Predicting the Performance of An Industrial HPGR With Severely Worn Rollers Using

DEM

V. A. Rodriguez, G. K. P. Barrios, T. M. Campos, H. A. Petit, *L. M. Tavares

Department of Metallurgical and Materials Engineering

Universidade Federal do Rio de Janeiro—COPPE/UFRJ
Athos da Silveira Ramos Av., 21941-972, Rio de Janeiro, RJ, Brazil

(*Corresponding author: [email protected])

Abstract

Applying dynamic modeling and simulation of high-pressure grinding rolls (HPGR) has grown in the last few years
in the minerals industry. In that regard, a pseudo-dynamic modelling approach based on the modified Torres and
Casali phenomenological model has been successfully applied to describe the continuous operation of the HPGR
for grinding iron ore concentrates. The model is capable of accurately describing the capacity, power, and Blaine
specific surface area (BSA) of the product based on operating conditions such as working gap, operating pressure,
roll velocity, and feed BSA. However, the model presents a deviation in predicting the performance variables as
rollers become progressively worn during months and years of operation. In general, for this particular
application, the rollers can operate up to 15,000 hours before being changed or repaired, during which the wear
profile has a direct influence on the process control strategy to keep the equipment capacity and the product
BSA. To better understand what happens inside the HPGR with the worn rollers, the present work simulated an
industrial-scale HPGR using the discrete element method (DEM) coupled to multibody dynamics and the particle
replacement model. Simulation results such as throughput, power, and working gap are compared with an
industrial survey, demonstrating good agreement between the two. The DEM simulation allows analyzing the
pressure and mass flow profiles along the roll length for both new rollers and worn rollers. DEM simulations were
able to describe the reduction in power, throughput, and fineness of the product-size distribution, as well as the
significant drop in pressure and an increase in mass flow in the centre of the rolls, which is the area that exhibits
the most severe wear.

Keywords

High-pressure grinding roll, DEM, simulation, modelling

Introduction

Modelling and real-time prediction of the performance variables of high-pressure grinding roll (HPGR) operation
presents a challenge, mainly at the industrial scale. Today, advanced modelling tools based on artificial
intelligence, data analysis, and phenomenological modelling can simultaneously assess multivariable
relationships between HPGR operational variables and predicted product targets (Chelgani et al., 2021; Tohry et
al., 2021). The HPGR can experience a great deal of roll wear, particularly when comminuting hard ores—due to
abrasion and indentation, significantly affecting the HPGR’s performance and service life (Lim et al., 1997).
Campos et al. (2023) proposed a pseudo-dynamic model (PDM) and successfully applied it to HPGRs pressing
iron ore concentrate at a VALE pelletizing plant. The PDM described throughput, power, and product fineness of
the HPGR, using online information on the operational variables such as roll speed, working gap, operating
pressure, and feed particle-size distribution (Campos, 2023). However, the PDM presents some deviations in the
description of the throughput explained by the change in the effective distance between the rolls (working gap
+ surface wear). A good strategy to study such phenomenon and make predictions of the HPGR performance are
simulations using the discrete element method (DEM). Rodriguez et al. (2021) studied the influence of the wear
pattern of the HPGR rolls in the throughput, compressive force, and product fineness, using the DEM coupled to
multi-body dynamics (MBD) describing the movable roll, in addition to the particle replacement model (PRM).
The present study applies the approach used by Rodriguez et al. (2021) to simulate the real wear profile of an
industrial-scale HPGR with 13,656 hours (h) of operation. The results of the performance variables of the worn
rolls are compared with those of new rolls; showing that due to the worn rolls a significant drop in pressure, an
increase in mass flow in the centre of the rolls (higher wear portion), and coarser product is observed.

Materials and Methods

We selected for this case study an industrial-scale HPGR from the production line of the Complexo de Tubarão
(Vitória, Brazil), pressing iron ore concentrates. The machine has a roll diameter of 2.25 metres (m), length of
1.55 m, with studs and check plates as the confinement system. It is connected to a supervisory system that
allows capturing information about operating pressure, operating gap, roll peripheral velocity, throughput, and
power consumption. Additional information on the machine can be found in Campos et al. (2021). In the present
work, information from the supervisory system was collected over a period of 13,656 h, where samples were
also collected at 4-hour intervals in both HPGR feed and product; these samples were analyzed to measure the
BSA using a Zünderwerke Ernst Brün GmbH PCBlaine-Star. To investigate the wear profile along the roll length,
the method relies on simply measuring the distance between the top of the studs and a metal strip placed in
front of the rolls along their axial position. Digital calipers were used to measure the distance between the roll
surface and the metal strip periodically, at 63 studs selected in both rolls, which allowed us to assess the rolls
from the beginning of roll lifetime up to nearly 15,000 h of operation.

