Original Research Papers
Simulation of Mechanical Degradation of Iron Ore Pellets in a Direct Reduction Furnace
https://ror.org/03490as77 
2018 Volume 35 Pages 216-225
Details
2018 Volume 35 Pages 216-225
The increase in production of steel in electric arc furnaces in recent years influenced directly the world production of direct reduction iron (DRI). Amongst the most widely used technologies for DRI production is the MIDREX® process. The behavior of the metallic charge used in these furnaces, mainly made up of iron ore pellets, influences directly their productivity. Fines contained in the feed are typically removed by screening, in order to prevent them from impacting negatively the productivity of the furnace. However, fines still may be generated inside the furnace as a result of the collisions amongst pellets and between them and internals of the furnace as they move downwards from the feed to the discharge of the shaft furnace. Predicting and preventing such mechanical degradation is critical in the furnace operation. The present work deals with the prediction of degradation of iron ore pellets during reduction in a MIDREX furnace. Collision energies involved in the vertical flow of pellets along a MINIMOD® MIDREX direct reduction furnace were estimated using the discrete element method. Using this technique along with a model of degradation that considered assumptions based on information from the literature it was possible to estimate the proportion of fines generated inside the reactor, which was consistent with plant practice. Finally, it was predicted that the generation of fines in the reduction zone of the furnace would vary from 3.4 % to 4.6 % as the throughput of the furnace increased in 50 %.
Because of the reduced supply of good quality lump iron ore in the world market and the growing environmental constraints to sintering, iron ore pellets began to occupy an important role as raw material for primary iron production. Their participation in the production of pig iron and direct reduction iron (DRI) has increased year after year due to their advantages in comparison to lump ore. Some of the advantages are their higher uniformity in size, greater control of chemical composition, besides their reduced amenability to mechanical degradation during handling and under reduction.
The increase in production of steel using electric arc furnace impacted directly the production of DRI, raising it from 40 Mt in the beginning of this century to 74 Mt in 2014 (MIDREX, 2015). Several processes are used in the production of sponge iron, however the most widely used is the MIDREX (Midland Ross Experimental) process, which responds for more than half of the world’s DRI production (MIDREX, 2014). In this process lump ore and/or iron ore pellets as used to compose the metallic charge, with iron ore pellets contributing to improve operational stability and increase productivity and/or metallization of charge, making this raw material the most attractive charge for producers of sponge iron in MIDREX furnaces (MIDREX, 2014).
In spite of their greater resistance to degradation, iron ore pellets still suffer from this problem, although to a more limited extent when compared to lump ore. Such degradation can take place either during handling from the pellet producer until loading the furnace, or under reduction, generating either coarse fragments or fines. Fines that are generated by handling prior to loading the furnace can be removed by screening. However, mechanical degradation can also occur inside the direct reduction furnace, as the result of the combined effect of weakening of the pellet structure due to reduction (Huang et al., 2012) and to the stresses which pellets withstand until they are discharged at the bottom of the furnace. When such mechanical degradation occurs inside the furnace, either coarse pellet pieces or fines may be produced. Indeed, such fines have been estimated to represent between 2 and 4 % in weight of the product, based on measurements in plants that screen the furnace discharge (MIDREX, 2011). Such fragments can cause loss of furnace productivity because they affect negatively the permeability of the pellet bed, thus resulting in lower metallization of the charge. In addition to that, these fines can contribute to the pellets sticking to each other resulting in the formation of clusters, which besides contributing negatively to the permeation of reduction gasses, also create a disturbance in the flow of the charge (Wong et al., 1999; Basdag and Arol, 2002). Therefore, prediction of the amount of fines that are generated inside the reactor owing to the use of a particular type of iron ore pellet would be useful. This information could be used to assist in selecting the optimal feed blend to the direct reduction reactor, besides enabling the choice of the operational parameters that can reduce the impact of the formation of fines. However, no study in the published literature has been found that addressed simulating such mechanical degradation of iron ore pellets in direct reduction furnaces.
