Professional Documents
Culture Documents
PII: S0032-5910(18)30151-7
DOI: doi:10.1016/j.powtec.2018.02.030
Reference: PTEC 13206
To appear in: Powder Technology
Received date: 30 September 2017
Revised date: 8 February 2018
Accepted date: 12 February 2018
Please cite this article as: A. Hossein Madadi Najafabadi, Abolfazl Masoumi, S. Mehdi
Vaez-Allaei , Analysis of abrasive damage of iron ore pellets. The address for the
corresponding author was captured as affiliation for all authors. Please check if
appropriate. Ptec(2017), doi:10.1016/j.powtec.2018.02.030
This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
ACCEPTED MANUSCRIPT
T
a
School of Mechanical Engineering, College of Engineering, University of Tehran,
IP
P.O. Box 11155-4563, Tehran, Iran.
CR
b
School of Physics, College of Science, University of Tehran,
US
P.O. Box 11155-4563, Tehran, Iran.
AN
Abstract
The abrasive wear of iron ore pellets (IOP) during transportation and storage in the iron making industry
M
influences process quality and leads to air pollution. Therefore, prediction of the abrasive wear of IOP
ED
enables one to eliminate possible damages during processing. In the present research, the abrasive wear of
IOP is analysed by numerical simulation using the discrete element method (DEM) and an analytical
PT
model. The variable parameters of the analytical model, including average normal contact force and
CE
sliding distance, were predicted from DEM simulation results. To evaluate the proposed numerical-
analytical technique, a standard abrasion resistance test of IOP (ASTM E279) was carried out and the
AC
experimental results were compared to numerical-analytical results. The comparison was made for two
types of IOP with different mechanical properties. The error of the numerical-analytical method in
prediction of abrasive wear of IOP was +21.7% and -34.4% for optimised quality and low quality pellets,
respectively.
Corresponding author. Tel/Fax: +98 9123841995 /+98 2188013029
Keywords: Iron ore pellet; Discrete element method; Abrasive wear; Analytical model
1. Introduction
Numerous studies have been performed addressing numerical simulation of granular materials in which
the surface wear of granules is predicted based on analytical models or experimental techniques. Previous
studies can be classified into two major categories. In the first category, abrasive wear and breakage of
T
IP
granules are carried out deliberately, as in comminution processes of mineral materials. The work of
CR
Cleary and Morrison [1] and Powel and McBride [2] are examples of such studies. In the second
category, the size reduction of granules is considered as undesired damage, and the aim is to reduce their
US
wear and breakage extents. In such researches, conveying, storage, and packing of granular materials are
studied in order to estimate the amount of damage that may occur during these processes. Studies by
AN
Ahmadian et al. [3] and Ning and Ghadiri [4] are good examples of works in this category.
M
Analytical, numerical, and experimental research approaches have been widely used to predict the
breakage and abrasive wear of granular materials. In these works, variable parameters (such as energy,
ED
force, and velocity) of the experimental or analytical models can be obtained from the numerical
PT
simulation. Ghadiri and Zhang [5, 6] propose one of the most important analytical models based on
fracture mechanics principles to evaluate the attrition of granular materials. They present an analytical
CE
model based on indentation facture mechanics to determine the impact attrition of granular materials.
Ning and Ghadiri [4] extended this model to predict the surface wear of catalyst beads in a shear cell test.
AC
From a computational point of view, one of the well-known methods for the simulation of granular
materials is the discrete element method (DEM), first proposed by Cundall and Strack [7]. Nowadays,
there are many commercial and open-source computer codes and software available for DEM simulation
of granular materials.
ACCEPTED MANUSCRIPT
One of the most important industrial granular materials is the iron ore pellet (IOP). IOP is the initial form
of iron production in direct reduction plants. Generally, production of this granular material contains two
major stages: agglomeration and firing. The chemical composition and production process factors affect
reduction behaviour and mechanical strength of IOP. Producing iron by direct reduction method is more
environmentally friendly than other common methods. However, the damage vulnerability of IOP to
T
collisions results in excessive wear and breakage damages during handling, storage, and reduction
IP
processes. These damages lead to size reduction of IOPs and generation of fines. Generated fines may
CR
lead to the following problems:
US
Clustering of IOPs during reduction process.
AN
Environmental pollution.
Gustafsson et al. [8, 9] used smoothed particle hydrodynamic (SPH) and multi-particle finite element
M
method (MPFEM) to simulate IOP flow and storage into a silo. They focussed on predicting the breakage
ED
damage of IOP and did not pay attention to the abrasive wear of IOP. Morrissey [10] and Thakur et al.
