HEAP LEACH SOLUTION TRANSPORT MODELLING FOR IMPROVED PROCESS CONTROL AND PRODUCTION FORECASTING

By

Oliver Kloiber-Deane

Element Process, Australia

Presenter and Corresponding Author

Oliver Kloiber-Deane

oliver@elementprocess.com.au

ABSTRACT A dynamic bulk solution transport model has been developed by Element Process and coupled with a leaching model to predict, over time, solution flows and grades in heaps, ponds, solvent extraction and the neutralisation plant of an overseas copper heap leach project. The model takes into account the ore stacking schedule, ore leach rates, irrigation rates, rainfall and evaporation, changes to on-flow solution and off-flow destination, and chemistry in solvent extraction and neutralisation to provide a comprehensive project-level process model tracking aqueous copper, iron, zinc and free acid. The model can be operated by project staff, and simulation results are used for medium- and long-term planning and process control. Project mineralogy is primary and secondary copper sulphides hosted in massive pyrite and marcasite. Operators face challenging leach chemistry and extensive solution flow lag due to the topographically-constrained layout of leach pads. In medium-term planning, pregnant leach solution grade predictions by the model feed back into irrigation scheduling and give advanced warning of problems to enable remedial action. In long-term planning, the model predicts the accumulation of iron, free acid and zinc in recirculating solutions. These results inform the implementation of engineering solutions to prevent problems and add value, and allows process change options to be explored in theory before pilot plants are built to prove the feasibility of any change. The model’s use in short-term planning continues to be explored. Planning in heap leaching is always challenging due to its inherent semi-batch nature, significant process response lag, and exposure to the environment. As such, potential exists for similar solution models to improve production and compliance outcomes at other sites by providing a practical, site-operated tool for forecasting solution volumes and grades across the operation. This paper details the development of the bulk solution transport model and its application in a commercial setting. Keywords: heap leaching, heap leach modelling, dynamic process simulation, residence time distribution, solution transport modelling

INTRODUCTION The new owners of an overseas copper sulphide heap leach project engaged Element Process to develop a dynamic model of the process, with the model to be maintained as a live ‘digital twin’ to enable the prediction of production rates and process response to changing inputs over time. Project mineralogy is primary and secondary copper sulphide hosted in massive pyrite and marcasite. A particular challenge faced by the new owners is the recent acceleration of excess acid and iron release to solution which now exceeds neutralisation plant (NP) capacity. Solution total dissolved solids (TDS) is high and is affecting copper solubility during the dry season, while high pregnant leach solution (PLS) free acid concentrations are reducing solvent extraction (SX) efficiency. Site topography is such that limited flat ground is available for new leach pads which has led to the development of very elongated, multi-lift heaps with no inter-lift liners that add significant lag into the off-flow response to on-flow irrigation changes, further complicating solution grade, volume and flow management. The project digital twin is expected to find specific application in predicting leaching rates; solution compositions, flows and volumes; and copper production rates under various scenarios, and will allow process change options including removal of spent ore and increased neutralisation capacity to be trialled.

MODEL ARCHITECTURE The model combines two existing software applications, Microsoft Excel and ITHACA®, into a single customised tool. ITHACA is a dynamic chemical process simulator developed and marketed by Element Process. The Excel component handles user inputs and leaching calculations while ITHACA simulates aqueous species, i.e. solution, flows throughout the flowsheet. As a result, the Excel component of the model is referred to as the leach model and the ITHACA component is referred to as the solution model. During simulation, data is automatically conveyed between the two components as required. ITHACA solves its flowsheet mass balance for every timestep to within a user specified accuracy, and the size of the timestep is automatically maximised as long as this accuracy is maintained. The leach model, however, operates on a fixed time step size of either one month, one week or one day, as selected by the user. One leach model timestep is referred to as a leach period. The model is operated via the Excel component. The following figure illustrates key aspects of model architecture:

Figure 1: Model Conceptual Structure

MS Excel

ITHACA

Input DataE.g. general:- leach curves- ore density- site rainfallE.g. for each time period:- stacked tonnes by parcel- stacked grade by parcel- irrigated % by pad- irrigation rate by pad- irrigation type by pad- off-flow destination by pad- pad temperature

Input Tables

Code- VBA- Object-oriented

Reports

Solution ModelCalculates through time:- HL solution inventory and grades- HL off-flows and grades- Ponds inventory and grades- Other major flows and grades- SX extraction based on PLS FA- NP performance

Input Data for Fluid Process ModelE.g.:- Aggregate Cu, Fe, FA leach rates- Aggregate volumes under irrigation- Aggregate irrigation rates- Rainfall and evaporation rates

Output DataE.g. for Cu, Fe and FA in each time period:- Tonnes produced by each ore parcel- Liquid inventory- Major stream flows and grades- Tonnes extracted (Cu) and produced (FA) by SX- Tonnes removed by NP

LEACH MODEL Much of the ore movement and leaching logic as it pertained to copper stacking and leaching had been developed by the client in an Excel/VBA calculator. Additional logic was added collaboratively to accommodate recent changes to leaching behaviour, and the leach model tool was then upgraded by Element Process to predict iron, zinc and acid leaching and generation rates; to interact with the solution model; and for improved user input and output reporting functionality. For each leach period during simulation the leach model reads user-input ore-stacking and irrigation schedules to determine copper, iron and zinc units available for leaching. Tonnes leached are then calculated from leach curves and the solution properties fed back from the solution model. Free acid generation is calculated from iron leached based on a calibrated relationship.