The present work uses a combination of the discrete element method (DEM) (Weerasekara et al., 2013) and models that describe pellet breakage and attrition to predict iron ore pellet degradation and fines generation in a vertical shaft direct reduction furnace in a one-way coupling strategy.
The overview of the simulated direct reduction furnace is presented in Fig. 1. The furnace consists of a feeding system, containing 16 conduits (called legs), a first section, named reduction zone, on the bottom of which the reformed gases are introduced. The charge then travels downwards and encounters the upper burden feeders, which are axles that are meant to oscillate at predetermined angles at a speed that is adjusted to promote uniform transverse movement of the load and to limit the size of clusters formed. The identification of the sections and the corresponding numbers of pellets involved in the simulations can also be found in Fig. 1. The charge then encounters the region where the cooling gases are injected, to then find the sets of middle and lower burden feeders, until it is discharged from the bottom of the furnace.
Overview of the MINIMOD® MIDREX shaft furnace (left) and view of the simulated sections (middle and right). The number of particles corresponded to the 25 mm spheres simulated in each section.
The furnace simulated is designated MINIMOD®, and it has the smallest production capacity among the industrial furnaces that use the MIDREX technology (Atsushi, 2010). With 4.25 m in diameter, it can produce about half a million tons of sponge iron per year. The selection of this scale in the present work was due to challenges associated to the number of pellets that are present in this furnace at any time, which is in the order of several millions, which is only a fraction of the billions that occupy the 2-million ton-per-year Super Megamod® furnace measuring 7.15 m in diameter (Atsushi, 2010).
In spite of being the smallest industrial scale furnace, the MINIMOD® still contains an exceedingly large number of particles for a computationally cost-effective simulation using DEM. In order to reduce the number of particles within the simulation, the furnace was first split from top to bottom into four sections for simulation in separate, which were numbered 1 to 4 (Fig. 1). In order to further reduce the number of particles, only a fraction of the volume, with periodic boundaries, was simulated in sections #1 to #3. The region around the lower burden feeders (#4), on the other hand, was simulated with the full particle population.
Still with the aim of reducing the computation effort in the simulations pellets were assumed to present spherical geometry, which is a reasonable assumption in the case of iron ore pellets. In addition to that, pellets with nearly twice the average diameter of the actual pellets usually used industrially (9–19 mm), that is, 25 mm, were simulated. In spite of their uniform shape and size, no crystallization structures were identified in simulating the different sections of the furnace.
The software used in the DEM simulations was EDEM® by DEM Solutions (Edinburgh, UK), version 2.7. In all simulations the Hertz-Mindlin contact model available in EDEM® was adopted, and only gravity was considered as external force. The material parameters used in the simulations were those estimated on the basis of detailed testing by Barrios et al. (2013) with unreduced fired iron ore pellets. It is assumed that the contact properties would not change as pellets undergo reduction in the furnace. These and other material properties (density and shear modulus) are summarized in Table 1.
Hertz-Mindlin contact and material parameters of iron ore pellets (Barrios et al., 2013).
| Steel | Pellet | |
|---|---|---|
| Poisson’s ratio | 0.3 | 0.25 |
| Shear modulus (MPa) | 7000 | 16 |
| Density (kg/m3) | 7800 | 3948 |
| Steel-pellet | Pellet-pellet | |
| Coefficient of restitution | 0.48 | 0.39 |
| Coefficient of static friction | 0.49 | 0.50 |
| Coefficient of rolling friction | 0.21 | 0.25 |
The speed of descent of metallic charge for each section was determined based on the simulated reactor with production of 58 t/h, which would correspond to an annual production in the order of half of a million tons. In addition to that, simulations were also conducted considering an increase in 50 and 100 % in the throughput of the system, resulting in simulated production of 86.8 and 116 t/h. These scenarios would correspond to cases when the residence time of the pellets in the furnace would be reduced significantly. The three systems of burden feeders were inserted in the model and the dimensions and modes of operation of these parts that turn were programmed on the basis of the parameters listed in Table 2.