[11] simulated IOPs through DEM by including moisture and fines factors and considered them on
PT
cohesion. Nabeel et al. [12] experimentally investigated the influence of IOP characteristics (weight,
CE
Prediction of the abrasive wear of IOPs during handling and storage based on a numerical-analytical
AC
approach has not been previously reported. In the present research, a numerical-analytical method for the
prediction of abrasive wear of IOPs has been proposed. The analytical model of abrasive wear of Ning
and Ghadiri [4] has been used, for which variable parameters are generated via DEM simulation.
Identification and optimisation of the effective physical and mechanical properties of IOP during
Determination of the damaging points within the equipment, such as chutes, diverters, and day
bins.
In order to validate experimental and analytical parts of this research, two types (normal and weak) of
T
IP
This section presents a brief description of determination methods for the physical and mechanical
CR
parameters of IOPs, for both analytical model and DEM simulation.
US
Here, 20 kg pellets were taken from production line of pelletizing plant of Mobarakeh Steel Company
(MSC) as the normal type IOP, and 20 kg of pellets were produced under inefficient conditions as the
AN
weak type IOP in laboratory of the pelletizing plant. In fact, the firing temperature was set to 1180-1220
ºC instead of 1300 ºC, to produce IOPs with a lower mechanical strength. The parameters and
M
characteristics of both types are presented in the following context. It is worth mentioning that a
ED
combination of three bed-levels have been chosen for testing each type of IOP.
PT
Both types of IOPs tested in present research contain 95% Hematite (Fe2O3) and their chemical
CE
The porosity of IOPs is measured in the laboratory of the pelletizing plant according to ASTM C20. The
porosity of tested IOPs have been 20 to 22% for both types of IOP.
The extent of damage (abrasive and breakage) of IOPs during handling, storage, and the reduction process
is strongly affected by the firing process parameters. The crushing strength (according to ASTM E382)
ACCEPTED MANUSCRIPT
and abrasion resistance (according to ASTM E279) are two standard tests to qualify the mechanical
quality of IOPs. The results of the abrasion resistance test and crushing strength test for both types of
IOP, are presented in Table 2. For the abrasion resistance test, according to ASTM E279, a one diameter
tumbler containing 11.3 kg IOPs rotates for 8 minutes at rotational speed of 25 rev/min.
T
IP
Size distribution of IOPs is identified according to ASTM E11. Table 3 presents a typical size distribution
CR
chart of IOPs produced in MSC. The acceptable size range for IOPs is 9 to 16 mm; pellets out of this
US
It can be observed from Table 3 that the major percentage of IOPs is within the range of 9 to 16 mm. The
unacceptable pellets account for only 10% of a total batch. Since the purpose of this study is to predict the
AN
abrasive wear of IOPs during handling and transport to the direct reduction plant, only the IOPs of fifth
M
and sixth ranges (9-12.5 and 12.5-16) are selected for simulation. The size range of 9 to 12.5 mm is
labelled as group 1 with a mean diameter of 10.75 mm and the range of 12.5 to 16 mm is labelled as
ED
Young’s modulus is the most important parameter in present study. However, conventional methods are
not applicable for measurement of this parameter for granular materials such as IOP. Here, Young’s
AC
Gustafsson [8] calculated Young’s modulus of IOP, based on Eq. 1, presented by Bruno et al. [13]. This
equation calculates Young’s modulus in terms of material porosity (ε), morphology factor (m), and
𝐸 = 𝐸𝑑 (1 − ε)𝑚 (1)
ACCEPTED MANUSCRIPT
Gustafsson [14] suggested a morphology factor m=2 for overlapping spherical pores and m=4 for
overlapping spherical solids in order to achieve IOP’s Young’s modulus according to Eq. 1.
As mentioned above, tested IOPs consist of 95% hematite and their porosity is 20 to 22%. Considering
Ed=160 GPa for hematite [19] and p=0.22, Young’s modulus for these IOPs is within the range of 59 GPa
< E < 97 GPa. Therefore, Young’s modulus of both types of IOP is set as 78 GPa for DEM simulation
T
and analytical modelling.
IP
Bruno et al. [13] consider Poisson’s ratio to be independent of porosity and set it to υ=0.2, according to
CR
Poisson’s ratio of hematite.
US
The determined values of Young’s modulus and Poisson’s ratio are applied for both types of IOP. This
seems to be one of the main factors of error in DEM simulation and the analytical model, especially for
AN
weak IOPs.
