Professional Documents
Culture Documents
1-s2.0-S0032591021000723-main
1-s2.0-S0032591021000723-main
Powder Technology
a r t i c l e i n f o a b s t r a c t
Article history: High-pressure grinding rolls have an important role in several iron ore pellet feed preparation circuits. Modeling
Received 26 June 2020 and simulation are potentially useful tools to improve their performance, but the applicability of mathematical
Received in revised form 30 November 2020 models from the literature in this particular application has not yet been demonstrated, although the model by
Accepted 1 December 2020
Torres and Casali has been identified as having good potential. The present work relies on several surveys on
Available online 24 January 2021
four HPGRs with variable diameters (1.4 to 2.25 m) and aspect ratios (0.88 to 1.45) operating in pelletizing plants
Keywords:
to test the model originally proposed by Torres and Casali, as well as modifications proposed to it.
Roller press Inadequacies have been found in the original model, which gave poor predictions of power, throughput and
Modeling product size distribution along the roll's length. Modification of the power equation, incorporating a machine-
Simulation dependent correction factor for the nip angle, as well as several modifications to the throughput model, including
High-pressure grinding an expression meant to account for the ejection of material from the edge of the rolls and extrusion, resulted in
Iron ore very good agreement between data from surveys and the model. Finally, incorporation of an empirical correction
equation to account for saturation and a more general expression for the roll profile led to good agreement be-
tween measured and predicted product size distributions, including their variation along the roll's length.
© 2021 Elsevier B.V. All rights reserved.
https://doi.org/10.1016/j.powtec.2020.12.075
0032-5910/© 2021 Elsevier B.V. All rights reserved.
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
245
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
In order to solve the system of differential equations, the authors used where Umax is the maximum peripheral roll velocity in operation for the
the equations for the cumulative breakage function [28] (Eq. 6) and HPGR in question, and φ, υ and τ are dimensionless parameters that
the specific selection function [29] (Eq. 7), given by: should be fitted from data. An additional modification introduced was
in the calculation of the operating gap, which has been proposed to be
γ β
xi x estimated from the distance between the top of the studs on the rollers,
Bij ¼ ϕ þ ð1−ϕÞ i ð6Þ
xj xj added to the depth of penetration of the studs in the particle bed [13],
giving:
2
ln sEi =sE1 ¼ ξ1 ln b
xi =b
x1 þ ξ2 ln bxi =b
x1 ð7Þ χ g ¼ χ gstud þ 2λχ p ð13Þ
where ϕ, γ, and β are breakage function parameters and sE1, ξ1 and ξ2 are where χgstud is the distance between the rollers measured from the top
selection function parameters that must be fitted to data, with bij = of the studs, χp is the stud penetration in the particle bed and λ is the
Bi−1, j − Bi, j. b
xi is the representative size of particles contained in size fraction of roll surface area corresponding to the autogenous layer. For
pffiffiffiffiffiffiffiffiffiffiffiffiffi
range and given by b xi ¼ xi xi−1 . the pilot-scale HPGR work the optimal parameter values were
Moreover, the authors also considered breakage rates that vary φ = 100, υ = 140, τ = 0.1 and λ = 0.65 [5].
along the axial roll position according to the parabolic assumed power In analogy to the throughput model, Campos et al. [5] also reported
profile and a constant holdup for each block. The breakage rates are poor agreement between the measured and the predicted power con-
then given by: sumption in a pilot-scale HPGR pressing iron ore pellet feed. Indeed, it
246
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
In the case of the pilot-scale HPGR pressing iron ore pellet feed, the
value of κ was estimated as 2.8, suggesting that Eq. (4) systematically
underestimated the nip angle in this operation [5].
In the case of the product size distribution, Campos et al. [5] intro-
duced a relatively minor modification to the Torres and Casali model,
by replacing the standard size-normalizable breakage function equation
(Eq. 6) by a non-normalizable breakage function when the particle size Fig. 1. Mean size distributions of the feed to the various HPGRs studied.
xi is lower than a critical size ω, given by [35]:
x η x γ β
x
i i which allowed capturing information on roll peripheral velocity, oper-
Bij ¼ ϕ þ ð1−ϕÞ i for xi <ω ð15Þ
ω xj xj ating gap, operating pressure, power consumption and throughput.