Operating parameters used for the burden feeders.
| Position | Angle of rotation (°) | Period of rotation (s) | Frequency (rad/s) |
|---|---|---|---|
| Upper | 45 | 180 | 0.0044 |
| Middle | 90 | 30 | 0.0524 |
| Lower | 60 | 10 | 0.1047 |
| Position | Shaft radius (m) | Burdenfeeders radius (m) | |
| Upper | 0.228 | 0.487 | |
| Middle | 0.080 | 0.175 | |
| Lower | 0.069 | 0.160 |
For instance, the computational effort was such that in order to simulate 25 seconds of operation of the charge in section #1, approximately 240 hours of processing in a state-of-the-art workstation were required. Indeed, simulations were conducted until steady-state conditions were reached. Upon completion of the simulations, data from each collision, discriminated between the normal and the tangential component, were collected and processed using a post-processing routine developed in the authors’ laboratory (Carvalho, 2013). Energies in the collisions were first equally split between each pair of pellets involved and then divided by the simulated particle weight, resulting in the specific collision energy. Fig. 2 shows that the magnitudes of the specific collision energies are very similar for the 25 mm pellets simulated throughout the work and the 12.5 mm here analyzed in the lower burden feeder region of the furnace, thus demonstrating the validity of considering the larger pellet sizes in the remainder of the simulations.
Cumulative distribution of specific energy dissipation in the lower burden feeder section from simulations considering pellets of different diameters (12.5 and 25.0 mm).
From the DEM simulations it was possible to analyze the profile of the compressive forces in the sections studied. Fig. 3 shows the influence of the depth in the pellet column inside the reduction zone (section #1) on the magnitude of compressive forces. In the following section (#2), shown in Fig. 4, the force applied was transferred from the previous section in order to allow simulating the entire column of pellets to which that section was subjected. In the figure it is observed that the magnitudes of the forces are similar to those presented on the basis of the column of reduction zone and the burden feeders work as a support to the column of pellets, withdrawing from the regions beneath it the effect of load from the column of pellets above. In sections #3 and #4, on the other hand, this method was not used, given the ability of the burden feeders of supporting the charge, reducing the magnitude of the contact forces (Fig. 5).
Simulation of the reduction zone (section #1), showing detail in the bottom layer of the column with the profile of the compressive forces.
Simulation showing the compression forces in section #2, influenced by the presence of the upper burden feeder.
Simulation showing the compressive forces in section #4 influenced by the presence of the lower burden feeder.
The energy dissipations in the collisions were analyzed in each section, with the larger frequency of higher energy collisions found closer to the bottom of the reduction column (Fig. 3), showing that there is a relative effect of depth of the pellets column. This effect becomes evident in Fig. 6, where the variation of the average normal energy dissipated in the collisions as a function of height and radius of furnace becomes evident.
Normal average energy transferred to particles as a function of normalized depth (0: bottom of reduction zone; 1: top) and radius of the section #1 of the reactor.
A summary of the energy dissipation spectra in the different sections of the furnace is presented in Table 3 for comparison. It becomes evident that the highest energy dissipation occurs in the reduction zone (section #1), whereas the sections around the burden feeders (sections #2–4) exhibit very low energy dissipations (Fig. 7).
Summary of the total energy dissipation distributions (in J/kg) of collisions by furnace section for a throughput of 58 t/h.
| Statistic (J/kg) | Furnace section | |||
|---|---|---|---|---|
| #1—Reduction Zone | #2—Upper Burdenfeeder | #3—Middle Burdenfeeder | #4—Lower Burdenfeeder | |
| Mean | 0.214 | 0.0013 | 0.0018 | 0.0023 |
| Median | 0.0035 | 6.82E-05 | 1.85E-04 | 2.99E-04 |
| 95th percentile | 0.294 | 0.0042 | 0.089 | 0.0103 |
| Standard deviation | 2.032 | 0.0094 | 0.0059 | 0.0093 |
Cumulative distribution of total energy dissipated in the four sections of the furnace simulated. Also shown for comparison are the fracture energy distributions of unreduced and estimated values after complete reduction.