M
Due to the non-homogenous and porous structure of IOP, usual measurement techniques for fracture
PT
toughness are not applicable. For measuring the fracture toughness, a special method, proposed by
Warren [16] was used for IOP. In this method, a sphere with radius R penetrates the material and the
CE
exerted force increases gradually up to the material fracture; fracture toughness is calculated using:
AC
𝐸 ∗ 𝑃𝐹𝑚𝑖𝑛 1/2
𝐾𝐶 = ( 𝑚𝑖𝑛 𝑅 ) (2)
𝑃𝐹𝑁
In which,
1 1−𝜈2 1−𝜈𝐼2
= + (3)
𝐸∗ 𝐸 𝐸𝐼
ACCEPTED MANUSCRIPT
E* is obtained via Eq. 3 where 𝜈 , E, 𝜈𝐼 , and EI, are Poisson’s ratio and Young’s modulus of the material
and spherical indenter, respectively. PFmin is the minimum fracture force and can be measured through the
𝑚𝑖𝑛
steps described by Warren [16]. 𝑃𝐹𝑁 is the normalised minimum fracture force, which has been provided
𝑚𝑖𝑛
for different Poisson’s ratios by Warren [16]. The expected value of 𝑃𝐹𝑁 for IOP is 1883 N, according to
T
The advantage of Warren’s approach is its simplicity. However, this has led to uncertainty in the results.
IP
His method is especially imprecise for porous and non-homogenous materials. Therefore, Warren’s
CR
method is expected to give a rough estimation of IOP fracture toughness.
US
Before the test, the assumptions used in Warren’s method should be checked for applicability to IOP. In
Warren’s tests, an acoustic emission transducer was used to detect the moment at which the crack
AN
instigates from a small surface flaw. Meanwhile, measuring the minimum fracture force via trial and error
M
is suggested [16]. The penetration surface of the specimen should be polished to reduce the surface flaws
depth to the range of 5 to 10 µm. Another issue is the difference between the elastic properties of the
ED
testing material and spherical indenter material. Thus, the material of the penetrating sphere should be
selected correctly in order to minimise the elastic mismatch [16]. For this reason, the sphere's material
PT
was chosen as soda lime glass, which has a Young’s modulus and Poisson’s ratio very close to that of
CE
IOP. Young’s modulus and Poisson’s ratio of soda lime glass are reported to be 72 GPa and 0.2,
respectively [17].
AC
The soda lime sphere with a diameter of 9 mm was placed in a Brinell indenter holder, mounted on an
IOP crushing strength test machine using a specific component (Fig. 1). The glass sphere was replaced for
A total of 25 pellets were polished and tested under Hertzian loading with the soda lime glass sphere;
similar tests were done for weak type IOPs. Minimum fracture force and standard deviation of results, as
well as calculated Kc, are presented in Table 4 for both types of IOP.
For validation, the procedure was repeated for two groups of normal IOPs from different production
batches. Although comparison of results shows an acceptable agreement, application of Warren’s method
T
for IOPs involves uncertainty during complexity and porous structure of the material. This uncertainty is
IP
expected to be one of the main factors of error in the analytical model used in the present study.
CR
2.6. Hardness
US
Huang et al. have found micro hardness of oxide pellet (IOP) to be constant in the whole regions
(periphery, core, and mantle) [18]. They specified the periphery region to be a 2 mm thickness from the
AN
pellet surface. However, Nabeel et al. have concluded that hardness of IOP differs from the outer layer to
M
Here, a digital micro hardness tester (HVS-1000) was utilized to measure the micro hardness of 25 IOPs
in the peripheral region, where it is expected to be exposed to abrasive wear. Table 5 presents results as
PT
the mean value of 25 tests and standard deviation of results for both types of IOP.
CE
Since, from the chemical and physical points of view, IOPs produced by different producers are very
similar to each other, in the present study, the coefficients of rolling friction and restitution are selected
based on results by Barrios et al. [19]. However, for coefficient of sliding friction of IOP an innovative
method is developed and utilized in this study, as described below. Contact parameters are assumed to be
To estimate the coefficient of sliding friction, in particle-particle contact or particle-wall contact, Barrios
et al. [19] proposed the pin-on-disk tribometer method. This method cannot yield an acceptable accuracy
for particle-particle contact for two reasons. First, the thin layer of iron ore resin, which covers the
rotating plate, does not represent the actual IOP properties because the production and hardening process
of IOP is quite complicated. Second, in the pin-on-disk tribometer test, the sliding contact is long and
T
continuous, while in real conditions, pellet contact occurs over a very short distance. In the present study,
IP
a special experimental setup was designed to estimate the coefficient of sliding friction (Fig. 2).
