In this method, the trend of grinding sound is extracted by qualitative trend analysis (QTA), grinding sound signal and mill power are used as controller inputs to control ore feed and water feed, and improved fuzzy C-means (FCM) is adopted to adapt parameters of fuzzy controller according to actual process conditions. Simulation results show that this proposed approach achieves less fluctuation of ore feed than manual control, ensures that ore throughput meets the target of this concentrator and reduces energy consumption of ball mill.

Keywords: Artificial intelligence, machine learning systems, and human machine systems, Control and optimization, Process modeling
Abstract: The automatic tuning problem of multiple-input-multiple-output (MIMO) controllers is considered within the framework of Bayesian optimisation and applied in simulation to a bulk tailings treatment process. The aim is to develop a model free, on-line, automatic tuner which can optimise the performance of a given controller to the task at hand. The automatic tuning procedure can be conducted during commissioning, when poor controller performance is observed or when the process has changed. Simulations indicate that the method is able to locate the optimal tuning parameters for the bulk tailings treatment process as compared to a de-coupled controller developed from a model of the process. The parameters were obtained from an objective function which was balanced and weighted according to the response required.

Keywords: Control and optimization, Mining operations, mineral processing, Measurement, sensors
Abstract: The paper presents the results of simulation tests of the dynamic characteristics of a three-product jig for coal separation and control systems for refuse discharge. The changes in the position of the stratified bed layers and the density of the separation layer in the product discharge zone as well as the static characteristics of the zone in two compartments of the jig were investigated. It was assumed that the main disturbance occurring in industrial systems are changes in the flow rate and the feed washability characteristics. The simulation model was verified in industrial tests in two coal preparation plants. An analysis of the effects of the control of the bottom product discharge system for nonlinear PI controllers and adaptive radiometric density meters with variable measurement time was carried out. The aim of control was to stabilize the density of the separation layer at the desired level. The parameters of the non-linear PI controller depended on the direction of changes in the feed flow rate. The control effects were compared with linear control systems and meters with a constant measurement time commonly used in industry. Proposed system improves product quality and reduces product losses in waste compared to conventional systems.

Keywords: Process modeling, Data mining and statistical analyses, Iron making, steel making
Abstract: The burn-through point (BTP) is an essential parameter in the sintering process, but there is no method to monitor it directly at the present. Traditional data-driven modeling strategies, on the other hand, are incapable of reliably predicting BTP. That is because real-world datasets feature input and output outliers, as well as a multicollinearity problem. To address these issues, a new stochastic configuration network based on error probability density function optimization (PDF-RPSCN) is proposed in this paper. Firstly, the weight of samples is calculated via generalized M-estimation using the residual value of the model and the distance information of the input vector in high-dimensional space to eliminate the influence of outliers with anomalies in the input and output directions. Then, to overcome the multicollinearity problem that exists in data samples, the least-squares method used in stochastic configuration network (SCN) is modified with a partial least-squares (PLS) method. At the same time, the gradient descent method is used to optimize the model parameters, making the real modeling error PDF shape close to the ideal PDF shape while considering the integral of squared deviation between the modeling error PDF and the modeling error target PDF as the performance index. Finally, a BTP modeling experiment based on actual industrial datasets is shown. The algorithm's general approximation performance has substantially improved, resulting in an easy-to-use model with enhanced accuracy and robust performance.

Keywords: Advanced process control, Control and optimization, Iron making, steel making
Abstract: The stable operation and the optimal thermal control of industrial blast furnaces are challenging due to the complexity of the multi-phase and multi-scale physical and chemical phenomena, the presence of fast and extremely slow dynamics with latency periods of more than 8 hours, the absence of direct measurements of key inner variables, and the occurrence of a wide range of unknown disturbances. Industrial blast furnaces are still operated in a semi-automated manner and the quality of the control depends on the skills and dedication of the operators. Model-based control schemes, operated either in closed-loop or as advisory systems, are an obvious option to achieve a smooth and energetically efficient operation of blast furnaces. This work proposes a hybrid dynamic model-based optimizing control scheme for achieving the desired operational objectives by tightly controlling the hot metal silicon content ([Si]) and the slag basicity (SB) at their desired set-points. These two variables are key product quality indices and indicators of the internal thermal status of the blast furnace process. Within the proposed framework, the optimizer regulates the fast dynamics of the blast furnace to counteract the unmeasured disturbances that are caused by the variations in the solid feed, subject to operational constraints. Simulation results on a large-scale industrial blast furnace demonstrate the potential of the proposed approach for an improved operation of the blast furnace process.

