Report on Computational Modelling '19

Computational Modelling '19 was held at the National Maritime Museum, Falmouth, UK, from June 11-12. The 7th in the series, which began in Cape Town in 2005, this was by far the smallest, with an attendance of 26 delegates representing 12 countries.

This begs the question as to whether this conference series has ran its course. Computational modelling techniques, such as DEM and CFD are now accepted tools in mineral processing, and their use is a feature of all current MEI Conferences, so is there a need for a small specialised conference dealing with developments and uses of these techniques? The argument for is that the conferences do have a multi-disciplinary flavour, the papers this year covering liberation and comminution, flotation, hydrometallurgy, pyrometallurgy, dewatering and materials handling. It is always useful to bring scientists from disparate groups together, as it promotes lateral thinking, but I would like the views of those who attended this year and in previous years.

The following is my report on the papers presented. More detail on people and actvities are in the posting of 12th June.

Tuesday June 11th

Paul Cleary

Particle based modelling is well suited to predicting complex multiphase flows within mineral extraction processes. Applications include comminution with particle breakage and flow in crushers and particle breakage and slurry transport within mills and separation, both wet and dry, by screens. In such models the coarser particulate phases are represented by DEM (Discrete Element Method) and the finer slurry or powder phases by SPH (Smoothed Particle Hydrodynamics). Emerging opportunities also exist in flotation where discrete phases for bubbles and particles can be coupled to continuous free surface liquid phases. These models have become quite sophisticated and continue to increase in the range of scales modelled and the fidelity of the physics represented as computer power continues to rise and computational modelling codes continue to develop. A fine keynote lecture from Paul Cleary, of CSIRO Data61, Australia, showcased a range of leading edge application examples of this modelling and discussed the increasing role of hybrid and multiscale methods.

Process simulators have become increasingly useful for design, optimization and control in mineral processing, evolving from simple models of isolated unit operations to complex integrated simulation environments. There has been an increase in the number of mineral processing simulation platforms available and publications on their application; however, a recent overview of the topic is missing in the literature.

Juliana Segura-Salazar

Juliana Segura-Salazar, of Imperial College, UK,  presented a critical review of the state-of-the-art of process simulators for the minerals industry, discussing and contrasting their capabilities and limitations, as well as identifying some of the challenges that need to be addressed. She discussed recent developments towards the application of these simulators in a more systemic context as a support tool for decision making within the minerals sector. This is particularly relevant if process simulation is to play a key role in overcoming the multiple challenges faced by mineral processing operations in the context of sustainability.

Luis Cisternas

As highlighted by Luis Cisternas, of Universidad de Antofagasta, Chile, mineral processing usually utilizes multi-unit separations because complete separation is seldom achieved in one unit. The analysis and design of these circuits have been an active area of research in the exploration for improvements in mineral processing. Linear circuit analysis (LCA) has been proposed as a simple tool for the analysis of mineral processing circuits. Clearly, LCA is a tool with several advantages, but it does have its limitations and disadvantages. Luis analysed some limitations and disadvantages of LCA. The results show that LCA must be used carefully because its application can introduce errors and incomplete analysis.

Component wear is a key consideration for the minerals industry given the abrasive nature of particle-laden flow such as mineral slurries. Studying wear through experimentation is time-consuming and expensive, thus, an accurate simulator to understand and predict wear is highly desirable. This allows for optimisation of design and operating parameters to improve performance, as well as enhance the service life of these parts. Lily Ip, of Imperial College, UK, described a finite element based CFD wear simulator, which includes the ability to model coupled solid-fluid motion of slurries through representative Lagrangian particles and perform dynamic remeshing of geometry in response to wear and moving components. The capabilities were illustrated through examples ranging from simple geometries to that of complex moving geometries such as centrifugal pumps.

