Simulating particle agglomeration in the flash smelting reaction shaft
Introduction
The pyrometallurgical flash smelting process is the main production method employed for extracting copper from its sulphide ores. The process involves the oxidation of the sulphur and other unwanted components, mainly iron, at high temperatures with oxygen enriched air. The ore concentrate and other reactant streams are fed through a burner into a furnace that is maintained at a temperature in the order of 1473–1573 K. The feed streams mix, disperse, ignite, and react. The sulphur is oxidised to form sulphur dioxide, which is removed out of the furnace with the waste gases and generally further processed to produce sulphuric acid. The iron present tends to be preferentially oxidised over the copper, as observed by Tsukada et al., 1981, Asaki et al., 1988. These reactions are highly exothermic and can generate sufficient heat such that the process can be operated autogenously. The reacted particles melt and collect at the base of the furnace, the settler region, where the iron oxides combine with introduced silica flux to form a slag phase that is immiscible with the copper-rich phase. These two molten phases separate and are removed from the furnace individually for further processing.
In a typical industrial flash smelting process the reactant streams are fed through a burner from the top of a vertically aligned cylindrical reaction shaft, where the particles heat, ignite, react, and melt as they travel downwards towards a rectangular box-shaped settler. The waste gases are removed via a separate offtake shaft.
Experimental work on the flash smelting process has ranged from industrial (e.g. Kemori et al., 1986, Parada et al., 2006) and pilot-plant (e.g. Asteljoki and Muller, 1987) scale sampling trials, to laboratory scale experiments (e.g. Jorgensen and Segnit, 1977, Jorgensen, 1983, Kim and Themelis, 1987, Perez-Tello et al., 2001a, Stefanova et al., 2004). Numerical models have also been developed in conjunction with the experimental work, with initial attempts involving simple one-dimensional (e.g. Themelis et al., 1988) or two-dimensional (e.g. Ruottu, 1979, Hahn and Sohn, 1990a) representations of the burner and reaction shaft, while advances in computing have allowed the consideration of more detailed fully three-dimensional representations of entire sections of the furnace (e.g. Solnordal et al., 2006a, Solnordal et al., 2006b). However, although this work has lead to an improvement in the understanding, design and operation of the process, further work is still needed to develop a fundamental understanding of control and optimisation (Parada et al., 2006).
One such area is the occurrence of dust in the offtake gas stream from the furnace (Jones and Davenport, 1996). This dust is made up of particles that are small enough to be entrained with the offtake gas instead of falling into the settler. Dust levels are typically in the order of 5 w/w% of the feed stream (Gonzales and Jones, 1993, Jones and Davenport, 1996), which represents a sizeable loss of product. The presence of dust also increases maintenance requirements due to the build-up of accretions in and beyond the offtake shaft (Jones and Davenport, 1996). Consequently, it is desirable to reduce the level of dust production.
There are three mechanisms of dust production, denoted here as: (i) mechanical, (ii) chemical, and (iii) physical. Small particles are formed mechanically when larger particles fragment due to the rapid internal build-up of gas from reaction and vapours from volatilisation at the high particle temperatures (Otero et al., 1991, Shook et al., 1995). Small particles are also formed by the chemical mechanism when volatile components in the gas phase condense (Jorgensen, 1980, Jorgensen, 1985). The physical mechanism does not describe the formation of small particles, but instead describes dust production by the entrainment of small particles that were initially present in the feed, either singly (Yli-Penttila et al., 1998) or collected in clusters that break up in the process (Debrincat et al., 2008a, Debrincat et al., 2008b).
Kimura et al., 1986, Kemori et al., 1988 investigated reacting particle behaviour using a pilot scale flash smelting furnace operated under industrial conditions. They collected water-quenched particle samples from various locations down the axis of the reaction shaft, which were analysed to determine the extent of reaction and size distribution. Their results indicated the occurrence of agglomeration where the average diameter of the particle increased from about in the feed to about at the base of the 4 m furnace. They also found that the larger agglomerate particles had reacted to a greater extent. They proposed that larger agglomerate particles had formed from particles that had reacted, heated-up, and become molten, and which had then collided and combined with other similarly reacted molten particles. They did not expect un-reacted solid particles to combine upon collision, but rather to bounce off each other instead.