CR
A cylindrical part was fixed to the spindle and a linear tool was installed on the cross slide of a high
precision lathe machine. Each component holds a pellet and due to spindle rotation and adjusting cross
US
feed, the tangential contact of pellets is properly simulated. A Kistler 3-axis dynamometer was assembled
on the cross slide. The tangential component of collision velocity was adjusted via spindle rotational
AN
speed and the normal contact force was adjusted by the cross feed. Thus, the collision velocity and normal
M
contact force were in the range obtained from the initial simulation results.
ED
The test was performed for different tangential velocities and normal forces, and the average coefficient
of sliding friction was obtained according to: Ft = µ.Fn, where Ft and Fn are tangential and normal
PT
components of contacting force and µ is coefficient of sliding friction. In addition, the test was performed
for contact between the pellet and steel. The results are presented in Table 6.
CE
Considering the fact that each IOP has a unique shape, simulation of IOP with a real shape is practically
impossible. Barrios et al. [19] discussed the adjustment of the coefficient of rolling friction in order to
reduce the effects of shape. They predicted the coefficient of rolling friction for both spherical and
overlapping models. They used EDEM software and simulated the pellets as overlapping spheres. In the
present work, due to limitations of post-processing the LIGGGHTS results in overlapping mode,
ACCEPTED MANUSCRIPT
simulations were performed in spherical mode. Therefore, the coefficient of rolling friction is set for the
sphere model based on the results presented by Barrios et al. [19], as given in Table 6.
Coefficient of restitution
Barrios et al. [19] also measured the coefficient of restitution through high-speed recording of drop tests
T
IP
3. Analytical
CR
3.1. Theoretical model of Ghadiri and Zhang
US
Ghadiri and Zhang [5] studied the chipping mechanism of semi-brittle materials by focusing on
microscopic and macroscopic damages in contacts. They classified wear damage of particles into two
AN
types: I) damage during sliding and rolling contact of particles on each other (abrasion), and II) wear
M
damage caused by impacts (erosion). They identified the aforementioned damages via subsurface lateral
cracks based on microscopic observations. The model of Ghadiri and Zhang, which is based on the
ED
𝑐 𝐻√𝑎
= (4)
𝑎 17.1𝐾𝑐
CE
Where H, Kc, c, and a are hardness, fracture toughness of material, crack length, and radius of impression,
AC
respectively. Using the model of Ghadiri and Zhang [5] for subsurface lateral crack extension, Ning and
Ghadiri [4] estimated the volume of abrasive wear of sliding spherical particles by:
5/4 3/4
0.07𝐹 𝜋1/4 𝐻 1/4 𝐹𝑛
𝑉𝑚 = 𝐾 𝜋5/4𝑛𝐻1/4 [2𝐿 + 17.1𝐾𝑐
] (5)
𝑐
Where H and Kc are hardness and fracture toughness of the colliding granules, respectively. F n and L are
normal contact force and sliding distance of the collision which are calculated using DEM simulation.
ACCEPTED MANUSCRIPT
Therefore, the volume of abraded fines during a process would be estimated using the conjunction of
A general form of Ning and Ghadiri model (Eq. 5) could be written as follows:
T
5/4 3/4
𝑚𝑤 = 𝐴 𝐹𝑛 (2𝐿 + 𝐵𝐹𝑛 ) (6)
IP
CR
Fn is the normal contact force in Newtons, L is sliding distance in metres, and mw is the wear mass of
collided IOP in milligrams. A and B are material constants, which are calculated as follows:
US
Where ρ indicates density of the IOP,
AN
7×104 𝜌
𝐴=𝐾 5/4 𝐻 1/4 (7)
𝑐𝜋
M
And,
ED
𝜋1/4 𝐻 1/4
𝐵= 17.1𝐾𝑐
(8)
PT
According to Eqs. 6, 7, and 8, the amount of wear in each collision is independent of pellet’s diameter and
CE
its radius in the contact point. Therefore, the calculated wear mass is independent of the shape and size of
the pellets.
AC
The values of A and B for normal and weak type IOPs, are shown in Table 7.