An approach often used to increase the productivity in direct reduction furnaces is to increase the flow of reforming gases coupled to the throughput of the furnace. However, an adverse effect of this is the increase in the degradation of the pellets inside the furnace (MIDREX, 2011). Table 4 compares predictions of the collision energy spectrum of the base case to two hypothetical cases in which the throughput was raised, showing that important increases in the average collision energies would result.
Summary of the energy dissipation spectra (in J/kg) of collisions for section #1 of the furnace as a function of throughput.
| Statistic (J/kg) | Estimated throughput | ||
|---|---|---|---|
| 58 t/h | 87 t/h | 116 t/h | |
| Mean | 0.214 | 0.238 | 0.251 |
| Median | 0.0035 | 0.0039 | 0.0038 |
| 95th percentile | 0.294 | 0.318 | 0.321 |
| Std. dev. | 2.032 | 2.536 | 2.539 |
Depending on the magnitude of the collisions that happen amongst the pellets and between the pellets and the walls and furnace internals as the charge moves downwards from the feed to the discharge, breakage of pellets may occur. Breakage of pellets can occur catastrophically, which is called body breakage, every time a pellet loses more than 10 % of its original weight in any event (Tavares et al., 2015). On the other hand, pellets may suffer only surface breakage or attrition, also called breakage by abrasion or chipping, which results in generation of fines, besides leaving the core of the original pellet nearly undisturbed, since it contains no less than 90 % of the mass of the original pellet prior to the collision event. In this later case, it is possible that the pellet becomes more amenable to breakage in a future stressing event. A model framework that has been developed to describe this in the context of degradation of ores during handling is described elsewhere (Tavares and Carvalho, 2011) and elements of it are here used to predict breakage in the MINIMOD MIDREX furnace.
The threshold between these two modes of breakage defined above is given by the mass-specific particle fracture energy of the pellets (Tavares et al., 2015). Fig. 8 presents the distributions of specific fracture energies for fired pellets contained in different size classes under slow compression conditions. A comparison between these and the distributions presented in Fig. 7 suggests that no pellets would suffer catastrophic breakage inside the furnace, since the energy required to break the weakest fired pellet is higher than the collision energy of highest magnitude inside the furnace. However, as pellets move downwards in the furnace, they undergo reduction, having their strength reduced. Work by Huang et al. (2012) demonstrated that the compressive strength—and therefore the fracture energy—of iron ore pellets drops significantly as reduction occurs (Fig. 9). As such, considering such estimate that particles loose in the order of 75 % of their compressive strength upon reduction (Huang et al., 2012) and assuming that the mass-specific fracture energies would be reduced in the same magnitude, it would be more likely that breakage of pellets might occur in the reduction section of the furnace (section #1). However, even considering that reduced pellets would further lose their fracture energies upon undergoing repeated stresses in the furnace, this effect may be considered negligible under the conditions simulated (Fig. 7).
Distribution of specific fracture energies of fired iron ore pellets.
Effect of degree of reduction on the compressive strength and surface hardness of iron ore pellets (data from Huang et al., 2012). Copyright: (2012) Powder Technology.
It is important to recognize, however, that the presence of coarse fragments, generated during handling prior to feeding the furnace, may be responsible for the appearance of lower fracture energies particles than well-formed pellets, which could still undergo body breakage inside the furnace.