Due to the large number of particles demanded in HPGR simulations, the authors selected a coarser particle-size
distribution, with a top size of 8 mm, to represent the iron ore pellet feed. This approach was already successfully
used in this application in previous studies (Rodriguez et al., 2021, 2022). From this and bench-scale tests, which
included measurement of the static and dynamic angles of repose (Rodriguez et al., 2022a), material and contact
parameters were calibrated. Additional material properties were estimated based on a piston-and-die test,
resulting in the selecting values of Poisson’s ratio of 0.25, particle density of 3,011 kilograms per cubic metre
(kg/m3) and shear modulus of 2 gigapascals (GPa) (Rodriguez et al., 2022a). Finally, values of breakage model
parameters were taken from previous publications (Rodriguez et al., 2021, 2022a). Besides the scaled particles,
the particles in the hopper level are an important aspect of the DEM simulation of HPGRs (Cleary & Sinnott, 2021;
Nagata et al., 2020; Rodriguez et al., 2022b). However, the large number also increases the computational effort,

1 | SAG CONFERENCE 2023 VANCOUVER | June 23, 2023

and it is recognized that the most significant effect of these particles is in applying pressure to the particle bed
sitting between the rolls. The strategy proposed by Rodriguez et al. (2022b) was adopted, where the final portion
of the hopper silo was filled with particles of even coarser size (60 mm diameter) and density (10,000 kg/m3)—
called column-pressure particles—to match the same pressure that would have been generated by piling up
additional feed particles. Computer aided design (CAD) drawings of the simulated HPGR unit were generated.
Two designs were carried out, one with the new rolls and the other representing the roll wear after 13,656 h of
operation. A force controlled in the longitudinal direction, x, described the translational dynamics in the floating
roll, while only the kinematics of the rotating movement in both rolls with a fixed angular velocity was described.
Being a one-dimensional model, it does not allow for skewing of the rolls, which was the object of another
publication (Rodriguez et al., 2022c). The roll movement started only when DEM particles filled the hopper from
a particle factory above the rolls. From this point on in the simulation, new particles were added continuously
until steady-state conditions were reached, then analysis was performed. To analyze the performance in the
virtual HPGR, the DEM simulation directly provides the total throughput and power of the HPGR, as well as axial
distribution along the rolls. The total power was calculated based on the torque required to rotate the rolls
multiplied by the rotation frequency. The mass flow rate in each block, and axial compressive force profiles were
estimated from a grid bin in the compression zone, presented in detail in Rodriguez et al. (2022a).

Results and Discussions

The results of the industrial survey are presented in Table 1. The pressure is reduced during the HPGR operation,
fact that directly affecting power consumption in the HPGR performance, decreasing it by 52% after 13,656 h.
During normal HPGR operation with new rolls, as the operating pressure decrease, both working gap and
throughput increase. However, Table 1 presents the opposite trend, with about a 60% reduction in gap and 3.2%
increase in throughput. The results of measurements taken at the metal strip during the operation are presented
in Figure 2. At initial operational times (6,504 h), smaller wear rates of the studs are observed (0.76 microns per
hour [µm/h]), while with additional operation time this rate increases to around 2.09 µm/h to reach the larger
20 mm in the centre portion of the roll. Until the first 6,504 h of operation, the stud's wear could be considered
homogeneous, with a tendency to the "bathub" profile along the rolls. However, as the HPGR operates, studs in
the centre zone suffer preferential wear, generating a parabolic profile. The larger gap of 20 mm in the centre
zone because of the wear could explain the relatively modest change in throughput, despite the working gap
reduction and considering that variations in feed characteristics are relatively limited.

Table 1—Summary of operating conditions and key variables of the industrial HPGR
Operational Time (h)
Operating Variables 0 1,248 4,704 6,504 13,656
Operating Pressure (bar) 75.0 75.1 56.7 66.8 44.5
Working Gap (mm) 12.4 13.1 12.4 9.8 5.1
Roll Velocity (m/s) 1.00 0.94 1.05 1.01 1.01
Power Consumption (kW) 1702 1447 1423 1541 1032
Throughput (t/h) 602.0 514.1 623.8 577.4 549.2
BSA Feed (cm²/g) 1720 1573 1606 1579 1589
BSA Product (cm²/g) 2070 1881 1859 1847 1834
Note: m/s = metres per second; kW = kilowatt; t/h = tonnes per hour; cm²/g = square centimetres per gram.

