SciELO - Scientific Electronic Library Online

vol.70 issue3Geological modeling of a stratified deposit with CAD-Based solid model automationThe Challenge to Scavenge IRON from Tailings Produced By FLOTATION A New Approach: The Super-WHIMS & the BigFLUX Magnetic Matrix author indexsubject indexarticles search
Home Pagealphabetic serial listing  

Services on Demand




Related links


REM - International Engineering Journal

On-line version ISSN 2448-167X

REM, Int. Eng. J. vol.70 no.3 Ouro Preto July/Sept. 2017 


Simulation of the Mineração Serra Grande Industrial Grinding Circuit

Thiago Oliveira Nunan1 

Homero Delboni Junior2 

1Engenheiro de Processo, AngloGold Ashanti, Nova Lima - Minas Gerais - Brasil.

2Professor, Universidade de São Paulo - USP, Escola Politécnica, Departamento de Engenharia de Minas e Petróleo.


Increasing throughput during the mining cycle operation frequently generates significant capital gains for a company. However, it is necessary to evaluate plant capacity and expand it for obtaining the required throughput increase. Therefore, studies including different scenarios, installation of new equipment and/or optimization of existing ones are required. This study describes the sampling methodology, sample characterization, modeling and simulation of Mineração Serra Grande industrial grinding circuit, an AngloGold Ashanti company, located in Crixás, State of Goiás, Brazil. The studied scenarios were: (1) adding a third ball mill in series with existing two ball mills, (2) adding a third ball mill in parallel with existing mills, (3) adding a vertical mill in series with existing mills and (4) adding high pressure grinding rolls to existing mills. The four simulations were carried out for designing the respective circuit, assessing the interference with existing equipment and installations, as well as comparing the energy consumption among the selected expansion alternatives. Apart from the HPGR alternative, all other three simulations resulted in the required P80 and capacity. Among the three selected simulations, the Vertimill alternative showed the smallest installed power.

Keywords: modeling; grinding; ball milling; vertical milling; simulations

1. Introduction

Mineração Serra Grande is a gold mining operation located in Crixás, State of Goiás, Brazil. The beneficiation plant processes gold ore from three underground and one open pit mines. The current process includes multi-staged crushing, followed by ball milling in closed configuration with hydrocyclones. A gravity concentration circuit is fed by part of the circulating load, while the grinding circuit product is thickened and leached with sodium cyanide. After leaching, the pulp is filtered, clarified and precipitated with zinc (Merrill Crowe process). The solid tailings are pumped to the tailings dam. Gold is thus produced from both Merrill Crowe and gravimetric circuits. Figure 1 shows the current Serra Grande plant flow sheet. Mineração Serra Grande (MSG) started its operation in October 1989 with a single ball mill, processing 1,200 t of ore per day. Currently, plant capacity is approximately 3,600 t/day.

Figure 1 Mineração Serra Grande Flowchart 

In 2008, the circuit was expanded by installing new equipment, together with various other actions, such as employing a better pumping system, hydrocyclone optimization, adequate ball charge, installing grates in the existing ball mill, as well as automation in the circuit. Further production increase was then focused on installing new equipment.

Figure 2 shows plant production and gold grade from 1990 to 2015. The chart shows a step change in gold production when the second ball mill was installed (2009), followed by a steady increase in following years resulting from optimization,together with a declining gold feed grade.

Figure 2 Plant production and gold grade history of Mineração Serra Grande 

MSG is currently studying alternatives for increasing current plant capacity from 1.3 MTPY to 2.0 MTPY. Apart from a 54%, increase in the current production, such an expansion would also result in further performance improvement by reducing operating costs.

2. Materials and methods

Sampling and data collection

This study began with a literature review to perform a survey campaign on the existing grinding circuit. The aim of sampling was to reduce the mass of a lot, without assigning significant changes to its properties. Data collection followed the sampling rules as proposed by Gy (1982).

Each selected stream was sampled for two hours during a steady-state period of the grinding plant. In some streams, automatic sampling systems were used, while manual sampling was carried out at all remaining selected points, as shown in Table 1.

Table 1 Information about the sampling campaign. 