Clearly, wear mass of a collision strongly depends on the normal contact force and wear mass in
4. Numerical
4.1. DEM
ACCEPTED MANUSCRIPT
In DEM the procedure starts with contact detection and the calculation of contact forces and
subsequently, the acceleration, speed, and new position of each particle. Hertz-Mindlin’s contact law is
used in this study. The impact force is calculated based on the normal and tangential overlap of two
spheres [20]. These calculations are performed for the entire simulation environment at each time step.
T
IP
𝐹 = (𝐾𝑛 𝛿𝑛𝑖𝑗 − 𝛾𝑛 𝑣𝑛𝑖𝑗 ) + (𝐾𝑡 𝛿𝑡𝑖𝑗 − 𝛾𝑡 𝑣𝑡𝑖𝑗 ) (9)
CR
where, 𝐾𝑛 is elastic constant for normal contact, 𝛿𝑛𝑖𝑗 is normal overlap, 𝛾𝑛 is viscoelastic damping
constant for normal contact, 𝑣𝑛𝑖𝑗 is normal relative velocity, 𝐾𝑡 is elastic constant for tangential contact,
US
𝛿𝑡𝑖𝑗 is tangential overlap, 𝛾𝑡 is viscoelastic damping constant for tangential contact, and 𝑣𝑡𝑖𝑗 is tangential
AN
relative velocity. These coefficients are calculated using Eq. 10 to 13.
4
M
𝐾𝑛 = 3 𝐸 ∗ √𝑅 ∗ 𝛿𝑛 (10)
ED
5
𝛾𝑛 = −2√ 𝛽√𝑆𝑛 𝑚∗ ≥ 0 (11)
6
PT
𝐾𝑡 = 8𝐺 ∗ √𝑅∗ 𝛿𝑛 (12)
CE
5
𝛾𝑡 = −2√6 𝛽√𝑆𝑡 𝑚∗ ≥ 0. (13)
AC
The coefficients Sn, St, β, E*, G*, R*, and m* are calculated using Eq. 14 to 20.
𝑆𝑛 = 2𝐸 ∗ √𝑅 ∗ 𝛿𝑛 (14)
𝑆𝑡 = 8𝐺 ∗ √𝑅∗ 𝛿𝑛 (15)
ln(𝑒)
𝛽= (16)
√𝑙𝑛2 (𝑒)+𝜋2
ACCEPTED MANUSCRIPT
1 (1−𝜐12 ) (1−𝜐22 )
= + (17)
𝐸∗ 𝐸1 𝐸2
1 1 1
=𝑅 +𝑅 (19)
𝑅∗ 1 2
T
1 1 1
IP
=𝑚 +𝑚 (20)
𝑚∗ 1 2
CR
Where E, G, υ, e, m, and R are Young’s modulus, shear modulus, Poisson's ratio, coefficient of
restitution, mass, and radius of a particle, respectively. The subscripts 1 and 2 stand for the two particles
US
in contact.
AN
Open-source LIGGGHTS code has been used for DEM simulation in this work. “LIGGGHTS stands for
LAMMPS Improved for General Granular and Granular Heat Transfer Simulations” and is based on
M
LAMMPS (Large Atomic and Molecular Massively Parallel Simulator), an open source molecular
ED
dynamics code created by "Sandia National Laboratories" for “massively parallel computing on
distributed memory machines” [21]. Both LIGGGHTS and LAMMPS are distributed as open source
PT
codes under the terms of the GNU General Public License [22].
CE
A major step in DEM simulation of granular material is to set the simulation parameters properly. This
step is complicated, especially when the material is not uniform and homogenous, both in terms of shape
AC
and mechanical properties. The physical and mechanical properties of the granular material define the key
parameters of simulation, including: number and diameter of particles, coefficient of restitution, Young’s
modulus, Poisson's ratio, and coefficients of static and rolling friction. Required parameters have been
The initial simulation was performed according to standard abrasion resistance test for IOP (ASTM
E279). An STL file of the tumbler was created and its rotational speed was set in LIGGGHTS code
according to the standard. The number and size of particles were set to the average diameter and volume
weight of the two major IOP groups, according to Table 3. The density of particles was set to IOP density
of 3948 kg/m3 [19] and other simulation parameters were assigned as described above.
T
To insure accurate results, a minimum time step was calculated based on the Rayleigh method (Eq. 21)
IP
and simulation outputs were stored for each time step which led to a massive amount of data production.
CR
𝜋𝑟√𝜌/𝐺
𝑇𝑆𝑚𝑖𝑛 = 0.1631𝜐+.8766 (21)
US
Where, r, ρ, G, and υ are radius, density, shear modulus, and Poisson's ratio of particle, respectively.