2 | SAG CONFERENCE 2023 VANCOUVER | June 23, 2023

 Figure 1—Wear profile measurements at the metal strip, taken on the industrial HPGR during 13,656 h of
operation; operating conditions are listed in Table 1.

The new and worn rolls after 13,656 h of operation (Figure 1) was designed in CAD to be simulated using the
DEM–MBD–PRM coupling according to the operational conditions presented in Table 1. DEM simulations were
able to capture the general trends from the experimental results (Table 2), since reductions in working gap,
throughput and power were observed as wear on the rolls progressively increased. Although the coarse particles
simulated in the feed prevent a direct comparison to experiments, a qualitative comparison is possible, which
shows that simulations captured the progressive reduction in BSA of the product as wear progressed. However,
a quantitative comparison between simulation results and surveys in Table 2 shows that deviations varied from
2 to as much as 29% for the various measures, in particular in the case of simulations where the rolls were fully
worn. The larger differences in the throughput predictions could be, at least in part, explained by the value of
particle density (3,011 kg/m3) calibrated in the simulation of the new rolls (Rodriguez et al., 2022a), which is
based on the flake density of the product under this condition (Campos et al., 2019). This value is likely to reduce
as the rolls wear, but further investigation are in progress by the research group to improve the DEM simulation
predictions under these conditions. The greatest advantage of a virtual experimenter, such as DEM, is associated
to the ability to illustrate the micromechanics of the machine operation. Figure 2 (left) shows the mass flow
profile along the rolls in simulations of the HPGR with new and fully worn rollers. The mass flow profile to the
new roll present higher values in the central portion of the rolls with a slight reduction towards the edges and
greater material transferred through the bypass gap. After 13,656 h of operation, the material passing through
the central region intensified the drop in the edges, but with similar quantity of materials ejected for the bypass
gap. On the other hand, the force profile, illustrated in Figure 2 (right), is the characteristic one in the new roll
with higher values in the centre and a drop pressure in the edge, while the worn roll generates a profile known
as “bathtub” with a significant drop in force in the middle portion of the rolls.

Table 2—Comparison between the survey data and prediction of the DEM–MBD–PRM simulations operating
with conditions presented in Table 1

New Rolls (0 h) 13,656 h

Operating Condition Survey Simulation % Relative Error Survey Simulation % Relative Error
Working Gap (mm) 12.4 12.1 2.0 5.1 3.6 28.7
Power (kW) 1702 1952 14.7 1032 764 25.9
Throughput (t/h) 602.0 671.3 11.5 549.2 398.5 27.4
1
BSA Increase (cm²/g) 350 1281 – 245 564 –
Note: 1Values in the DEM simulation were calculated using the prediction of the particle replacement model and the
empirical equation proposed by Campos et al. (2019).

3 | SAG CONFERENCE 2023 VANCOUVER | June 23, 2023

Figure 2—Mass flow (left) and axial compressive force (right) profile along the rolls of the DEM simulation of an
industrial HPGR with new rolls and after 13,656 h of operation; vertical grey dashed lines represent the
projection of the rolls length

Conclusions

An industrial survey of an HPGR performed was carried out to study the influence of roll wear on key variables.
During 13,656 h of operation, throughput, power, working gap, and wear profile were measured showing minor
wear rates at the initial operations times (6,500 h) and then a significant increase reaching 20 mm of wear in the
centre zone of the rolls.

The DEM-MBD-PRM coupling has been used to study the influence of wear on the performance of this industrial
HPGR. Predictions were able to capture the general trends observed in the surveys, namely, the reduction in
operating gap, throughput, power, and product fineness when comparing the new rolls with those after 13,656 h
of operation. Quantitative comparisons between simulations and surveys showed that deviations were moderate,
especially when results for the worn rolls were compared.