Sampling Procedure Sampling point
Manual Mill discharges
Automatic Hydrocyclone feed
Manual Hydrocyclone Underflow
Automatic Hydrocyclone Overflow
Automatic New Mill Feed

Table 2 shows the grinding circuit operating data as obtained during the sampling period for mass balance calculations.

Table 2 Process data obtained during sampling period. 

Parameter Unit Mill_01 Mill_02
New Feed t/h 114.9 42
Mill discharge Pulp Density t/m3 1.54 1.47
Water Flow in Mill discharge m3/h 110 50
Cyclone Feed Pulp Density t/m3 1.54 1.47
Cyclone Overflow Pulp Density t/m3 1.25 1.18
Cyclone Pressure Bar 0.71 0.58
Motor Power kW 1061 406
Bin Level % 72.2 39.5

Table 3 shows the equipment main characteristics as currently installed at MSG industrial grinding circuit.

Table 3 Equipment main characteristics. 

Parameters Circuit1 Circuit 2
Ball mill internal diameter (m) 3.48 2.57
Ball mill internal length (m) 5.34 3.74
Hydrocyclone Diameter (m) 0.508 0.508
Number of hydrocyclones in operation and installed 3/5 1/2

Further information about sampling of this work can be found in Leite (2016).

Ore characterization

Samples obtained in the survey campaigns were sent to the Laboratory of Simulation and Control (LSC) of the University of São Paulo for screening, as well as for specific gravity assessment and comminution testing, which included the Bond Work Index, Drop Weight Test, Piston Press Test and Jar Mill Grinding Test.

The Bond Work Index (BWI) was performed to estimate energy requirements for ball milling using the Bond equation shown below, together with the Rowland (1982) efficiency factors - EF.

E=10Wi1P801F80·EFi (1)

Drop Weight Tests (DWT) were also performed to calibrate the Whiten´s ball mill model, while a Piston Press Test (PPT) was carried out to calibrate the High Pressure Grind Roll (HPGR) model used throughout simulations. Both DWT and PPT procedures were carried out according to Napier-Munn et al., 1996. DWT and PPT resulted in A and b parameters, as obtained from equation 2, together with respective breakage matrices.

t10=A1ebEcs (2)

Jar Mill Grinding Test (JMGT) was performed to estimate energy consumption for an industrial vertical mill. The energy calculation for the JMGT was carried out through equation 3, following Metso procedures described by Wills, 2016.

E=6.3.D0.3.sen51222.44D2.44·3.23Vp·CS·10.12910Cs·t.mbmm60 (3)

E = specific energy consumed during JMGT; D = mill internal diameter;

Vp = mill volume fraction filled with grinding media; Cs = fraction of the mill critical speed; t = time jar operation; mb = media mass; mm = ore mass

Mass balancing was carried out using experimental data obtained during the sampling period. This procedure included estimating best flow rates and size distributions around the entire grinding circuit.

Equipment and process models

The Nageswararao (2004) model was used for modeling the industrial hydrocyclones. The model includes both operation and design data, together with partition curve parameterization. Calibration constants were back calculated for model fitting exercises.

The adapted Perfect Mixing Model proposed by Whiten (1976) was used to model industrial ball milling.

The grinding kinetic parameter (r/d*) was determined for each ball mill during the model fitting exercises, as described by Napier-Munn (1996).

The HPGR model proposed by Morrell/Tondo/Shi (1997) includes three breakage zones i.e. the pre-crusher zone, the edge effect zone and the compression zone. The throughput model component uses a standard plug flow model version that has been used extensively by manufacturers and researchers. Power consumption is based on throughput and specific comminution energy input. (Morrell et al., 1997).

3. Results and discussion

Ore characterization

The BWI test performed in the surveyed grinding circuit feed sample resulted in 11.6 kWh/sht. Such a value was used to estimate the overall grinding circuit energy consumption. The combination between such an energy consumption and the stipulated 2.0 MTPY resulted in 624 kW power to be installed in the additional parallel ball mill.

The appearance function and breakage parameters as obtained from DWT, carried out on surveyed samples are shown in Tables 4 and 5.

Table 4 Appearance Function data - DWT. 

t10 t75 t50 t25 t4 t2
10 2.9 3.7 5.6 22.7 49.5
20 5.8 7.5 11.2 42.5 77.3
30 8.7 11.2 16.8 59.4 90.4

Table 5 Breakage parameters - DWT. 