Care was taken to guarantee that the cheek plates were closely adjusted,
where η and ω are additional fitting parameters. with a gap smaller than 1.5 mm in respect to the rolls, and the autoge-
nous main frame and wear parts were in very good condition, with no
3. Materials and methods noticeable signs of wear. Process information from each survey was
gathered when the supervisory system indicated that the HPGR was op-
3.1. HPGR industrial-scale tests erated under steady-state conditions.
Table 1 summarizes the main operating conditions as well as the
High-pressure grinding rolls operating in different pelletizing plants main material characteristics for the experimental campaigns on each
at the Complexo de Tubarão from company Vale (Vitória, Brazil) were of the HPGRs, while Fig. 1 compares the mean size distributions of
the object of the present study. The pelletizing plants are fed with a each feed. The coarser feed size corresponding to the HPGR #4 is consis-
blend of iron ore concentrates from different mines, being composed tent with its pre-grinding role in the circuit. The feed size distributions
predominantly by iron oxides, mainly hematite, with minor amounts of the feed to HPGRs #1 and #2 are finer and consistent with the
of quartz [36]. A total of four experimental campaigns were conducted regrinding pre-pelletizing role of the machines downstream from ball
in four different industrial-scale machines (#1 to #4) to assess the per- mills operating in closed-circuit. Although with the same regrinding
formance of HPGRs in pellet feed pressing. Two different circuit designs role, the HPGR #3 is fed with material with slightly coarser top size,
were surveyed: roller presses #1 to #3 operated after ball milling, in since it operates downstream from ball mills that operate in open cir-
regrinding pre-pelletizing process, whereas roller press #4 corresponded cuit. A snapshot of HPGR #2 in operation is presented in Fig. 2.
to an HPGR in a pre-grinding operation, that is, prior to ball milling. The specific gravity of the feed in the various campaigns was mea-
Moreover, campaigns were conducted with a moisture content of 8 ± sured using a Helium pycnometer and is given as 4.9 t/m3. The bulk den-
0.5%, which is recognized to be close to the maximum value that is toler- sity was obtained from the ratio between the sample mass and a known
ated in HPGR operation involving iron ore pellet feed [1]. volume and is given by 3.0 t/m3. Finally, the flake density was measured
Among the HPGRs, three of them (#1 to #3) were manufactured by
Thyssenkrupp Polysius and the remaining one (#4) was manufactured
by KHD/WEIR. All machines were connected to a supervisory system,
Table 1
Summary of the main characteristics of the HPGR as well as of the feed material in the var-
ious surveys.
HPGR
#1 #2 #3 #4
247
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
using preserved flakes from one operation, and the average value was Table 2
3.5 t/m3, which was used throughout the simulations. Summary of the base case conditions selected to fit parameter κ, as well as its optimal
value (Eq. 14) for each HPGR.
During the course of the surveys, samples were collected along the
axial roll position during the HPGR discharge in order to assess the var- HPGR
iation in size reduction along the roll length. Samples were collected in #1 #2 #3 #4
seven different points using a manual sample cutter introduced under-
Operating pressure (bar) 41 59 82 48
neath the rolls, which was quickly removed. Immediately after that, the Roll velocity (m/s) 0.78 0.83 1.54 1.30
HPGR operation was interrupted and samples from the feed and the Measured power (kW) 525 897 2325 814
product conveyors were collected from belt cuts. κ (−) 2.0 2.5 2.6 2.6
All samples were carefully quartered, dried, and subjected to wet
size analysis by laser scattering in a Malvern Mastersizer 2000 (Malvern
Instruments Inc) and Blaine specific surface area (BSA) in a PCBlaine-
Star (Zünderwerke Ernst Brün GmbH).
N h i2
f obj ¼ ∑ log W HPGR HPGR
calc ðiÞ − log W exp ðiÞ ð16Þ
i¼1
Fig. 3. Comparison of experimental and predicted powers calculated with the Torres and Casali model (a) and the modified Torres and Casali model (b) for all experimental surveys,
discriminated by machine.