This finding of agglomeration seemingly contradicted earlier work by Kellogg and Themelis (1983) who predicted by calculation that the particle number densities were too low for collisions (and subsequent agglomeration) to occur in flash smelting. Themelis et al. (1988) later addressed this contradiction by developing a one-dimensional numerical model of the flash smelting process that predicted the collision and agglomeration of particles, that were fed as molten. Unfortunately few details of conditions and parameter values were given. Nevertheless, their single set of results compared qualitatively well with experimental data and the role of agglomeration of molten particles was made clearer. Notably, they identified agglomeration of molten particles as a method of reducing dust production. They also identified the need for further research to establish the behaviour and potential of agglomeration. However, since then, no further work focused on investigating or understanding the formation of agglomerates within the flash smelting furnace has been conducted (Donizak et al., 2005).
As discussed above, agglomeration has the potential to reduce dust losses from the flash smelting furnace. This work aims to investigate agglomeration in the flash smelting process by developing a numerical model of the reaction shaft that improves on the attempt of Themelis et al. (1988). This model is used to identify important process variables that influence agglomeration.
Section snippets
Turbulent gas-particle flow
Fig. 1 shows the reaction shaft flow geometry within which the agglomeration of molten particles is investigated. The often complex burner inlet flow geometry of industrial reaction shafts is simplified to that of flow through a sudden expansion. As a further simplification, the reaction shaft is assumed to behave in an axi-symmetric and steady-state manner, although the work of Sutalo et al., 1998a, Sutalo et al., 1998b shows the actual behaviour is more likely both three-dimensional and
Standard case
Before considering the effects of variables on the flow, transport, and agglomeration behaviour, a standard case is solved to identify and establish trends predicted by the numerical model described above. The industrially relevant inlet and boundary conditions specified for this standard case are listed in Table 2.
Relevance of the model
Fig. 7 shows that as the particles fall and collide down the 5 m reaction shaft, under the standard case conditions, the average particle diameter increases from an initial value of to a peak value of around . This approximate five fold increase in is closely comparable to the experimental data of Kimura et al., 1986, Kemori et al., 1988 cited previously in Section 1, as well as the numerical predictions of Themelis et al. (1988). The favourable comparison of the present
Conclusions
A steady-state, two-dimensional, axi-symmetric model of turbulent particle-laden gas flow for a flash smelting reaction shaft has been presented that predicts the agglomeration of particles as they melt. This model supersedes the earlier one-dimensional attempt of Themelis et al. (1988), with the extent of agglomeration predicted comparable to their results and the published experimentally collected data of Kimura et al., 1986, Kemori et al., 1988. The particles are found to heat and melt
Acknowledgements
The first author wishes to thank Andrew Campbell, Nic Croft, Andrew Kyllo, and Melissa Trapani for helpful discussions. Financial support from an Australian Postgraduate Award (for DRH), BHP Billiton, and the University of Melbourne is gratefully acknowledged.
References (64)
Collision rates of small particles in a vigorously turbulent fluid
Chemical Engineering Science
(1975)- et al.
Characterisation of inter-particle forces within agglomerated metallurgical powders
Powder Technology
(2008) - et al.
Influence of particle properties on the size of agglomerated metallurgical powders
International Journal of Mineral Processing
(2008) - et al.
On the solution of population balance equations by discretization: 1. A fixed pivot technique
Chemical Engineering Science
(1996) - et al.
The numerical computation of turbulent flows
Computer Methods in Applied Mechanics and Engineering
(1974) - et al.
Self-sustained oscillation of a submerged jet in a thin rectangular cavity
Journal of Fluids and Structures
(2001) - et al.