A 54.4
b 0.84
IQ 45.4

The appearance function and breakage parameters as obtained from PPT carried out on -6.35 +4.75 mm size fraction are shown on Tables 6 and 7.

Table 6 Appearance Function data - PPT. 

t10 t75 t50 t25 t4 t2
10 3.3 4.3 6.3 17.4 31.7
30 10.6 13.4 19.1 46 72.1
50 18 22.7 31.4 72.7 100

Table 7 Breakage parameters - PPT. 

A 28
b 1.44
IQ 40.3

JMGT was carried out for 3, 5 and 10 min grinding periods. Table 8 shows the results obtained in terms of specific energy and resulting product P80.

Table 8 Batch grinding test results. 

Time (min) Specific Energy kWh/t P80 (mm)
0 0 165
3 2.54 81.9
5 4.24 60.7
10 8.48 43.8

Model calibration

The obtained sample values are similar to the calculated data and resulted in a consistent mass balance, as well as adequate fitted models.

Figure 3 shows experimental and calculated size distributions obtained for each individual stream around the MSG industrial grinding circuit.

Figure 3 Experimental and simulated size distributions as obtained for individual grinding circuit streams 

The calibrated parameters obtained from hydrocyclone modeling are shown in Table 9, while Table 10 shows the parameters obtained from ball mill modeling.

Table 9 Nageswararao hydrocyclone model calibration for MSG hydrocyclones. 

Model Parameters Hydrocyclone Nest 1 Hydrocyclone Nest 2
D50 Constant - KD0 8.14E-05 7.89E-05
Capacity Constant - KQ0 510.7 601.9
Volume Split Constant - KV1 7.15 9.11
Water Split Constant - KW1 10.66 14.44
Sharpness of Efficiency Curve - Alpha 2.01 2.81

Table 10 Whiten model calibration for MSG ball mills. 

Knot Size (mm) Ln (r/d *)
Mill_01 Mill_02
1 0.2 1.674 1.503
2 1.5 3.712 3.966
3 5 5.232 5.733


Four circuit alternatives were assessed through simulations for increasing the current 1.3 MTPY capacity to the stipulated 2.0 MTPY for the expansion project. Each alternative was simulated to obtain the respective mass balance and equipment design, together with the installed power and energy consumption. Simulations were carried out with calibrated models using JKSimMet 6.0 software.

Each simulated alternative is described as follows.

Alternative 1 - Additional ball milling line in series

The first alternative consisted in simulating an additional ball mill in the existing grinding circuit. The third ball mill would regrind the product of the two existing ball mills, as shown in the Figure 4 flow sheet.

Figure 4 Additional ball milling stage flow sheet 

The two existing ball mill lines were thus simulated for the 2.0 MTPY increased throughput, therefore producing a relatively coarser product, in this case a P80 equals to 165 µm. The third ball mill was thus designed to grind such an intermediary product to the stipulated P80 of 109 µm.

The designed ball mill showed 3.2 m in diameter and 4.6 m in length, operating at 35% ball charge, 70% critical speed and 60 mm steel ball top size. The calculated ball mill installed power was 618 kW-.

Alternative 2 - Addition ball milling line in parallel

The second alternative comprised of simulating an additional ball milling line in parallel with the two existing ones, as shown in the Figure 5 flow sheet.

Figure 5 Additional ball milling line flow sheet 

The designed ball mill resulted in the same dimensions as obtained in Alternative 1, i.e. 3.2 m in diameter and 4.6 m in length, operating at 35% ball charge, 70% critical speed and 60 mm steel ball top size. The calculated ball mill installed power was 618 kW-.

Alternative 3 - Additional vertical mill

The third alternative consisted in simulating a vertical mill to regrind the product from the existing two ball mills. Figure 6 shows the simulated circuit flow sheet.

Figure 6 Vertical mill flow sheet 

As per Alternative 1, the existing ball mill circuit product showed a P80 of 165µm for processing 2.0 MTPY.

In order to calculate the required energy for a vertical mill in reducing the P80 from 165µm (feed) to 109 µm (product), the graph showed in Figure 7 was used. Such a graph resulted from the JMGT carried out specifically for such a purpose. According to Figure 7, the required energy for such an operation was calculated as 1.71 kWh/t, which resulted in 416 kW for a 243 t/h throughput. A Metso VTM-800 was selected considering safety factor suggested by the manufacturer Wills, 2016.