A more detailed analysis of the simulation demonstrated that wear resulted in a highly heterogeneous force
profile along the rolls because of the presence of around 20 mm of wear in the centre of the roll. Such a profile
led to a significant reduction in power and product fineness, as also observed in plant operation. The mass flow
profile also was also significantly different in both cases, with a relatively homogeneous pattern for the new rolls,
whereas there was higher mass transfer in predicted in the centre of the rolls due to the severe wear in this
region. These changes in the axial performance of the HPGR due to the worn roller could be adapted in the PDM
(Campos, 2023) and the hybrid DEM–PBM approach (Rodriguez et al., 2023) to make predictions of the product
particle-size distribution of the rolls. In spite of its limitation in quantitatively describing the fineness of the
product, the simulation approach was demonstrated to be a powerful tool in the design of HPGRs in the future.

Acknowledgments

The authors are grateful for funding by the Brazilian Agencies FAPERJ (grant number E-26/205.848/2022),
Conselho de Desenvolvimento Tecnológico e Científico (CNPq), and the Vale S.A. for the technical and financial
support.

4 | SAG CONFERENCE 2023 VANCOUVER | June 23, 2023

References

Campos, T. M, Bueno, G., & Tavares, L. M. (2021). Modeling comminution of iron ore concentrates in industrial-
scale HPGR. Powder Technology, 383, 244–255. https://doi.org/10.1016/j.powtec.2020.12.075

Campos, T. M., Petit, H. A., Freitas, R. O., & Tavares, L. M. (2023). Online prediction of pressing iron ore
concentrates in an industrial HPGR. Part 1: Modeling approach. Minerals Engineering, 201(December 2022),
108206. https://doi.org/10.1016/j.mineng.2023.108206

Chehreh Chelgani, S., Nasiri, H., & Tohry, A. (2021). Modeling of particle sizes for industrial HPGR products by a
unique explainable AI tool- A “Conscious Lab” development. Advanced Powder Technology, 32(11), 4141–
4148. https://doi.org/10.1016/J.APT.2021.09.020

Cleary, P. W., & Sinnott, M. D. (2021). Axial pressure distribution, flow behaviour and breakage within a HPGR
investigation using DEM. Minerals Engineering, 163(December 2020), 106769.
https://doi.org/10.1016/j.mineng.2020.106769

Lim, W. I. L., Campbell, J. J., & Tondo, L. A. (1997). The effect of rolls speed and rolls surface pattern on high
pressure grinding rolls performance. Minerals Engineering, 10(4), 401–419.
https://doi.org/10.1016/S0892-6875(97)00017-4

Nagata, Y., Tsunazawa, Y., Tsukada, K., Yaguchi, Y., Ebisu, Y., Mitsuhashi, K., & Tokoro, C. (2020). Effect of the roll
stud diameter on the capacity of a high-pressure grinding roll using the discrete element method. Minerals
Engineering, 154(May), 106412. https://doi.org/10.1016/j.mineng.2020.106412

Rodriguez, V. A., Barrios, G. K. P., Bueno, G., & Tavares, L. M. (2021). Investigation of Lateral Confinement , Roller
Aspect Ratio and Wear Condition on HPGR Performance Using DEM-MBD-PRM Simulations. Minerals, 11,
1–21. https://doi.org/https:// doi.org/10.3390/min11080801

Rodriguez, V. A., Barrios, G. K. P., Bueno, G., & Tavares, L. M. (2022a). Coupled DEM-MBD-PRM simulations of
high-pressure grinding rolls. Part 1: Calibration and validation in pilot-scale. Minerals Engineering,
177(January), 107389. https://doi.org/10.1016/j.mineng.2021.107389

Rodriguez, V. A., Campos, T. M., Barrios, G. K. P., Bueno, G., & Tavares, L. M. (2022b). A Hybrid PBM-DEM Model
of High-Pressure Grinding Rolls Applied to Iron Ore Pellet Feed Pressing †. KONA Powder and Particle
Journal, 40(July), 262–276. https://doi.org/10.14356/kona.2023011

Rodriguez, V. A., Barrios, G. K. P., & Tavares, L. M. (2022c). Coupled DEM-MBD-PRM simulations of high-pressure
grinding rolls . Part 2 : Investigation of roll skewing. Minerals Engineering, 178(October 2021), 107428.
https://doi.org/10.1016/j.mineng.2022.107428

Tohry, A., Yazdani, S., Hadavandi, E., Mahmudzadeh, E., & Chelgani, S. C. (2021). Advanced modeling of HPGR
power consumption based on operational parameters by BNN: A “Conscious-Lab” development. Powder
Technology, 381, 280–284. https://doi.org/10.1016/J.POWTEC.2020.12.018

5 | SAG CONFERENCE 2023 VANCOUVER | June 23, 2023