Although unlikely to suffer body breakage, pellets inside the furnace will be subjected to stresses that will certainly be responsible for surface breakage or attrition. A model that describes the relationship between the stressing energy and the generation of fines as the result of surface breakage is (Cavalcanti, 2015)
| (1) |
where k is the proportion of fines generated in each collision event, Eo is a reference specific collision energy, taken as 10 J/kg, and a and λ are dimensionless fitting parameters to be calibrated from experimental data. It is worth noting that in the case where the constant λ is equal to 1, this model becomes equivalent to the one proposed by Ghadiri and Zhang (2002). E is the effective mass-specific energy in each collision, obtained from a combination of the normal component of the energy dissipated in the normal direction of the collision added to 57 % of the contribution from the shear component. This composition of the collision energy was estimated on the basis of impacts of pellets against angled targets (Cavalcanti, 2015). Fig. 10 presents data for an unreduced iron ore pellet, comparing it to the model fit, resulting in the values of a and λ equal to 0.00029 and 1.16, respectively. The fitted model parameters, however, are only valid for unreduced pellets. Huang et al. (2012) demonstrated that the hardness of an outer layer of the pellet reduces as pellets undergo reduction (Fig. 9). Recent work in the authors’ laboratory (Cavalcanti, 2015) proposed a relationship between the Vickers hardness of the region close to the surface of unreduced iron ore pellets to their amenability to surface breakage (Fig. 11).
Experimental results and model fit for fines generation due to attrition of fired pellets as a function of effective collision energy (Cavalcanti, 2015). Model coefficient of correlation (R2) of 0.97. Copyright: (2015) Universidade Federal do Rio de Janeiro.
Relationship between parameter a from Eqn. 1 and Vickers hardness in the periphery of unreduced iron ore pellets (Cavalcanti, 2015). Copyright: (2015) Universidade Federal do Rio de Janeiro.
Combining now the results from Huang et al. (2012) and Fig. 11, it is possible to estimate the change in abrasion response as a function of degree of reduction. This is illustrated in Fig. 12, which shows predictions on the effect of effective collision energy on the mass loss in a single collision of a pellet inside the furnace with different levels of reduction.
Predicted percentage of fines generated in pellets with different levels of reduction as a function of effective collision energy.
A comparison of Fig. 7 and Fig. 10 demonstrates that the collision energies found in the furnace are still comparatively low, so that significant extrapolation in the model became necessary.
Degradation in the furnace, assuming stresses are insufficient to cause body breakage, only surface breakage (attrition), can then be predicted considering that pellets will undergo collisions from the feed to the discharge of the furnace flowing like a piston, so that
| (2) |
| (3) |
where wi is the fraction of pellets in size class i, n is the number of size classes, E is the effective specific energy applied to each pellet in each collision and h is the fractional depth in the furnace.
The proportion of material contained in the finer size range (fines) generated is then calculated by
| (4) |
The challenge in applying the equations above lies in the fact that kj, the proportion of degraded material by surface breakage, not only varies with collision energy, but also with the degree of reduction and, therefore, the vertical normalized position in the furnace h. Fortunately, work by Parisi and Laborde (2004) made it possible to estimate the degree of reduction of the pellets along the vertical axis of the furnace, from the feed to the end of the reduction section, that is, the entrance of the point of entry of the reforming gases. Parisi and Laborde (2004) proposed this model on the basis of data from two industrial MIDREX plants. Fig. 13 shows outputs of their simulations, which were used as the basis for the present work. The figure also presents predictions of the surface breakage parameter a from Eqn. 1, now considering the model described in Fig. 11, which nearly doubled in magnitude from the feed to the end of the reduction section.
Variation of the degree of reduction (based on Parisi and Laborde model) and the surface breakage parameter a in Eqn. 1 as a function of depth in the reduction section of the furnace.
The assumptions considered in the present work were:
Simulations were carried out considering the base case (Table 4), which resulted in Fig. 14. It shows that the fines generation in the reduction zone is 3.4 %. Considering this, as well as the additional contribution of fines generation in the other sections (#2 to #3, of the burden feeders), the final predicted value for the amount of fines generated inside the reactor is estimated as 3.5 % (Table 5). This result shows good agreement with those seen in industrial plants that produce cold DRI and perform screening of discharge (2 to 4 %) (MIDREX, 2011). With this value considered for fines generation, the resulting production loss would be 17600 t/year approximately.