AN
The visualization of simulation was performed in open-source Paraview software. The output data of
M
LIGGGHTS were prepared for Paraview, using Pizza compiler. Fig. 3 displays Nine screenshots of
4.3. Post-processing
PT
As kinetic energy can be a good general indicator of a system’s condition, variations of this parameter for
CE
Obviously, the system reaches a steady state in a short time and kinetic energy fluctuates with a particular
pattern. Therefore, data related to a period of 300,000 time steps (1.2 s) seems to be sufficient for
analysis. The simulation process has been repeated 3 times and post-processing has been done for 3
periods of every simulation. Comparison of results shows a very good repeatability, thus 300,000 TS (0.5
The data of pellet-pellet contacts and pellet-wall contacts were obtained via post-processing of
LIGGGHTS outputs using a special code in C#. These data include: tangential and normal components of
force, and sliding distance of each contact in every time step. As shown in Fig. 5 each collision takes
place during several time steps and normal force changes along the contact time, thus, post-processing
code should detect and collect data of collisions through the whole period.
T
Normal force of a collision is calculated by averaging normal force data of the collision time steps and
IP
sliding distance of a collision is calculated by summing sliding distance data of the collision time steps.
CR
5. Results and discussion
US
5.1. Numerical-analytical results AN
Firstly, outputs of 300,000 TS of simulation was stored and post-processed for collisions with normal
force more than 7 N. Applying the analytical model to the simulation results gave a prediction of abrasive
M
wear for each collision. Total mass loss (produced fine particles) is calculated as the sum of wear mass of
ED
collisions.
PT
Table 8 presents a summary of results obtained from 300,000 TS (1.2 second) of simulation for pellet-
pellet and pellet-wall collisions for the normal type IOP. The wear mass of pellet-pellet collisions is
CE
multiplied by two since both pellets are worn during such collisions.
AC
The total wear mass for 8 minutes of the tumbler test (200 cycles) is estimated to be 301.35 g for the
normal type IOP, which is calculated by extrapolation of results presented in Table 8. Similar calculations
have been done for weak type IOPs; the proposed numerical-analytical approach estimated the wear mass
Primary analysis of the results shows that approximately 10% of collisions have taken place with a
normal force of more than 50 N which leads to nearly 90% of total wear. Therefore, the major fraction of
ACCEPTED MANUSCRIPT
simulation results have not contributed to wear. Table 9 presents a summary of results for collisions with
According to the results presented in Table 9, the total fine particles produced by collisions with a normal
force of more than 50 N in 8 minutes of the tumbler test is estimated to be 269.1 g based on the
numerical-analytical approach. This value is 10.7% less than fine particles produced by collisions with a
T
normal force of more than 7 N. Similar comparison for collisions with a normal force of less than 7 N,
IP
have shown that, despite huge number of collisions, the amount of wear mass is negligible compared to
CR
the total wear mass.
US
It can be concluded that post-processing can be done for collisions with a normal force of more than 50 N,
which leads to a significant post-processing simplification. Such a simplification would be greatly helpful
AN
in industrial scale applications of the proposed numerical-analytical method.
M
In order to validate the numerical-analytical approach, a comparison was done between numerical-
analytical estimation and experimental results for the abrasion resistance test (ASTM E279).
PT
The experimental results (according to Table 2), numerical-analytical estimation, as well as error of
CE
5.3. Discussion
Experimental validation has verified the precision of the proposed numerical-analytical method. Thus, at
this step the causes of discrepancy between the numerical-analytical estimation and experimental results
1. The intrinsic error of the analytical model (Ning and Ghadiri have mentioned to the point that their
3. Ignoring some data (related to the collisions with Fn < 7 N) in post-processing section.
4. Some fines have been produced as the result of fragmentation of some IOPs during the tumbler test.
5. Some fines have been lost during the gathering and weighing in the tumbler test (negligible).
T
IP
According to Table 2 only 0.6% of fragmented particles are generated during the tumbler test, therefore
the extent of fragmentation damage of normal type IOPs have not been considerable. Therefore, fine
CR
particles produced during fragmentation of IOPs (item 4) is negligible.
US
In fact, the first three items have considerable effects on the error of the numerical-analytical estimation.
Unlike the first two items, the magnitude of error caused by the third item is approximately specified
AN
(nearly -10 percent).
