INCREASING THE PRODUCTIVITY IN A MAGNESITE PILOT PLANT COMBINING OPEN BALL MILL CIRCUIT AND HIGH FREQUENCY SCREENER

Open circuit and high frequency screener have already been adopted by some mining operations, however not taking advantage of its full potential. This work presents a new approach, combining them to have a better use of the available energy and, thus, increasing productivity. This concept could fit not only the studied case but also any other ore, since it is based on the milling inner processes and the external classification. The studied magnesite process has currently four parallel ball mills operating in closed circuit with hydrocyclones, grinding the ore for the silicates flotation cells. The modification proposed changes the circuit to operate with three mills in open circuit with parameters that improves inner classification and particles transport and breakage. Their product are collected together to feed a high frequency screener, a more efficient equipment when compared to hydrocyclones. The screener oversize feeds a fourth ball mill to continue grinding and prepare the ore for the next steps, dewatering and reverse flotation. On the other hand, the screener undersize goes to desliming and reverse flotation equipment. According to the data collected in the pilot plant and the modeling for the industrial site, this modification could promote 27% increase in the productivity, due to 17.6% increase on mills feed and 7.8% in mass recovery, caused by loss reduction on desliming and flotation. The gain is expressive due to some inefficiency of the current circuit and still need to be validated on the industrial site.

However, to take advantage of these characteristics, the circuit must count on an efficient classification process.The high frequency screener, recently developed for industrial mining process, has been proving some advantages when compared to the traditional hydrocyclone for closed circuits [6].
Nevertheless, none of these works have combined the benefits of open circuit and high frequency screener to increase the circuit productivity.

Milling Modeling
The comminution modeling was performed using the population balance model, as reviewed by Mazzinghy and Alves [7,8].The modeling is based on the specific parameters determination on both selection function and breakage function.
The selection function is determined by the parameterization of 0 ∝ , 1 ∝ , 2 ∝ and crit d , discribed by equation 1 [9]: On the other hand, the breakage function is modeled by the variables 0 β , 1 β and 2 β , showed by equation 2 [10]: ( ) Finally, these parameters must be associated to the available energy for comminution and the equipment size, as demonstrated by equations 3 and 4 [11,12]: ( ) Using these equations, it is possible to predict the particle size distribution given the feed particle size and the required mass rate.

Milling Efficiency Comparison
The Bond Efficiency was established to compare industrial circuits and has been used to measure industrial efficiency [13].It compares the operational work index with

INTRODUCTION
Magnesite, apart from sea water magnesia, is the main feedstock for dead burned magnesia (DBM) production, one of the most important raw materials for basic refractories.RHI Magnesita is the most important DBM producer in Latin America, in its Brumado, Bahia State, Brazil, operation.The ore is extracted from an open pit mine named Pomba.
Despite having one of the purest magnesite reserves, silicates still must be removed from the Pomba ore, in order to achieve the desired DBM chemical composition.In order to do so, the ore has to be ground prior to reverse flotation, so the silicates can be separated from magnesite.This concentration is important to produce a high performance DBM, and the magnesite is transformed into magnesia trough a double step firing in high temperature shaft kilns.
The current concentration process starts with an impact crusher, preparing the particle size distribution to feed the ball mills.The milling operation has four ball mills working in closed circuit with hydrocyclones.This conventional circuit promotes a high circulating load and overgrinding, thus reducing the selectivity in the following operations, desliming and flotation, which is detrimental to mass and metallurgical recoveries.
The recent Chinese environmental measures reduced the DBM availability in the market, since China is the key magnesia producer, pushing the prices up and the demand for other DBM sources.Given this scenario, the Brumado operation aims to increase the DBM production and the concentration process was identified as the bottleneck [1].

MATERIALS AND METHODS
Firstly, a review on grinding concepts was carried out to determine the milling parameters and draw up the pilot plant trials.Also, the comminution mathematical modeling was revisited to base the scale up procedure.Afterwards, the pilot plant tests were drawn up to meet the selected condition and the milling operation was modeled.