Prediction of the evolution of fines generation relative to the depth of the furnace in reduction zone with the effect of loss of mechanical strength caused by the reduction.
Summary of the predictions of fines generation for the different sections of the furnace for the feedrate of 58 t/h.
| Section | Estimated fines generation (%) |
|---|---|
| #01—Reduction zone | 3.42 |
| #02—Upper Burdenfeeder | 0.07 |
| #03—Middle Burdenfeeder | 0.01 |
| #04—Lower Burdenfeeder | 0.03 |
| Total | 3.53 |
It is important to remember that the reference value mentioned only corresponds to discharge and other fines generated inside the reactor, e.g. in take-off, for correct comparison with the value found by the model, must also be computed. As such, the predictions by the model may still be conservative. If the change in the amenability of the pellets to surface breakage resulting from reduction is not accounted for, then predictions would be of only 2.64 % of fines, which very likely underestimates the proportion of fines generated in the furnace.
Simulations have also considered the case when the throughput of the furnace is increased (Table 4). The increase in throughput by 25 and 50 % would increase the fines production to 4.26 and 4.63 % in the reduction zone, demonstrating the increased importance of this problem.
Based on simulations of motion of iron ore pellets in a MIDREX MINIMOD furnace using the discrete element method and to the application of a simple degradation model, it was possible to conclude that:
The authors would like to thank Vale S.A. for sponsoring the research and for permission to publish the work and for CNPq for partially sponsoring the work. The authors also thank DEM Solutions for the use of the software EDEM under the Academic Program.
Fernando Oliveira Boechat
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/35_2018009_15.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
Fernando Oliveira Boechat is PhD student at Universidade Federal do Rio de Janeiro. He completed his bachelor’s degree in Industrial Engineering from Faculdade Integradas Espirito Santense and his master’s degree in Metallurgical and Materials Engineering from Universidade Federal do Rio de Janeiro. He is currently a lecturer at Centro Universitário Católico de Vitória, teaching courses in Industrial Engineering. His research interests include modeling of industrial, mainly metallurgical processes.
Rodrigo Magalhães de Carvalho
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/35_2018009_16.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
Rodrigo Magalhães de Carvalho is an Associate Professor at Universidade Federal do Rio de Janeiro. He holds a bachelor’s degree in Chemical Engineering and Metallurgical and Materials Engineering master’s and D.Sc. degrees at Universidade Federal do Rio de Janeiro, under Prof. Luís Marcelo M. Tavares supervision. He joined the faculty of the Department of Metallurgical and Materials Engineering at Universidade Federal do Rio de Janeiro in 2014, where he is the deputy-head of the Laboratório de Tecnologia Mineral. His research interests include discrete element method in mineral processing and steelmaking applications, mathematical modeling and simulation of comminution processes. He received Minerals Engineering International Young Person’s award and International Mineral Processing Congress Young Authors Award in 2014 and has received a number of awards from Brazilian Association of Metallurgy, Materials and Mining (ABM) as a co-author with Prof. Luís Marcelo Tavares.
Luís Marcelo Tavares
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/35_2018009_17.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
Luís Marcelo Tavares is a Professor at Universidade Federal do Rio de Janeiro. He received his bachelor’s degree in Mining Engineering (honours) and his master’s degree from Universidade Federal do Rio Grande do Sul. He was awarded a Ph.D. degree in Extractive Metallurgy at the University of Utah under the supervision of the late Professor R.P. King. He has been a member of the faculty of the Universidade Federal do Rio de Janeiro (Brazil) since 1998, where he is head of the Laboratório de Tecnologia Mineral and has been Department chairman. His research interests include advanced models of comminution, particle breakage, physical concentration, classification, iron ore processing and development of pozzolanic materials. He is a founding member of the Global Comminution Collaborative (GCC) and has received a number of awards from the Brazilian Association of Metallurgy, Materials and Mining (ABM).