Keywords: Process modeling, Ferrous metal, Control and optimization
Abstract: This paper presents presents an approach to compute the optimal voltage and impedance setpoints that achieve the optimum electrical efficiency of an EAF for operation at a fixed power level. In the optimization framework, an electric arc model and an EAF process model are embedded into a dynamic optimization framework that aims at minimizing the radiation losses of the process. A numerical case study shows a reduction of 5% of the electrical power consumption of the process and an increment of 2% in the energy efficiency. It is demonstrated that the electrical power consumption of the process can be reduced under a mono voltage profile setting of the EAF.

Keywords: Process modeling
Abstract: Hall-Héroult process is an electrolysis method to produce aluminum at industrial scale. It is based on an electrochemical reaction that requires an alumina dissolution in a bath solution. The hazardous operational conditions make it difficult the development of a sensor for continuous measurement. Moreover, local variations of alumina concentration throughout the pot cell arise during daily operations. This paper presents a modeling procedure to obtain a spatio-temporal dynamic representation of alumina concentration distribution. From the convection-diffusion relations, the alumina source term is analyzed and expanded to obtain the relations between the available signals and the output. The goal is to develop a system that is able to predict the obtained measurement by taking into account transport properties. The model is validated with industrial data and compared with other models.

Keywords: Control and optimization, Mining operations, mineral processing
Abstract: In this paper, a robust control strategy for surface wave mitigation in copper converters is presented. In copper converters, the purification of copper is carried out by injection of air into the molten bath through lateral tuyeres. The constant rate of air injection produces undesirable oscillation and splashing of the bath in the surface diminishing the lifetime of the internal cover. An H infinity mixed sensitivity approach is proposed to robustly control the air injection rate in order to eliminate the modes of oscillation in the surface even in the case when uncertainty in the parameters of the model and noise in the measurements are present. The effectiveness of the proposed approach is shown by simulations and by comparison with a non-robust LQG control strategy.

Keywords: Mining operations, mineral processing
Abstract: Many plants across the globe are taking advantage of control, optimization and automation. The adoption of the fourth industrial revolution has increased the opportunities for industrial applications. This presentation summarizes learnings from solutions implemented in the mineral processing industry. The session highlights the need for solid process control and an APC foundation to maximize plant benefits and enable digital transformation using analytics, big data, machine learning, AI, etc. Ten practical lessons from the operating floor will be shared. The lessons are inspired by mistakes and successes of actual implementations from the plant's point of view. Questions like where to start, what to avoid, what to look for, what to expect, how to prioritize and more will be part of the discussion.

Keywords: Production planning, distribution and logistics, plant management, Mining operations, mineral processing, Control and optimization
Abstract: This article presents a new geometallurgical framework that builds empirical prediction models of key performance indicators of the processing plant and integrates them into a simultaneous stochastic optimization model for mine production scheduling. The prediction models are created by tracking blended rock properties and matching them with observed responses from the operating processing plant. Several applied case studies in a gold mining complex are presented building prediction models of ball mill throughput, steel ball consumptions, and reagent consumptions from collected data in a gold leaching operation. Finally, the integration of these geometallurgical prediction models into stochastic mine production scheduling is demonstrated to monetize the gained insights. Compared with conventional planning, the resulting mine plans increase yearly profit and are more likely to align with production targets.

Keywords: Control and optimization, Data mining and statistical analyses, Mining operations, mineral processing
Abstract: By applying the theory of constraints (TOC) at 23 operations across every product group and globally including Australia, Canada, North America and Mongolia, Rio Tinto has realised significant performance benefits. This has required the centralisation of historian data, the creation of new data transformation and visualisation approaches (using PowerBi and OSI PiVision) while allowing individual sites flexibility to draw from this framework to create bespoke tools. Subsequently identified constraints were either exploited or had their performance elevated. The constraint utilisation (CU) tool calculates the % CU of each critical piece of equipment in the process. The equipment with the highest CU is then identified as the process constraint for which other components are subordinated. Analysis of the results highlights the potential to increase capacity if the constraint is exploited. The tool also allows for a Pareto view of constraint sequences such that the latent or potential value of constraint exploitation or debottlenecking can be determined.

Keywords: Fault diagnosis, process monitoring, Control and optimization, Data mining and statistical analyses
Abstract: A typical Canadian Mining production facility operates 100s of PID control loops and maintains an even larger quantity of production assets as the foundation of an automated production process. Control loop performance monitoring (CLPM) solutions leverage the information stored in a mining facility’s data historian. They actively evaluate PID control loops on a plant-wide basis and identify issues that both undermine production performance and eventually put critical assets at risk of unplanned downtime. This paper leverages insights gained from the near continuous monitoring and assessment of data from a mine located in Quebec, Canada and operated by Agnico Eagle Mines Limited (Agnico Eagle).