Milling Parameters
An efficient milling takes advantage of the ore mineralogy and slurry rheology to promote a comminution with better energy application, improved inner classification and enhanced particles transportation, consequently, using less energy to liberate the ore [2].
The key parameters to carry out an efficient milling are solids percentage, media charge and circulating load.The optimal solids percentage is not too high that could be detrimental to internal classification and transportation and not too low that could decrease the probability of the particles to suffer media impact [3].The media charge objective is to promote impact and transfer the energy supplied to the mill, and the filling should not be high enough to reduce ( ) Since the Bond work index is a standard test often used for mills sizing, if the operational work index is lower than the Bond value, the operation works is surpassing its designed conditions.Therefore, the higher the Bond efficiency the better is the process.

Pilot Plant Design
The pilot plant flowsheet was designed to simulate the proposed industrial circuit with equipment that is adequate to obtain the modeling parameters, especially for the comminution circuit.
Two tests were drawn up, the first one reproducing the first milling, with open circuit and selected conditions, to acquire parameters for modeling.In this same circuit, following the mill, a high frequency screen took place to carry out the classification.The screener undersize proceeded to desliming and flotation and the oversize was stored for the second test, simulating the secondary milling.
Since the secondary milling was never tested for this ore before and the operational work index was unknown, the primary milling conditions were used on the second pilot trial, in order to model this process; the desliming and flotation were performed in bench scale equipment with samples of the mill product.
Figure 1 displays the pilot plant flowsheet and Table 1 shows the pilot mill dimensions and the parameters chosen for the trials.
As mentioned before, the secondary milling behavior was uncertain.However, it was chosen to sustain the primary milling conditions and alter the necessary variables on the scale up, through simulation, in order to achieve the desired productivity and particle size distribution.
Considering the pilot plant designed and the parameters chosen, the trials were then carried out.

Ore Characteristics
The analyzed ore mineralogy is composed mainly by magnesite, talc, chlorite and quartz, identified by X-Ray Diffraction.The main contaminant for the product is SiO 2, its concentration in the process feed is approximately 3% and it must reach less than 0.30% to achieve the specification, measured by X-Ray Fluorescence in calcined basis.The MgO content is also important; it has to be above  98% in the concentrate, starting near to 96% on the feed using the same analytical methodology.

Modeling Software
The Moly-Cop Tools was used for milling modeling and simulation.This software was developed by a grinding media supplier (Moly-Cop) to aid its customers to improve their processes.

Primary Milling Test, Modeling and Scale up
The first evaluated condition concerned the primary milling, modeling the comminution process and analyzing desliming and flotation of the screener underflow.Table 2 shows the main modeled milling parameters and experimental particle size distribution obtained on pilot plant trial.
With these parameters the process was scaled up for the projected industrial site.The industrial circuit has four ball mills equally divided in two sizes and power, 184kW and 110kW.Table 3 exhibits the predicted operation conditions compared to the current parameters.The current situation, for all comparisons, was modeled and simulated based on samples took on the circuit roughly at the same time as the sample for pilot plant trials was taken As shown in table 3, the milling capacity could be increased, using only three mills for primary grinding.However, the fourth mill must be used in the secondary milling, since 30% of the primary milling product is retained in the screener.These results agree with previous works [5,14].
Another key benefit of these conditions is the overgrinding reduction, illustrated in Figure 2.
Since silica is the main contaminant, the analysis was focused on this component.The overground magnesite is picked up among the smaller particles, reducing the SiO 2 content in these fractions.This is unfavorable for the process since energy was spent to grind unnecessarily these particles, so they will be lost in desliming, reducing the mass recovery and, consequently, productivity.This behavior was also observed for different ores [15].