Lagrangian and Eulerian models for simulating turbulent dispersion and coalescence of droplets within a spray
Applied Mathematical Modelling
(2006) The description of a mathematical model for the flash melting of Cu-concentrates
Combustion and Flame
(1979)- et al.
CFD modelling of the flow and reactions in the Olympic Dam flash furnace smelter reaction shaft
Applied Mathematical Modelling
(2006) - et al.
Oxidation of molten copper matte
Metallurgical Transactions B
(1988)
Transport Phenomena
Velocity gradients and internal work in fluid motion
Journal of the Boston Society of Civil Engineers
Heat Transfer: A Practical Approach
PHYSICA – a software environment for the modelling of multi-physics phenomena
Zeitschrift fur Angewandte Mathematik und Mechanik (ZAMM)
On the evolution of mathematical modelling of single-step flash smelting of copper concentrates
Progress in Computational Fluid Dynamics
Radial and axial turbulent flow measurements with an LDV in an axisymmetric sudden expansion air flow
Journal of Fluids Engineering
On predicting particle-laden turbulent flows
Applied Scientific Research
Simulation of turbulent swirl flow in an axisymmetric sudden expansion
AIAA Journal
Numerical simulation of unsteady turbulent flow in axisymmetric sudden expansions
Journal of Fluids Engineering
Mathematical modeling of sulfide flash smelting process: part I. Model development and verification with laboratory and pilot plant measurements for chalcopyrite concentrate smelting
Metallurgical Transactions B
Mathematical modeling of sulfide flash smelting process: part II. Quantitative analysis of radiative heat transfer
Metallurgical Transactions B
Heat transfer mechanism in ignition of nickel sulphide concentrate under simulated flash smelting
Proceedings of the Australasian Institute of Mining and Metallurgy
Single particle combustion of chalcopyrite
Proceedings of the Australasian Institute of Mining and Metallurgy
Vapourization during the combustion of a complex copper concentrate
Bulletin of the Proceedings of the Australasian Institute of Mining and Metallurgy
Copper flash smelting simulation experiments
Proceedings of the Australasian Institute of Mining and Metallurgy
Cited by (28)
Distribution and evolution of particles in flash converting furnace under different operational conditions
2023, Transactions of Nonferrous Metals Society of China (English Edition)CFD modelling and optimization of oxygen supply mode in KIVCET smelting process
2019, Transactions of Nonferrous Metals Society of China (English Edition)Optimal control strategy of working condition transition for copper flash smelting process
2016, Control Engineering PracticeCitation Excerpt :In this paper, we assume that minor species such as PbS and ZnS in the matte and Cu2O, PbO, ZnO, CaO, and Al2O3 are not considered, and the SiO2, N2 and H2O are chemically inert. In fact, several works have been described change process of major species and temperature by using mass/energy balances method (Liu et al., 2012; Higgins, Gray & Davidson, 2009; GUI et al., 2007). Based on desulfurization ratio of copper concentrate, the developed model (Liu, Gui, Xie & Yang, 2014) couples dynamic mass balances on each species with equilibrium relationships for major component (Cu, Fe, S, SiO2, et al.) to form a system of differential and algebraic equations, which verify the effectiveness.
Dynamic modeling of copper flash smelting process at a Smelter in China
2014, Applied Mathematical ModellingCitation Excerpt :The flash smelting process is widely used throughout the world for copper production, accounting for about 50% of global capacity for primary copper production [1].
Sticking of iron ore pellets during reduction with hydrogen and carbon monoxide mixtures: Behavior and mechanism
2013, Powder TechnologyCitation Excerpt :A. Aran had revealed that the higher the gangue amount in ores the smaller the sticking tendency was. On the other side, ores with higher gangue portion might cause lower melting temperature which could also induce sticking greatly [8,12,13]. In the aspect of reducing agent, the pore structure of iron obtained by CO reduction was coarser than that by H2 reduction and the result of small additions of hydrogen was that the fibrous iron growth decreased and stopped if the hydrogen addition increased greatly [8,14,15].