Insights from simulation of CIL/CIP circuit Part 1 – model development

Leaching

The representation of values in the ore as consisting of a fast leaching component and a slow leaching component is similar to the way in which flotation has been analysed and modelled (Cameron et al., 1971; Imaizumi and Inoue, 1965). While it is not claimed that this is a rigorous physical representation of the leaching properties of an ore, nonetheless it provides a convenient empirical way of representing the leaching behaviour and so long as it matches the leach profile over the range of leaching times under consideration then valid simulations may be obtained (Fig. 2). With some ores an additional parameter could be justified to represent a fraction of the values which cannot be leached at all. Such a situation was not considered in this work.

OPEN IN VIEWER

Figure 2

Leaching model for gold fitted with two components (typical data)

In a batch leaching situation (1) where F_rem is the fraction of silver or gold remaining unleached after leaching time T, f is the fraction of the gold or silver which is fast leaching and α and β are the leaching rates of the fast and slow components (reciprocal time).

It follows from this that if tanks are fully mixed, then in a continuous flow situation, where f_in is the fraction of fast leaching material entering, the fraction of values unleached in any tank is (2) The leaching rates and fast leaching fraction are primary inputs to the leaching model for simulation. They can be estimated from the leaching profile down the bank. Leaching rates, and hence the parameters for the leaching rate equations, would be expected to be a function of the ore type, the grind and the general chemical environment, particularly the pH, available cyanide and whether oxygen is used to enhance the leaching process.

Unexpected deviations from a normal leaching profile may be due to lack of cyanide or aeration (or oxygenation) deficiencies. Identification of such deviations may be obscured by short term variations in ore (order of hour to hour) changing the material from which the contents of any tank have been derived.

In order to fit the model to any circuit, data must therefore be obtained by sampling the feed to the circuit and the tailings from each tank. A series of campaigns may be required to obtain sufficient data to accommodate short term changes in parameters due to variations in head grade and ore type. The number of such ‘runs’ needed to obtain reasonable averages may be reduced by sampling each tank in succession with a delay which allows for the residence time of slurry in each tank, thus removing a large part of the dynamic effect of short term changes in ore properties and feedrate. The solids, solution and carbon in all samples are assayed for both gold and silver to give the assay profiles for both components down the bank on each occasion. If it is desired to extrapolate simulations reliably to a more barren tailing, the leaching data may be extended by bottle rolling final tails for an additional period or special plant runs may be performed at reduced feedrate. Adjustment can be made for the difference between continuous flow and mixed conditions in a tank and batch conditions in a bottle.

Carbon adsorption

Ideally the sampling of solution and carbon would be taken immediately prior to the movement of carbon on to the next tank up the train. In this way the carbon will have had maximum opportunity to equilibrate fully to the solution concentration in a tank. In practice because of head grade variations the solution concentration in a tank does not remain steady and timing sampling to correspond to carbon movement is difficult to coordinate. However, if the approach of the carbon to equilibrium loading in any tank after movement occurs at a reasonable rate (e.g. within a few hours), then there may be quite a big window for satisfactory samples to be obtained. While active carbon adsorption rates from pure solutions in the laboratory may occur quite rapidly (order of minutes) in a plant with solids concentration of ∼40%, the activity of the carbon as measured under laboratory conditions has only a secondary influence as adsorption rates are largely controlled by the mixing in the tanks and diffusion through the boundary layer close to the available carbon particles.

As concentrations of the gold and silver as cyanides in solution (up to 40 g t⁻¹) and on the carbon (<3%) are relatively low, the adsorption of gold and silver on active carbon in the model has been represented by the Freundlich equation (Treybal, 1955) (3) where L is the carbon loading, s is the solution concentration of gold or silver and A and N are the adsorption constants. The model assumes that gold and silver adsorb independently and that there is no interaction or conflict at typical loadings. The model also assumes that adsorption and desorption may both occur in the circuit provided the carbon has adequate residence time in each tank. Given carbon residence times in the tanks typically of the order of at least 10 h, this requirement should be met in practice. Hence, the loading on the carbon removed from any tank is considered to be in equilibrium with the solution in the tank regardless of the solution profile. It is especially important for this to be the case with the solution in the tank from which the final loaded carbon is removed.

Mass balance

Values enter the circuit primarily contained in the ground ore fed to the circuit. Water being recycled to the circuit may contain values, especially water returned from a thickener which is intended to recover unused cyanide and values which would otherwise be lost in solution in the tailings. Regenerated carbon may also contain some residual values. The mass balance for the circuit and for any tank may be written as follows (4) where G is the feedrate of ore, h is the head grade of the feed to the entity (circuit or tank), t is the tailings grade, C is the carbon advance rate, L is the carbon loading, p is the per cent solids in the slurry and s is the solution concentration. L_out and s_out are related by the adsorption equation. t is calculated progressively down the bank, the tailings from any tank becoming the head to the next tank. For the circuit overall, h is the initial head and t the tails from the final tank.

These equations were programmed in Excel, solid and solution grades in the tailings from one tank being the input to the next and the carbon output from one tank being the carbon input to the preceding tank. The number and function (leach-only or leach and adsorption) of the tanks was set in any version of the programme and the tank sizes could be input. Additional normal inputs apart from the leaching and adsorption parameters were the feedrate of ore, the per cent solids in the slurry, the carbon advance rate and the residual loadings of the regenerated feed carbon. If the effect of return of water containing values being recycled from a thickener was being explored then a further input was the fraction of water in final tails which was recycled. A direct analytical solution of the complex equation set cannot be obtained readily, so the circuit was generally solved by progressive approximation of the solution concentration of the final tails. Alternatively, the final solution concentration for either gold or silver could be chosen and one of the other operating variables changed to give the selected concentration.