248
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
Fig. 5. Comparison between experimental and calculated throughputs using the original Torres and Casali model (a) and the model modified in the present work (b) for all industrial-scale
HPGRs.
4.2. Throughput
Uρa χ c
Ug ¼ ð18Þ
Experimental results are shown in Fig. 4 that illustrate the relation- ρg χ g
ship between the specific throughput and the specific operating force
for all surveys in the different HPGRs. As already observed by Van der where Ug is the material velocity and χc is the critical aperture calcu-
Meer and Leite [37] and confirmed by Campos et al. [6] for pressing lated from [15]:
iron ore concentrates in pilot-scale HPGR in pre-grinding operation, an
χ c ¼ χ g þ D 1− cos α ip ð19Þ
inverse relationship appears between the specific throughput and the
specific force. Campos et al. [6] suggested that such variation is a conse-
where χg is the operating gap from Eq. (13).
quence of the variation in operating gap. However, different from the
Incorporating the modifications associated to the increase in mate-
results presented by Campos et al. [6] for the pilot-scale HPGR, Fig. 4
rial velocity along the extrusion zone, besides the ejection of material
shows that no single simple relationship exists that is valid for the dif-
between the edge of the rolls and the cheek plates, parameters of
ferent industrial-scale HPGRs surveyed.
Eq. (12) can now be fitted from data. Fig. 6 shows the reasonable agree-
With the aim of assessing the applicability of the Torres and Casali
ment between model fit (line) and measured values for the different
[15] model to the industrial survey data, predictions from Eq. (1) are
HPGRs over the entire range of operating operating conditions sur-
compared to experiments in Fig. 5a, which shows very poor agreement
veyed. As such, the model suggests that between about 10 and 35% of
between model and experiments, as already observed in a previous
study involving the pilot-scale HPGR pressing iron ore pellet feed [5].
However, the application of the modified version of the model (omitted
for brevity) only improved marginally the predictions, demonstrating
the need for further changes in the Torres and Casali model, beyond
those already presented in section 2.2.
One aspect that was neither included in the original model nor in its
modified version is the acceleration of the material along the compres-
sion zone, recognized by some authors [12,13,32] to be important. This
effect has been incorporated in the model by using an analogy between
the HPGR and a lamination process used in metallurgy. Indeed, a mass
balance of a lamination process may be described as [38]:
where the terms on the left of the equal sign refer to the material be-
fore the lamination and those on the right refer to the material after
the process, in which L is the roll length, Δ0 is the initial thickness,
ρsp is the specific gravity, U0 is the initial sheet velocity, Δf is the final
thickness and Uf is the final sheet velocity. In order to keep the mass
balanced on both sides of the equation, Dieter and Bacon [38] sug-
gested that Uf > U0. The present work uses Eq. (18) to calculate the
material velocity as a constant value along the extrusion zone in the
HPGR. For pressing in a confined bed an additional modification must
be incorporated by using the packing density before (bulk density) Fig. 6. Relationship between the product of the dimensionless gap and the dimensionless
and after the operation (flake density). As such, rearranging Eq. (17) rolls speed raised to parameter τ and the estimate of the proportion of material ejected
gives: from the edges of the rolls and the cheek plates. Data identified by machine.
249
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
Table 3
Summary of the fitted parameters for all HPGRs.
Parameter Values
Very good agreement in both the coarser and the finer parts of
the size distribution are required in pelletizing, since both extremes
have a significant impact on iron ore pellet characteristics [39]. A
well-known way to capture these features of the size distribution
in the iron ore pelletizing industry is through measurement of the
amount of material passing the 45 μm sieve and the Blaine specific
Fig. 7. Feed and product size distributions from surveys with HPGR #1 discriminated by surface area [39]. A model, presented in the appendix, has been
specific force (and energy) at a constant throughput of 400 t/h. used to estimate the BSA from the predicted size distributions [5].