Lily Ip with Stefan Kirsch

Taking us to the lunch break, Stefan Kirsch, of Robert Bosch Packaging Technology B.V., The Netherlands and Technical University Dresden, Germany, discussed how vertical discharge and filling processes of granular media have been successfully described with the aid of DEM in the past. However, a major challenge is often the identification of accurate interaction parameters (model calibration) such as friction and restitution coefficients. In-situ bulk calibration has been shown to be an efficient approach to generate high fidelity models. However, Stefan showed that in the case of vertical transfer and especially when working with low particle counts, randomness intrinsic to the process, can be detrimental to efficient and reliable implementation. A second constraint is shortcomings in the physical model of DEM and differences between lab scale and industrial equipment. These factors can lead to systematic errors in the calibration process and may restrict model transferability. Stefan described a study where several example media were assessed to deduce suitable calibration and quantitative validation strategies for DEM models.

Grinding and liberation were the topics discussed in the five papers in the afternoon session.

Feed size distribution is one of the important parameters that affect ball milling and can be tailored to optimally produce a desired mill product. To that end, Ngonidzashe Chimwani, of the University of South Africa, described how batch tests were used to investigate the effect of feed tailoring in ball milling using a South African gold ore. Narrow-size feeds of the ore were prepared and milled with sets of single-size balls. From the experimental data obtained, milling parameters of the gold ore were determined, validated and then used to simulate the breakage patterns of different feed size distributions for further analysis, with the purpose of finding the feed size distribution that produces the highest amount of the desired size range for gold leaching.

Ngonidzashe Chimwani with Bertil Palsson and Jarrod Hart

Discrete element methods (DEM) have provided the ability to resolve the complex phenomena experienced by ore within comminution devices such as tumbling mills. The new developments in DEM techniques and the corresponding increase in computational power has made it more feasible to study the movement of individual ore particles as they traverse a tumbling mill. A range of information can be extracted on individual particle impact histories, such as kinetic energy at the time of collision, velocities of individual particles, per-particle stress tensors, impulse and collision energy. DEM codes have been modified to provide the required information that can be analysed in an associated software to investigate the impact of varied operational parameters on the resultant granular dynamics, and we were looking forward to Olumide Ogunmodimu, of JKMRC, Australia, exploring the reaction of individual ore particles to the breakage environment within a tumbling mill using DEM techniques, studying the stressing conditions that would lead to either impact or abrasive breakage. Unfortunately, however, entirely due to no fault of his own, Olumide was unable to attend the conference but his paper is available on the Proceedings USB.

Simon Larsson

Modelling of wet grinding in stirred media mills requires the simultaneous modelling of grinding media, a moving internal stirrer, and the pulp fluid. Simon Larsson of Luleå University of Technology, Sweden, showed how wet grinding in a stirred media mill was simulated using coupled incompressible computational fluid dynamics (ICFD) and discrete element method (DEM) and finite element method (FEM) simulations. The grinding media was modelled with the DEM, the pulp fluid flow using the ICFD and the mill structure using the FEM. The combined model was used to predict the wear rate in the system. The model can be used to study process parameters and their effect on the wear map distribution and power consumption. The two-way fully coupled ICFD-DEM-FEM preserves the robustness and efficiency, and it allows the use of large time steps for the fluid with very low computation times.

The main roles of liners are to protect the mill shell and promote effective ball motion for grinding. For this reason the liner profile is carefully selected to ensure that the productivity is maximized and due liner replacement is made when this objective is no longer met. These issues have been extensively studied on shell liners as mill relining is a significant cost component of ball milling. To date, not much has been written about end-liners and the kind of forces they are subjected to. In his second paper,   Ngonidzashe Chimwani described a DEM simulation scheme which was conducted to look at how ball size distribution, mill filling, end-liner configuration and shape affect the distribution of forces acting on the few liners that were used as a case study to understand end-liner wear and damage.

Mapilane Madiba

In mineral processing plants there is a challenge in turnaround time for issuing results timely to reduce delays in the decision-making process. Mineral liberation is considered as one of the efficiency drivers for downstream processes such as flotation. Thus, delayed decision could be costly. Mapilane Madiba of University of South Africa, described the development of a procedure to quantify and minimise the time analysis for mineral liberation.

After a good first day of presentations, a number of hardy souls braved the cold and windy conditions for a 3 mile walk around the Pendennis Headland and to awaiting drinks in one of Falmouth's oldest pubs, the Chain Locker.