M
Focusing on the results for normal type IOPs, the first item seems to be the main factor, which leads to a
ED
+21.7% disagreement. The percentage of error has changed to -34.4% for weak type IOPs, which seems
to be due to uncertainty in IOP parameters as the main reason, and the fine particles produced due to
PT
fragmentation of weak type IOPs during the tumbler test, as the secondary reason.
CE
Among IOP parameters, Young’s modulus and fracture toughness have been the most important factors
influencing the precision of the numerical-analytical method. Especially Young’s modulus, which has
AC
been effective on both the analytical model and with DEM simulation, is specified to be the most
6. Conclusions
Numerical simulation and experimental work were carried out to analyse the wear of iron ore pellets.
Required parameters for DEM simulation and the analytical model were obtained in section 2. In the
ACCEPTED MANUSCRIPT
numerical section, the open source LIGGGHTS code was used to calculate the normal force and sliding
distance of collisions, which are variables in the analytical model of IOP wear. The theoretical model of
Ghadiri and Ning [4] for abrasive wear of granular materials was incorporated to predict IOP wear. The
In the experimental work (based on ASTM E279) using 11.3 kg of IOPs, corresponding to 1660
T
pellets, tumbled in a 1 m diameter drum, for 8 minutes produced 247.5 g of fine debris. The
IP
numerical simulation of the same process, but for a much shorter time of 300,000 time steps,
CR
corresponding to 1.2 s of operation, which predicted 753.4 mg of debris. Once this value is
extrapolated for the whole experimental operation time of 8 minutes, the total debris is estimated
US
as 301.3 g.
AN
A similar piece of work was carried out on weaker IOPs, for which the amount of debris for 8
minutes of operation was 1548.1 g (as measured) and 1016 g (as predicted by simulation).
M
Therefore, the difference between the model predictions and experimental measurements of
abrasive wear of two types of IOP produced in the laboratory were +14.7% and -22.1%.
ED
Intrinsic errors in the theoretical model of Ghadiri and Zhang, as well as numerical data loss due
PT
normal type IOPs. Therefore, consideration of a correction factor can improve estimation
CE
accuracy of the numerical-analytical approach for this type of IOP. As for the inefficient IOPs,
determination of Young's modulus and fracture toughness of IOPs are highly critical in the
AC
As of yet, the fracture toughness and hardness have not been recognised as significant properties
of IOP responsible for wear and abrasion in the literature. However, according to this study, these
properties are highly influential and should be considered in any analysis of wear. Further
Acknowledgments
The experimental and simulation parts of this research were done with equipment through the support of
MSC. Authors are grateful to the “Chief Operating Officer” and departments of “R&D”, “Hardware &
Data Center”, “Laboratory of Pelletizing Plant”, and the “Product Laboratory” of the Mobarakeh Steel
Company for technical assistance. The authors are also grateful to Azad University of Najafabad, for
T
providing a dynamometer for the testing of sliding friction for this research.
IP
CR
References
[1] P. W. Cleary, R. D. Morrison, Understanding fine ore breakage in a laboratory scale ball mill using
US
DEM, Minerals Engineering 24.3 (2011) 352-366.
[2] M. S. Powell, A. T. McBride, What is required from DEM simulations to model breakage in mills?,
AN
Minerals engineering 19.10 (2006) 1013-1021.
[3] H. Ahmadian, A. Hassanpour, M. Ghadiri, Analysis of granule breakage in a rotary mixing drum:
M
experimental study and distinct element analysis, Powder technology 210.2 (2011) 175-180.
ED
[4] Z. Ning, M. Ghadiri, Distinct element analysis of attrition of granular solids under shear deformation,
[5] M. Ghadiri, Z. Zhang, Impact attrition of particulate solids. Part 1: a theoretical model of chipping,
[6] Z. Zhang, M. Ghadiri, Impact attrition of particulate solids. Part 2: experimental work, Chemical
AC
[7] P. A. Cundall, O. D. L. Strack, A discrete numerical model for granular assemblies, Geotechnique
[8] G. Gustafsson, H. –Å. Häggblad, P. Jonsén, P. Marklund, Determination of bulk properties and
fracture data for iron ore pellets using instrumented confined compression experiments, Powder
[9] G. Gustafsson, H. –Å. Häggblad, and P. Jonsén, Multi-particle finite element modelling of the
compression of iron ore pellets with statistically distributed geometric and material data, Powder
[10] J. P. Morrissey, Discrete element modelling of iron ore pellets to include the effects of moisture and
T
[11] S. C. Thakur, J. P. Morrissey, J. Sun, J. F. Chen, J. Y. Ooi, Micromechanical analysis of cohesive
IP
granular materials using the discrete element method with an adhesive elasto-plastic contact model,
CR
Granular Matter 16.3 (2014) 383-400.