As such, Fig. 10 compares measured and predicted values of the
BSA (a) and of the percent passing the 45 μm sieve (b) in the prod-
uct for all surveys with HPGRs #1 to #3. The results demonstrate
material would have been ejected from the edge of the rolls in opera- good agreement for surveys carried out with specific compressive
tion, which is in general agreement with estimates by Lim et al. [12] forces below about 2 N/mm2. Indeed, when higher specific forces
and Van der Ende et al. [20]. The optimal values of the parameters, and specific energies were used, the model tended to overestimate
valid for all industrial-scale HPGRs, were τ equal to 0.1, υ equal to 190 the product fineness.
and φ equal to 100, in which only υ needed to be refitted from the esti- These results suggest the appearance of a well-known phenome-
mates obtained from the pilot-scale study [5]. non in confined breakage that is associated to high compressive
A comparison between the survey results and the improved model forces, in which the energy input to the bed is no longer effectively
fit is presented in Fig. 5b, which demonstrates the very good agree- used in generating new surfaces [23–25,40]. Such saturation is influ-
ment between them. As such, although Fig. 4 did not show a straight- enced by a number of variables, but can be studied under controlled
forward relationship between the operating conditions and the conditions in a piston-and-die apparatus. In order to illustrate such
measured specific throughput, the model was able to incorporate saturation, the concept of energy utilization, defined as the ratio be-
these effects. tween the increase in surface area of the material and the input en-
ergy [23], has been used. Results from testing particles, contained
4.3. Product size distribution both in narrow size ranges and as complete size distributions in a
piston-and-die apparatus and carried out by Campos et al. [23] are
Typical size reduction results from the industrial surveys are pre- presented in Fig. 11. It shows that energy utilization is maximum at
sented in Fig. 7, which shows the significant effect of specific compres- the lowest input specific energies studied, which was just high
sive force on the product size distribution, demonstrating the enough to surpass initial bed rearrangement [23]. Defining an effi-
importance of this variable in HPGR performance. ciency variable as the ratio between the actual energy utilization
In order to use the modified Torres and Casali model for describing reached at a given input specific energy and the maximum energy uti-
the breakage response from the surveys (Table 1), the present work as- lization, results are given in Fig. 12, which have been fit to the empir-
sumes validity of the non-normalized breakage function (Eq. 15) pa- ical equation:
rameters fitted by Campos et al. [5] from pressing iron ore
concentrates in the pilot-scale HPGR. Besides these, two of the parame- " #
ters in the selection function (Eq. 7) were also assumed to be constant Ei Λ
ΨðEi Þ ¼ exp − ð20Þ
for all industrial-scale HPGRs. These values are listed in Table 3. The re- E0
maining parameter sE1 in Eq. (7) has been fitted on the basis of a base
case for each HPGR, selected between those in which the specific energy
is intermediate to low. where Ψ is the proposed efficiency variable, Ei is the input energy, E' is a
Fig. 8 presents the fitted breakage function [5], which shows the material parameter, given in kWh/t, and Λ is a dimensionless fitting pa-
point of inflexion at 13 μm, as well as the specific selection functions rameter. Results from Fig. 12 show Eq. (20) can describe the data rea-
fitted for each HPGR. Results from Fig. 8b are able to highlight the rela- sonably well for pressing iron ore concentrates in a piston-and-die
tively higher specific breakage rates for tests performed in the HPGR #1 test. Moreover, a recent work by the authors showed the great similarity
(Pre-grinding), showing smaller variations among the other surveys. in the energy utilization for pressing different iron ore concentrates in a
In order to assess the model predictive capabilities to deal with piston-and-die apparatus and a lab-scale HPGR, thus justifying the ap-
different operating conditions, Fig. 9 compares experimental and simu- plication of these results to calibrate the model for industrial-scale
lated product size distributions from surveys with HPGRs #1 and #4. HPGR operations [36]. Nevertheless, Campos et al. [36] also showed that
These results demonstrate the very good ability of the model in the energy utilization varies with particle size and size distribution. This
predicting the product size distribution both when the HPGR is fed suggests that caution should be exercised in use of Eq. (20) to simulate
with a coarser feed size distribution, such as in a pre-grinding operation changes in feed size distribution.