In the 16th century Pendennis Castle moat

Welcome ale in the Chain Locker

Wednesday June 12th

Mapilane Madiba of University of South Africa was back on the podium first thing to further review correction methods to solve limited representativeness in mineralogical analysis of measured liberation data. Although progress is being achieved, there are still challenges facing correction methods. Therefore, the advantages and disadvantages of specific issues, such as point spread function, excitation energies and the monochromatic assumption of X- Ray source, facing the correction methods were discussed.

Flotation then dominated the morning session. Luis Cisternas, of Universidad de Antofagasta, Chile discussed how regrinding of rougher concentrate, scavenger concentrate, and/or cleaner tailing is common practice to liberate valuable mineral attached to gangue particles. However, the methodologies proposed for the design of flotation circuit based on optimization usually do not consider regrinding. Luis analysed the effect of regrinding in the design of flotation circuit structures.

Guichao Wang

Guichao Wang, of Southern University of Science & Technology, China, showed how direct numerical simulations of a particle-bubble collision system composed of monosized spherical solid particles and air bubbles in a quiescent liquid and homogeneous isotropic turbulence have been performed. Particle-bubble collisions in a quiescent liquid were first simulated and compared to the existing experimental work of particle-bubble collisions. A DNS model for studying the effect of turbulence on the collisions between particles and bubbles was then developed. Turbulence was found to first increase particle-bubble collisions to a certain extent, and, with further increase turbulence intensity, the particle-bubble collision rate was reduced.

Phil Schwarz described how the flow field, gas dispersion and solids concentration in a flotation cell fitted with an Outotec flotation mechanism were studied at CSIRO Minerals Resources, Australia, using both experimental and multi-phase CFD modelling. He showed that the complex multi-phase CFD model can be used to predict the major characteristics determining flotation kinetics: namely turbulence, bubble size and distribution, and solids distribution. The CFD flotation model can therefore be used for investigations on the design and operation of the cell, for example, in parametric studies on the effect of stirring speeds and aeration.

Phil Schwarz (centre) with Guichao Wang and Russia's Dmitrii Maiorov
at the Chain Locker

Asmaa Hadane

Flotation is an important separation process for beneficiation of sedimentary phosphate rocks. In order to evaluate hydrodynamic efficiency, Asmaa Hadane, Mohammed V University in Rabat and Mohammed 6th Polytechnic University, Morocco, showed how a CFD multiphase model of water-phosphate flow was used to investigate the main hydrodynamic criteria namely: the agitator power consumption and the solid distribution in the flotation cell depending on the impeller rotational speed and the phosphate volume fraction.  A second CFD study considered the injection of air bubbles. She described work on the air dispersion and the solid distribution in the complete cell volume, in order to analyse the phosphate-air contacts within the flotation tank, which represent the key elements to establish convenient conditions to maximize the flotation performance.

Several ore deposits are polymetallic with two, three or more base metals. These base metals include copper, lead, zinc, cobalt, nickel, molybdenum, and pyrite among others. Usually, polymetallic ore is separated by sequential flotation circuits for each base metal. In his 3rd paper of the conference, Luis Cisternas argued that the methodologies proposed in the literature for the design of flotation circuits consider only one base metal and there is no literature for the design of integrated flotation circuit for polymetallic ores. Luis proposed a procedure for the design of polymetallic integrated flotation circuits, the methodology considering a superstructure to represent the set of feasible flotation circuits, disjunctive expressions for the selection of operation conditions of flotation stages, and a mixed integer nonlinear programming model. The integrated flotation designs were compared with sequential flotation circuits, and advantages and disadvantages identified. 