Some mass balance considerations

The overall mass balance provides some preliminary insights into the primary relationships between the main operational variables. A key design and operating consideration is selecting a carbon advance rate. If a particular loading is targeted for say gold, then a corresponding advance rate would be (5) This simple relationship underlines that the fact that the advance rate is the most important aspect of carbon management in circuit performance. The quantity in each tank is only important in ensuring that the carbon has sufficient residence time to come close to being in equilibrium with the solution and that it is in sufficient concentration for requisite quantities to be readily moved from tank to tank to give the desired advance rate.

The fractional recovery R, in the circuit is (6) To a first approximation L_in may be considered to be negligible. If it is assumed that the ore is very fast leaching so that all the values are leached in the first tank, then the maximum possible solution concentration in the first tank will be (7) Substituting from equations (3) and (7), equation (6) above then becomes (8) This assumes that there is sufficient tankage for the values to be adsorbed from solution so that carbon entering the first tank in the train which contained carbon is already well loaded. If this ratio R, is significantly greater than one, then the solution concentration in the first tank must be less in practice than the maximum possible. If the ratio is significantly less than 1 this implies that values are being fed at a rate greater than they can be recovered on the carbon and a solution loss will result. This could mean that the carbon advance rate C, is too low so that leached values cannot be taken out of the circuit at a high enough rate. In the same way if A (the primary constant in the adsorption relationship) is low, a low carbon loading will demand an unduly high value of C in order to avoid solution loss. A low A may be due to inadequate regeneration, degradation or fouling of the carbon.

As N is typically in the range 0·65–0·85, the value of R proves to be relatively insensitive to the head grade h, a doubling producing only a change of ∼20% in R. In contrast, the recovery from solution may be very sensitive to pulp density, an increase in p of 6% from say 40 to 46% increases R by ∼20%.

These general considerations form a background for the simulation studies which embody a more detailed and precise representation of the whole leaching and adsorption circuit.

Model parameters

A standard set of parameters were used in most simulations with the model. The leaching parameters are shown in Table 1 below. These were derived from a range of plant data obtained from surveys of gold and silver in the solids and solution, and carbon loadings from tank to tank. The parameters represent an ore in which the gold leached readily but the silver, in contrast, was more resistant to extraction. The surveys encompassed a range of feedrates, plant configurations with seven or ten tanks in series and operation with and without oxygen addition.

OPEN IN VIEWER

Table 1

Leaching parameters

	No supplementary oxygen		With oxygen
	Gold	Silver	Gold	Silver
Fraction fast leaching f	0·85	0·45	0·88	0·55
Leaching rate fast fraction α/h−1	0·75	0·80	4·00	2·00
Leaching rate slow fraction β/h−1	0·035	0·020	0·030	0·012

The effect of oxygen addition on the parameters is typical of the data examined. There is a relatively small increase in the fraction which is fast leaching, a major increase in the rate at which this fraction leaches and a decrease in the leaching rate of the slow leaching fraction. The leaching profiles generated by these parameters are illustrated in Fig. 3 below. These are for an ore feedrate of 209 t h⁻¹ at 38% solids to a series of tanks of 840 m³ active volume. The residence time in each tank was very close to 2 h.

OPEN IN VIEWER

Figure 3

Standard leach profiles used in models

While the initial leaching is much quicker with oxygen, in one tank with oxygen almost as much leaching occurs as in three tanks without oxygen, surprisingly after 12 tanks (24 h) the final amount leached was found to be little more with oxygen (0·7% additional recovery for gold and 0·5% more for silver). As the number of tanks is reduced the differences with the use of oxygen are increased.

Adsorption parameters used in the standard simulation with the model are given in Table 2 below. The parameters are scaled for expressing the carbon loading and solution concentrations in g t⁻¹. They represent the adsorption of active carbon which is a seasoned inventory, i.e. carbon in a plant which has been in operation for a sufficient period for the carbon to have reached the equilibrium blend from carbon which has been regenerated many times through to fresh make up carbon.

OPEN IN VIEWER

Table 2

Adsorption parameters

	Gold	Silver
A	2600	1900
N	0·8	0·7

The adsorption relationships are shown in Fig. 4 below for a normal range of carbon loadings.

OPEN IN VIEWER

Figure 4

Adsorption relationships

These relationships were developed for ores where there were no important quantities of competing species (other metal cyanides) and show that, in general, gold was more readily adsorbed than silver.

The ‘standard’ ore being processed in the simulations was presumed to have 5 g t⁻¹ gold and 35 g t⁻¹ silver. Feedrates and residence times were generally kept in a range such that recoveries of gold would be >85% and silver 30–70%. Thus feedrates were generally 150–300 t h⁻¹, slurry ∼38% solids and total slurry residence times of about 14–30 h. Carbon advance rates were generally constrained to maintain a carbon loading of at least 2000 g t⁻¹ and were typically in the range of 6–10 t/day.

While the model studies illustrate behaviour characteristic of CIL/CIP plants and generate some broad principles of circuit design and management, decisions in relation to any particular plant will be very dependent on head grades, the actual range of leaching rates for any ore, the adsorption parameters pertaining to a seasoned carbon inventory, metal prices and the cost of any indicated changes. However, in this study, a particular percentage change in silver recovery would currently have only about one eighth of the value of the same percentage change in gold recovery with current gold and silver prices (Au, $1600; Ag, $28).