(Fig. 9b), and a finer feed size distribution in the regrinding pre- As such, it is hereby proposed that Eq. (8) of the Torres and Casali
pelletizing process (Fig. 9a). model is modified in order to account for the breakage saturation, by
250
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
Fig. 8. Cumulative breakage function for selected parent particle size classes (a) and specific selection functions (b) fitted for all HPGRs using sE1 = 0.20 t/kWh (#1), 0.26 t/kWh (#2), 0.15 t/
kWh (#3) and 0.40 t/kWh (#4).
applying the constant Ψ to reduce the breakage rates at each block and length. Results are presented in Fig. 14, showing the significant variation
size class, giving: of both BSA (Fig. 14a) and percent passing the 45 μm sieve (Fig. 14b)
along the length of the rolls. As already reported by Campos et al. [6]
Pk P for pressing iron ore pellet feed in a pilot-scale HPGR and suggested
Si,k ¼ sEi Ψ ð21Þ
Hk Q by other authors [14,18], it shows the edge effect, with a coarser product
formed close to the edges of the rollers and a finer product being gener-
Fig. 13 then compares experimental and predicted values of BSA (a) and ated in the central zone. However, Fig. 14 does not show the parabolic
percentage passing the 45 μm sieve in the product of tests carried out profile suggested in Eq. (9) as precluded by the Torres and Casali
with HPGRs #1 to #3 after applying Eq. (21). Results demonstrate the model, since a region of approximately constant fineness appears in
very good description of both measures using the proposed model mod- the central part of the rolls. The figure also shows that in the edges
ification, which yielded good predictions even at high compressive nearly no size reduction occurs, since both BSA and the proportion pass-
forces. ing the 45 μm sieve in the product are only marginally higher than in
the feed.
4.4. Axial variation of product size distribution along the rolls´ length As such, an alternative expression to Eq. (9) has been sought to de-
scribe such observed profile. It is proposed that the axial normalized
In selected surveys, samples from underneath the rolls were col- pressure profile (Pk′) can be described as a periodic function based on
lected, so as to map the axial variation of the product along the rolls the Fourier Transform:
Fig. 9. Comparison of experimental and predicted product size distributions in regrinding pre-pelletizing at 700 t/h and specific force of 0.9 N/mm2 (HPGR #1) (a) and pre-grinding at
500 t/h and specific force of 2.5 N/mm2 (HPGR #4) (b).
251
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
Fig. 10. Comparison between measured and predicted BSA (a) and percent passing the 45 μm sieve (b) in the product from surveys with HPGRs #1 to #3 using the modified Torres and
Casali model presented in section 2.
Fig. 11. Variation of the energy utilization as a function of specific input energy for pressing
Fig. 12. Comparison between experimental and fitted values for the efficiency in creating
different iron ore concentrates from a narrow feed size range of 150–125 μm (A-C) and
new surface for different iron ore concentrates from a narrow feed size range of
complete feed size distribution (D) in a piston-and-die apparatus.
150–125 μm (A-C) and complete feed size distribution (D) in a piston-and-die
apparatus, with E´ = 9.8 kWh/t and Λ = 0.81.
4 100 1− cos nπ −μ ðn2 π2 Þ
P 0k ¼ ∑ e sin nπyk ð22Þ
π n¼1 2n
where μ is a single fitting parameter that can be selected to match the Fig. 15 compares experimental and predicted values of BSA (a) and per-
profile shape, varying in the range of 0.001 to 0.1. yk is the normalized centage passing the 45 μm sieve in the product using Eq. (24), compar-
average position along the roll length given by yk ¼ yk−12þyk , in which yk ing to results obtained using the parabolic relationship proposed by
is given by: Torres and Casali (Eq. 9). It clearly shows that the new proposed ex-
k pression is able to more closely describe the shape of the profile ob-
yk ¼ ð23Þ served in the plant survey. Minor deviations were found in Fig. 15b
Nb
with Eq. (24) overestimating the measured percent passing in 45 μm,
where Nb is the number of blocks. Therefore, the power profile along even though still showing a relatively good agreement. These results
the axial roll position, previously described in Eq. (9), would be replaced showed that the value of μ of about 0.01 is able to describe the approx-
by: imately trapezoidal profile observed in Fig. 15. It is important to
emphasize that when μ is approximately equal to 0.1, Eq. (22) ap-
κα ip P0 proaches the parabolic profile given by Eq. (9), thus demonstrating its
P k ¼ 2F m sin U N k ð24Þ
2 B wide applicability.
∑ P 0j
j¼1
252
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
Fig. 13. Comparison between measured and predicted BSA (a) and percentage passing the 45 μm sieve (b) in the product from surveys on HPGRs #1 to #3 with different specific
compressive forces using the modified Torres and Casali model with the description of the breakage saturation effect given by Eq. (20).