Keywords: Mining operations, mineral processing, Process modeling, Automation, instrumentation
Abstract: The representation of hierarchical control strategies within discrete event simulation (DES) enables mine automation by supporting multiphase re-engineering projects. Indeed, simulation-testing allows refinements of the control strategy, while assisting the selection and sizing of equipment, e.g. in debottlenecking a proposed automated haulage system for an underground mine. Moreover, discrete rate simulations (DRS) is a type of DES that is well-suited for the early stages of mine re-engineering projects, since it supports system-wide dynamic mass balances, and alternating modes of operation in response to geological uncertainty; the DES detailing of localized automated technology within a DRS representation of the global mining system ensures that automation design decisions are oriented toward system-wide metrics, which assess the overall competitivity of the mine. The current paper describes an adaptation of DRS/DES formulations for multiphase mine automation projects, allowing successive detailing of critical aspects that distinguish the mining context from other industrial settings.

Keywords: Mining operations, mineral processing, Artificial intelligence, machine learning systems, and human machine systems
Abstract: LaRonde mill is facing two major challenges, the complexity and the variability of the ore being processed along with the aging of its assets. To maintain production efficiency though time, the facility is turning towards artificial intelligence (AI). At the LaRonde mill, efforts were first put forward to collect reliable data owned by its users. The steps undertaken can be split into two categories: business wide management system and data from equipment and people. The road to introduce AI into day-to-day operations is not without its challenges and complexity. The expertise is in constant evolution, requiring added knowledge of computer programming, robotic, network systems, etc. LaRonde is pushing forward to take advantage of the opportunities AI shows, like advanced control, teleoperation and digital twins.

Keywords: Production planning, distribution and logistics, plant management, Manufacturing, assembling, Artificial intelligence, machine learning systems, and human machine systems
Abstract: This work addresses the optimal scheduling for a multi-line steel hot rolling mill with two parallel production lines. Groups of products are manufactured at one of these lines in an alternate sequence. An optimal schedule minimizes the sum of the required setup times between all products and optimally fills up retooling times necessary at one line by processing a group of products at the other line. This requires both sequence optimization and, for some groups, the optimal selection of jobs. The optimization problem is a combination of traveling salesman and orienteering problems and considers all groups simultaneously. To iteratively solve the combined problem, an algorithm consisting of a heuristic simulated annealing and a local search procedure is employed. The effectiveness of the algorithm is analyzed in a case study.

Keywords: Control and optimization, Hot/cold rolling, forming, forging, Advanced process control
Abstract: In this work, the material flowrate control of a hot rolling finishing mill for steel bars is evaluated in simulation. First, a reduced model for control purpose is developed and described. This model uses nonlinear equations from the rolling theory and leads to the nonlinear control structure. The simulations are tested on a hot rolling mill simulator that implements a more detailed process behavior and is outside of the scope of this work. For controlling this plant, a Nonlinear Model Predictive Control (NMPC) is tested in various conditions. It is compared as a benchmark with a classical Linear Quadratic Regulator with integral action (LQI). The controller is implemented with help of CasADi and the control problem is solved by a SQP solver. The NMPC uses an underlying EKF as disturbance observer to handle the reference tracking problem.

Keywords: Control and optimization, Hot/cold rolling, forming, forging, Advanced process control
Abstract: In flat steel rolling, roughing mills are used to roll slabs into thin plates within 5 to 7 consecutive reversing rolling passes with decreasing thickness. Some of these mills are equipped with edger rolls to control also the width of the plate. Edger rolls are usually mounted upstream of the roll gap. In the best case, a perfect cuboid-shaped plate without camber and thickness wedge leaves the roughing mill. In practice, however, lateral asymmetries like temperature gradients and thickness inhomogeneities may cause the plate to rotate in the roll gap with respect to the vertical axis and to form a camber or thickness wedge or both. In the worst case, these effects not just deteriorate the product quality but also jeopardize plant equipment like side guides. To improve the product quality and reduce wear of plant equipment, active control of the motion and camber of the plate is desirable. In this work, a mathematical model for both the planar motion and the resulting shape of the plate is briefly summarized. Based on this validated model, a model predictive control strategy that calculates optimal control inputs (roll gap tilt and lateral position of edger rolls) for all remaining rolling passes is developed. The performance of the proposed control concept is demonstrated in simulation studies.

Keywords: Fault diagnosis, process monitoring, Hot/cold rolling, forming, forging, Ferrous metal
Abstract: Unwanted oscillations are common source of problems in industrial processes that contain rotational elements. The procedures for detecting and isolating such periodic disturbances are often based on the application of FFT and subsequent analysis in the frequency domain. The approach presented here modifies this procedure: all rotating elements in the system are assumed to produce unwanted oscillations, and the resulting hypothetical power spectral density is continuously compared to the real power spectrum generated by the main process variable. When the real oscillations occur, the results of the comparison indicate their probable source(s). The approach is first demonstrated and tested on simulated data and then verified on real recorded data. The method is primarily intended for inspection of a cold rolling mill whose data were used for verification.