[12] Nabeel, Muhammad, Andrey Karasev, and Pär Göran Jönsson, Evaluation of Dust Generation during
US
Mechanical Wear of Iron Ore Pellets, ISIJ International 56.6 (2016): 960-966.
[14] G. Gustafsson, H. –Å. Häggblad, and P. Jonsén, Characterization modelling and validation of a two-
ED
point loaded iron ore pellet. Powder technology 235 (2013) 126-135.
[16] P. D. Warren, Determining the fracture toughness of brittle materials by Hertzian indentation,
[17] W. Martienssen, H. Warlimont, Springer handbook of condensed matter and materials data, Springer
[18] Huang, Zhucheng, Lingyun Yi, and Tao Jiang. "Mechanisms of strength decrease in the initial
reduction of iron ore oxide pellets." Powder technology 221 (2012): 284-291.
DEM simulation of iron ore pellet handling, Powder technology 248 (2013) 84-93.
ACCEPTED MANUSCRIPT
Experimental analysis of tablet properties for discrete element modeling of an active coating process,
[21] LAMMPS, LAMMPS User Manual, Sandia National Laboratories, USA, 2014. (Retrieved from
http://lammps.sandia.gov/doc/Manual.html).
T
[22] GPL, GNU General Public License, 2014 (Retrieved from http://www.gnu.org/licences/gpl.html).
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
T
Fig. 3. 3D diagrams of analytical model of wear mass as a function of normal force and sliding distance.
IP
CR
Fig. 4. Fifteen screenshots of simulation of the tumbler test.
US
AN
Fig. 5. Periodic changes of total kinetic energy of the system during 2.5 cycles of tumbler rotation.
M
ED
Fig. 6. Normal force in time steps of a pellet-wall collision (TS=4e-6, Collision velocity=5m/s).
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 2: Results of crushing strength and abrasion resistance tests for normal type and weak type IOPs
T
IP
Table 5: Micro hardness results
CR
Table 6: Summary of contact parameters of IOP
US
Table 7: The material constants of analytical model for both types of IOP
Table 8: A summary of numerical-analytical method for normal type IOPs, for 300000 TS (FN ≥ 7N)
AN
Table 9: A summary of results for normal type IOP for 300000 TS (FN ≥ 50 N).
M
Table 1
Mass% 65.81 0.52 2.88 0.51 0.07 0.32 0.02 0.61 0.55 0.3 0.002
T
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 2
Results of crushing strength and abrasion resistance tests for normal type and weak type IOPs
T
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 3
<0.6 2.3
0.6-6.3 1.4
T
6.3-8 1.1
IP
8-9 2.6
9-12.5 30.9
CR
12.5-16 59.5
16-18 2.2
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 4
Tests results
IOP type 𝑃𝐹𝑚𝑖𝑛 (N) 𝑃𝐹𝑚𝑒𝑎𝑛 (N) SD Kc (M.Pa.m1/2)
Normal IOP 1205 1967 388 2.36
Weak IOP 314 506 156 1.20
T
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 5
Tests results
IOP type Hmean (MPa) SD
Normal IOP 5112 878
Weak IOP 2554 301
T
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 6
Contact type
Parameter
Pellet-pellet Pellet-steel
Coefficient of sliding friction 0.43 0.47
T
a
Coefficient of rolling friction 0.21 0.25a
IP
Coefficient of restitution 0.48a 0.39a
a
Barrios et al. [19]
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 7
IOP type A B
Normal 0.105 8.81x10-6
Weak 0.225 1.46x10-5
T
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 8
A summary of numerical-analytical method for normal type IOPs, for 300000 TS (FN ≥ 7N)
Normal force Sliding distance Wear mass per Number of Total wear mass
(N) (µm) collision (mg) collision (mg)
T
-3 -5
Pellet-wall 7 1363 9e 226 4.5x10 1.75 35076 307.74
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 9
A summary of results for normal type IOP for 300000 TS (FN ≥ 50 N).
T
wall (10.7% of total)
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Table 10
T
IP
CR
US
AN
M
ED
PT
CE
AC
ACCEPTED MANUSCRIPT
Highlights
Analytical model for abrasive wear of IOP based on model of Ghadiri and Zhang.
T
IP
CR
US
AN
M
ED
PT
CE
AC
Graphics Abstract
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6