Fig. 14. Variation of BSA (a) and percentage passing the 45 μm sieve (b) along the axial roll position for surveys in HPGR #1 with a constant throughput of 400 t/h at different specific forces.
5. Conclusions improved predictions of product size distributions for all HGPRs over
the range of specific forces studied. Finally, an expression for the
Direct application of the model proposed by Torres and Casali power profile along the axial roll length, which allowed to describe
[15] to data from surveys of four HPGRs operating in pelletizing the profile observed in the surveys that approaches a trapezoid, pro-
plants led to poor predictions of power, throughput, product size vided a very good fit to the survey data.
distributions at high specific forces and product fineness along
the roll's length.
Modification of the equation to predict operating power by incorpo- Declaration of Competing Interest
rating a machine-dependent constant to be fitted from data allowed
matching the predicted power with the values collected in the plant The authors declare that they have no known competing financial
surveys. Modification of the throughput model by incorporating both interests or personal relationships that could have appeared to influ-
extrusion of material between the rolls and ejection of material be- ence the work reported in this paper.
tween the edge of the rolls and the cheek plates allowed the model to
produce very good fit to the survey data. Acknowledgements
The use of a non-normalizable breakage function, coupled to a func-
tion that accounted for the inefficiency in usage of input energy in par- The authors would like to thank Vale for financial and technical sup-
ticle breakage, proposed on the basis of piston-and-die tests, led to port to the investigation, as well as the Brazilian Agencies CNPq (grant
253
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
Fig. 15. Comparison between experimental and predicted BSA (a) and percentage passing the 45 μm sieve (b) in the product along the roll axial position using Eq. (24) with μ = 0.01
(Eq. 22) and the parabolic profile (Eq. 9). Survey with HPGR #1 with a specific force of 1.5 N/mm2 and a constant throughput of 700 t/h.
" #1
x2i þ x2i−1 ðxi þ xi−1 Þ 3
xi ¼ ðA:2Þ
4
where xi and xi−1 are, respectively the lower and the upper sizes of the
interval.
The authors then proposed an empirical expression to calculate
Blaine specific surface area (BSA) [41]:
where G and J are constants that must be fitted to data. On the basis of
data from pressing iron ore concentrates in a pilot-scale campaign, Cam-
pos et al. [6] estimated the optimal values of G = 0.996 and J = 18.0.
However, given the differences in material characteristics between that
study and the feed to the various HPGRs in the present work, given by a
blend of different ore type in the Complexo de Tubarão from Vale S.A.
(Vitória, Brazil), a need for re-estimating the parameters became evi-
dent. This resulted in the constants G equal to 1.398 and J equal to
27.97. The constants G and J were calibrated based in the equivalent
spherical specific surface area (SSA) given in m2/kg and, therefore, the
Fig. 16. Comparison between experimental and calculated values for the Blaine specific BSA (Eq. A.3) will be calculated in m2/kg. Transformation of this value
surface area (BSA) using Eqs. (A.1) to (A.3). to cm2/g requires multiplying the values achieved in Eq. A.3 by 10. A
comparison between experimental data and the model fit is given in
Fig. 16, which shows that the average difference between measured
number 310293/2017-0), CAPES and FAPERJ (grant number E-26/ and fitted results was 100 cm2/g.
200.592/2018) for partially funding the work.