Secondary Milling Test, Modeling and Scale up
The same procedure was adopted for the secondary milling, with the modeling parameters, experimental particle size distributions and hydrocyclone dimensions detailed in table 4.
From the primary milling simulation, the second step would be fed with 30tph, therefore, the scale up was carried out to verify if the available mill could fit this demand.Table 5 illustrates the main characteristics of the secondary milling compared to pilot plant data.
To maintain the flotation selectivity, the secondary milling P 80 could not exceed 212µm.In fact, the simulated   secondary milling product fits within this specification, thus proving that the process is feasible.Is this case, the closed circuit was chosen because the ore is homogeneous both in particle size and mineralogy, reducing milling efficiency and not justifying the investment in another high frequency screener.However, the reduced media charge and solids percentage were maintained, as they aid the inner classification and transport, therefore, increasing the milling efficiency as reported previously [16].
The pilot plant was carried out using open circuit since it is not possible to close circuit with hydrocyclones in this installation.The secondary simulation was carried out combining the pilot plant results and the current hydrocyclones modeling.
With the grinding circuit verified, the flotation was evaluated for both milling circuits.

Flotation
In order to simplify the nomenclature, the process undergone by the screener undersize was named primary milling circuit and that corresponding to the oversize, secondary milling circuit.
The primary milling product was analyzed in the pilot plant, verifying the consequences of different screen openings (300µm and 212µm) and the process recovery.On the other hand, the secondary milling product was analyzed in bench tests.Table 6 exhibits the results obtained.
Considering previous trials, this pilot plant can reproduce the flotation and desliming industrial performance.Hence, the results found in the pilot tests were expected to be reproduced on the industrial process, as shown in Table 6.
Moreover, the screen opening must be carefully analyzed, given the SiO 2 results obtained in the pilot tests.The concentrate generated from the 300µm screen undersize has SiO 2 percentage higher than the accepted value (0.30%), whereas the one produced from the 212µm screen undersize lies within the specification.This result is expected, since the mechanical reverse flotation presents inferior selectivity for coarse particles [17].
Finally, the secondary milling circuit desliming and flotation were evaluated.Nevertheless, these processes were analyzed only in bench scale tests, as the pilot plant does not have the necessary hydrocyclone to provide a closed milling circuit.
As a result, the secondary milling product was dried and screened in a 212µm sieve to produce the material for the tests.Afterwards, the desliming was performed in a 20µm sieve prior to flotation.The results of this procedure are in Table 7.
In this case, fewer fines particles were generated in the secondary milling.Hence, the desliming mass recovery was higher and yet the SiO 2 percentage in the flotation concentrate fell far below the specification.As a matter of fact, in the industrial process the silica content should be higher, because a result this low could be detrimental to the subsequent processes.Therefore, for the designed operation the considered value of SiO 2 was higher and, hence, the recovery would be increased.

CURRENT AND DESIGNED PROCESSES COMPARISON
In summary, Table 8 exhibits the comparison of the current process, using the conventional closed circuit mill, with the proposed selective milling process combining open circuit and high frequency screen.
As a result, the process modification could bring 17.6% increase in the feed capacity and 27% in the industrial productivity.The productivity raise is higher than the feed increase, as the desliming and flotation selectivity are improved solids percentage (55%), thus enhancing the inner classification and the transport in the mill.Desliming and flotation test in pilot plant and bench scale confirmed the process feasibility and the benefits of the selective milling for the subsequent operations, as well.
As a result, the productivity could be increased by 27% and the mass recovery is expected to raise 7.8%.
Since these results were obtained from pilot and bench scale trials, they still must be validated in the industrial installation.Moreover, all trials used magnesite as feedstock and the benefit for other ores needs to be tested.

Figure 2 .
Figure 2. Chemical composition per particle size on milling process.

Table 2 .
Primary milling modeling parameters

Table 3 .
Comparison between current and projected circuits

Table 4 .
Secondary milling modeling parameters

Table 5 .
Secondary milling simulation compared to pilot plant results

Table 6 .
Primary milling desliming and flotation results

Table 8 .
Summarized comparison of current and designed processes