A better understanding of controlling factors in mineral flotation will undoubtedly be possible using advanced CFD simulations over entire full-scale industrial cells. Such macro-scale simulations require sub-models to quantitatively describe the rates of bubble-particle attachment and detachment: the huge number of particles and bubbles in an industrial cell mean that macro-scale models cannot resolve the details of bubble-particle interaction. Micro-scale models and experimental investigations can be used to improve sub-model formulas for attachment and detachment rates. In his second paper, Phil Schwarz, of CSIRO, showed how micro-scale models of bubble-particle interaction are extended to account for near-field hydrodynamic effects, ie effects that have influence on particle motion near the bubble interface. These near-field effects include the mobility of the bubble interface (which influences the shear rate near the interface); the so-called lubrication force (which slows the particle approach); lift force; and modification to the drag force on the particle. The micro-scale CFD simulations show that the lubrication force is the most significant of these additional affects, and that the degree of interface mobility has an important influence on the magnitude of the other additional forces. Recommendations for sub-models for use in macro-scale CFD models were discussed.

Yann Foucaud

Continuing with the flotation theme after lunch, Yann Foucaud, of Université de Lorraine, France, stressed that understanding the adsorption mechanisms of reagents is a key step to enhance flotation. New depressants and collectors, which are more efficient, selective, and environmental friendly have been developed, but few experimental methods can currently identify the surface molecular mechanisms with accuracy and confidence. Yann showed that atomistic simulations aid understanding of the mechanisms involved in reagent adsorption.

Derek Machalek

 

After Yann's presentation, there was a complete change of direction. Derek Machalek, of the University of Utah, USA, showed that rotary kilns require large fans to blow air through them to support processing of minerals and if fan controls can be modified to respond to rapid changes in electric demand, they can become valuable grid assets. Due to their considerable thermal inertia, kilns and the associated fans are traditionally operated continuously in a steady manner to avoid process disruption. However, advanced model predictive control (MPC) can allow for the rigid process to be operated flexibly. Ultimately the flexible process can respond to grid demand requirements by ramping the kiln fan.

In a further paper Derek showed how the evolution of the electrical grid requires flexibility in electricity consumption. Given the tremendous amount of electricity consumed by mineral processing, these facilities could become a grid asset if they can leverage sources of flexibility. While facilities generally try to maximize ore throughput, de-watering represents one source of flexibility due to the holdup capacity of the water table itself and storage tanks to which the water is pumped. A second source of flexibility is intermediate product transportation with storage capabilities at each end. By developing predictive automation algorithms, these holdup capacities can be effectively leveraged. This makes the facility able to respond to grid signals, which can save on demand charges, while also becoming an asset to the grid.

Alireza Eslamian

Alireza Eslamian, of ESS Engineering Software Steyr GmbH, Austria, presented a novel coupled DEM-SPH method that simulates agglomeration and spheronization of powder mixtures to produce pellets. Optimal design of iron ore pelletizers, which allows generation of pellets of uniform size and form, is a common challenge and demands a deep insight into the physics of powder mixture accumulation and agglomeration. Iron ore pellets are spheres ranging between 6 and 16 mm, used as feed for blast furnaces, and usually contain 64-72% Fe, while the residuals consist of limestone, dolomite, and bentonite as binder.  Alireza said that no method for simulating the formation of pellets out of raw material mixtures in pelletizers has yet been proposed. He showed how a coupled DEM-SPH method has been developed by ESS. SPH is used to simulate the raw material mixture as a continuum media, while DEM simulates the formation of pellets in connection with raw material mixture.

It is standard practice to never stop the drive on a thickener, even during a shutdown, unless the bed in the thickener is drawn down to such an extent that the soft bed remaining will not pose a problem during restart. To avoid this, thickeners are typically equipped with an uninterrupted power supply to ensure non-stop raking. In the final paper of the conference Justin Jacobs, of Paterson & Cooke Ltd, USA, described how a client was interested in being able to stop and restart the thickener rake as needed, without drawing down the thickener bed. For this case, hydraulic transport and deposition of the thickener underflow required a specific rheology, eliminating the drawing down option. Justin showed how a complete rheology characterisation, using CFD established the drive size required to make a thickener restart under load possible. The CFD model was validated using an existing thickener rake with a known geometry, fluid rheology, and motor torque measurements.

Justin Jacobs (right) relaxing with delegates at the Chain Locker

Despite the low numbers this has been a productive conference and I will, of course, advise if we intend to continue with this series. All the papers presented are available on USB from MEI.

Twitter @barrywills