References
Appendix A
[1] F.P. Van der Meer, Roller press grinding of pellet feed, Experiences of KHD in the
Iron Ore Industry, AusIMM Conference on Iron Ore Resources and Reserves Estima-
Given that the model equations provide as output the size distri- tion, Perth 1997, pp. 1–15.
bution in the HPGR discharge, an equation must be available to esti- [2] G. Ehrenraut, R. Rao, Experience with roller press in the pellet plant Kudremukh Iron
Ore Company Ltd, Aufber. Tech. 10 (2001) 469–476.
mate the BSA from it. Zhang and Napier-Munn [41] proposed an [3] F.P. Van der Meer, Pellet feed grinding by HPGR, Miner. Eng. 73 (2015) 21–30.
expression to predict the BSA according to the size distribution, [4] A. Abazarpoor, H. Halali, R. Hejazi, M. Saghaeian, HPGR effect on the particle size and
given by: shape of iron ore pellet feed using response surface methodology, Physicochem.
Probl. Miner. Process. 53 (2017) 1–9.
[5] T.M. Campos, G. Bueno, G.K. Barrios, L.M. Tavares, Pressing iron ore concentrate in a
6 N wi
SSA ¼ ∑ ðA:1Þ pilot-scale HPGR. Part 2: modeling and simulation, Miner. Eng. 140 (2019) 105876.
ρsp i¼1 xi [6] T.M. Campos, G. Bueno, G.K. Barrios, L.M. Tavares, Pressing iron ore concentrate in a
pilot-scale HPGR. Part 1: experimental results, Miner. Eng. 140 (2019) 105875.
[7] T.M. Campos, G. Bueno, L.M. Tavares, Scale-up of HPGR in size reduction of iron ore
where SSA is the equivalent spherical specific surface area, ρsp is the spe-
pellet feed, Proc. XXX Int. Miner. Process. Congr. SAIMM, Cape Town 2020, pp. 1–10.
cific gravity, wi the fraction retained in each size class and xi is the har- [8] S. Rashidi, R.K. Rajamani, D.W. Fuerstenau, A review of the modeling of high pres-
monic mean of each size class, given by: sure grinding rolls, KONA Powder and Part. J. 34 (2017) 125–140.
254
T.M. Campos, G. Bueno and L.M. Tavares Powder Technology 383 (2021) 244–255
[9] D.W. Fuerstenau, A. Shukla, P.C. Kapur, Energy consumption and product size distri- [25] J.T. Kalala, H. Dong, A.L. Hinde, Using piston die tests to predict the breakage behav-
butions in choke-fed, high-compression roll mills, Int. J. Miner. Process. 32 (1991) ior of HPGR, 5th Int. Conf. Autogenous and Semi-Autogenous Grinding and High-
59–79. Pressure Grinding Roll Technol, Vancouver 2011, pp. 1–14.
[10] L.G. Austin, R. Weller, I. Lim, Phenomenological Modelling of the High Pressure [26] T. Mütze, Modelling the stress behaviour in particle bed comminution, Int. J. Miner.
Grinding Rolls, Proc. XVIII Int. Miner. Process. Congr, AUSIMM, Sydney, 1993 87–96. Process. 156 (2016) 14–23.
[11] L.G. Austin, M.P. Trubeljal, H.M. Von Seebach, Capacity of high-pressure grinding [27] M.J. Daniel, S. Morrell, HPGR model verification and scale-up, Miner. Eng. 17 (2004)
rolls, Min. Metall. & Expl. 12 (1995) 65–73. 1149–1161.
[12] W. Lim, J. Campbell, L. Tondo, The effect of rolls speed and rolls surface pattern on [28] L.G. Austin, P.T. Luckie, The estimation of non-normalized breakage distribution pa-
high pressure grinding rolls performance, Miner. Eng. 10 (1997) 401–419. rameters from batch grinding tests, Powder Technol. 5 (1972) 267–271.
[13] W. Lim, K. Weller, Some benefits of using studded surfaces in high pressure grinding [29] J.A. Herbst, D.W. Fuerstenau, Scale-up procedure for continuous grinding mill design
rolls, Miner. Eng. 12 (1999) 187–203. using population balance model, Int. J. Miner. Process. 7 (1980) 1–31.
[14] S. Morrell, W. Lim, F. Shi, L. Tondo, Modelling of the HPGR crusher, in: S.K. Kawatra [30] K.J. Reid, A solution to the batch grinding equation, Chem. Eng. Sci. 20 (1965)
(Ed.), Comminution Practices, SME 1997, pp. 117–126. 953–963.