Figure 7 Estimated P80 as a function of the specific energy for different grinding times - JMGT 

Alternative 4 - Additional HPGR

The fourth alternative included a HPGR in a single pass (open circuit) for providing a finer size distribution to the existing ball mills. Such a finer size distribution would thus increase the installed ball milling capacity to the required 2.0 MTPY. Figure 7 shows the simulated circuit flow sheet.

Figure 8 HPGR circuit flow sheet 

Based on simulation results, the selected equipment was one that had 1200 mm in roll diameter by 750 mm in roll length, with a 6.35 mm working gap, 324 ts/m³h specific throughput (m dot) and 1.48 m/s roll speed.

Even though the simulations indicated that the existing grinding circuit would only achieve the required capacity of 2.0 MTPY for a finer feed, HPGR benchmarking indicated that a realistic product would not be finer than a 2500 µm P80. For such a feed size distribution, the existing grinding circuit product would show P80 of 141 µm, therefore coarser than the required value (P80 of 109 µm).

Alternative comparison

A summary is shown in Table 11of the equipment selected for the simulated alternatives with required power, installed power and P80 of the product for each case.

Table 11 Additional equipment selection summary. 

Simulated Alternative Description Aditional Euipament Required Power (hp/kW) Installed Power(hp/kW) Grinding Circuit Product P80 (mm)
1 Second ball miling stage Ball Mill,o=3.2m, L=4.5m 830/618 900/671 105
2 Additional ball milling line Ball Mill,0=3.2m, L=4.5m 830/618 900/671 105
3 Vertical mill stage Vertical Mill - VTM 800 724/540 800/597 105
4 Additional crushing stage with HPGRR HPGR0=1.2m, L=0.75m 837/624 590/880 141

4. Conclusions

The grinding circuit of Mineração Serra Grande was surveyed for obtaining consisted and representative operating data, which in turn were used for model fitting and simulations, the latter using JKSimMet simulator.

BWI, DWT, PPT and JMGT were carried out on selected samples for obtaining comminution characterization parameters. Four simulation alternatives were selected for increasing the grinding circuit capacity from current 1.3 MTPY to 2.0 MTPY. In each case the simulations resulted in designing the additional crushing/grinding equipment, and respective installed power. The alternatives included (1) an additional ball milling stage, (2) an additional ball milling line, (3) an additional Vertimill stage and (4) an additional crushing stage by using a HPGR piece of equipment.

Apart from the HPGR alternative, all other three resulted in the required P80 for the 2.0 MTPY circuit capacity. Among the three selected simulations, the Vertimill alternative showed the smallest installed power.


The authors acknowledge the support provided by AngloGold Ashanti throughout the entire work.


GY, P. M. Sampling of particulate materials theory and practice. Amsterdam: Elsevier, 1982. [ Links ]

LEITE, T. O. N. Modelagem e simulação do circuito de moagem da Mineração Serra Grande. São Paulo: Escola Politécnica, 2016. 195 p. [ Links ]

MORRELL, S., SHI, F., TONDO, L. A. 1997. Modelling and scale-up of high pressure grinding rolls. In: INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC), 20. Proceedings... Germany, 1997. [ Links ]

NAGESWARARAO, K. et al. Two empirical hydrocyclone models revisited. Minerals Engineering, v. 17, n. 5, p. 671-687, 2004. [ Links ]

NAPIER-MUNN, T. J. et al. Mineral comminution circuits: their operation and opti-mization. Indooroopilly: SMIJKRMC, 1996. 413 p. [ Links ]

ROWLAND, C. A. Selection of rod mills, ball mills, pebble mills and regrind mills. In: Design and installation of comminution circuits. New York: SME/AIME, 1982. p. 393-438. [ Links ]

WHITEN, W. J. Ball mill simulation using small calculators. Proceedings AusIMM... p. 47-53, 1976. [ Links ]

WILLS, B. A., FINCH, J. A. Wills' mineral processing technology: an introduction to the practical aspects of ore treatment and mineral recovery. Butterworth-Heinemann, 2015. [ Links ]

Received: November 05, 2016; Accepted: April 25, 2017

Creative Commons License This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.