[15] M. Torres, A. Casali, A novel approach for the modelling of high-pressure grinding [31] E. Vyhmeister, L. Reyes-Bozo, R. Rodriguez-Maecker, C. Fúnez-Guerra, F. Cepeda-
rolls, Miner. Eng. 22 (2009) 1137–1146. Vaca, H. Valdés-González, Modeling and energy-based model predictive control of
[16] C.L. Schneider, V.K. Alves, L.G. Austin, Modeling the contribution of specific grinding high pressure grinding roll, Miner. Eng. 134 (2019) 7–15.
pressure for the calculation of HPGR product size distribution, Miner. Eng. 22 (2009)
[32] K. Schönert, U. Sander, Shear stresses and material slip in high pressure roller mills,
642–649.
Powder Technol. 122 (2002) 136–144.
[17] H. Dundar, H. Benzer, N. Aydogan, Application of population balance model to HPGR
[33] H. Kellerwessel, High pressure particle bed comminution, state of the art, applica-
crushing, Miner. Eng. 50-51 (2013) 114–120.
tion, recent developments, Eng. Min. J. (1996) 45–52.
[18] U. Lubjuhn, K. Schönert, Material Flow in the Acceleration Zone and Throughput of
[34] R. Klymowsky, N. Patzelt, J. Knecht, E. Burchardt, An overview of HPGR technology,
High Pressure Roller Mills, Proc. XVIII Int. Miner. Process. Congr, AUSMM, Sydney,
Proceedings of the SAG Conference 2006, pp. 11–26.
1993 161–168.
[35] R.P. King, Modeling and Simulation of Mineral Processing Circuits, 1st ed.
[19] M.J. Daniel, HPGR Model Verification and Scale-up, Masters Thesis Julius
Butterworth-Heinemann, 2001.
Kruttschnitt Mineral Research Centre, University of Queensland, 2002.
[20] R. Van der Ende, H. Kanapp, F. Van der Meer, Reducing edge effect and material by- [36] T.M. Campos, G. Bueno, V.R. Alfonso, A.-C. Böttcher, F. Mayerhofer, A. Kwade, L.M.
pass using spring-loaded cheek plates in HPGR grinding, 7th Int. Conf. Autogenous Tavares, Relationships between particle breakage characteristics and comminution
Grinding, Semi-autogenous Grinding and High Pressure Grinding Roll Technol, Van- response of fine iron ore concentrates, Miner. Eng. (2020) (submitted for publica-
couver 2019, pp. 1–12. tion, under review).
[21] J. Quist, C.M. Evertsson, Simulating pressure distribution in High Pressure Grinding [37] P.F. Van der Meer, I.A. Leite, Aspects of HPGR in iron ore pellet feed preparation, 6th
Rolls using the discrete element method, 8th Int. Comminution Symp., Cape Town Brazilian Iron Ore Aggl. Symp, ABM, São Paulo 2018, pp. 102–115.
2012, pp. 1–10. [38] G.E. Dieter, D. Bacon, Mechanical Metallurgy, vol. 3, McGraw-Hill, New York, 1986.
[22] G.K. Barrios, L.M. Tavares, A preliminary model of high pressure roll grinding using [39] K. Meyer, Pelletizing of Iron Ores, 1st ed. Spring-Verlag, New York, 1980.
the discrete element method and multi-body dynamics coupling, Int. J. Miner. Pro- [40] T. Mütze, Energy dissipation in particle bed comminution, Int. J. Miner. Process. 136
cess. 156 (2016) 32–42. (2015) 15–19.
[23] T.M. Campos, G. Bueno, L.M. Tavares, Confined bed breakage of fine iron ore concen- [41] Y.M. Zhang, T.J. Napier-Munn, Effects of particle size distribution, surface area and
trates, Minerals 10 (2020) 666. chemical composition on Portland cement strength, Powder Technol. 83 (1995)
[24] J. Liu, K. Schönert, Modelling of interparticle breakage, Int. J. Miner. Process. 44 245–252.
(1996) 101–115.
255