SOLUTION MODEL DEVELOPMENT The solution model tracks flows and accumulations of water, copper, iron, zinc and free acid through heaps, ponds, solvent extraction and neutralisation. Copper, iron, zinc and acid enter solution as it flows through the heaps. Electrolyte bleed acid make-up is added to SX and SX copper exchange acid enters raffinate solution. Copper exits the model predominantly via solvent extraction with some losses in neutralisation, and other species exit through neutralisation and any bulk solution removal events. Rain capture and evaporative water loss are modelled and Project solution management procedures for controlling pond levels through wet and dry seasons are implemented.

Heap Leach Flowsheet Several simplifications have been made to the flowsheet in order to maximise simulation speed. Some ponds with similar functions have been grouped together, and for each leach period pads that flow off to the same destination – Raffinate Pond, Intermediate Leach Solution (ILS) Pond or PLS Pond – are grouped together into a single model heap for that period. Consequently, the solution model contains three heap leach models, corresponding to the three possible on-flow/off-flow combinations of RAF-ILS, ILS-ILS and ILS-PLS, as illustrated in the following simplified solution model block flow diagram:

Figure 2: Solution model block flow diagram

Cu

Residue

Heap Heap Heap

SX

NP

Residence Time Distribution Solution flows through heaps via the voids between lumps and particles of stacked ore. Voids in stacked ore come in two types: relatively large voids between rocks and agglomerates through which liquid can flow freely under the influence of gravity, and small voids, or pores, between particles and within rocks and agglomerates in which, once filled, liquid remains static due to surface tension overcoming gravity. These types are referred to as free voids and pore voids, respectively. The residence time of any given liquid element in the Project heaps can range up to several weeks, depending on the degree of solution hold-up, plug flow, short-circuiting and back-mixing it encounters. The likelihood of any given solution element encountering these phenomena is characterised by the residence time distribution (RTD) of the heap.1 The heap RTD is important to model because of the significant delay and smoothing added by a heap to any changes in solution properties, which in turn affects the grade and volume of PLS, throughput and extraction in SX, and neutralisation plant performance at any given time. During normal operation, pore voids are saturated with solution while free voids are partially saturated2 and the degree of saturation of free voids determines the rate of liquid percolation through the heap. As the rate of on-flow increases, free void saturation (heap liquid hold-up) increases and as a result the rate of off-flow increases until it matches the new on-flow rate. The shape of this saturation-flow curve is unique to each ore and crush size, and is frequently quantified during early testwork for heap leach projects. Off-flow data was found for one of the Project heaps where the only change was a step-change in on-flow solution grade, and this was used for calibration of the saturation-flow model. The selected heap consisted of one lift of ore at that time. The heap off-flow solution grade change profile for this period indicated a minor degree of plug flow, evidenced by the minor onset delay in the change of off-flow grade, so as an initial trial the model lift was constructed from two process volume elements connected in series. In general, the more volumes in series the greater the degree of plug flow exhibited. The drainage layer was expected to exhibit a significant degree of plug flow and so was initially modelled with a large number of volume elements in series. However, trials demonstrated little change in overall RTD until the number of drainage model volume elements was reduced below three and so this number of volume elements was selected for the heap model. The saturation-flow curve generated during historical laboratory testwork on Project ore was taken as the starting point and tuned until the model off-flow grade response matched site data. The saturation-flow relationship ultimately determined was:

f = 200 × s5

Where:

f = solution flux (L/m2/h)

s = free void saturation (v/v)

The following figure compares original testwork and fitted-model saturation-flow curves, illustrating the significant reduction in the permeability of real, long-term leached ore:

Figure 3: Model versus actual off-flow response

The resulting fit of modelled off-flow grade behaviour relative to site data is illustrated below:

Figure 4: Model versus actual off-flow response

A total of four lift models was included in each model heap. Irrigation and Rain Capture Irrigation solution is applied to lift 1, 2, 3 or 4 depending on which is the top lift in any given leach period. The rate of irrigation is the aggregate of rate for each pad assigned to that model, as specified by user input irrigation flux and percent irrigation schedules. Rain input joins irrigation input for each heap model. For each leach period the volume of rain captured by the model heap calculated from site rainfall and aggregated user input pad rain capture areas. Evaporative Loss The two main drivers of evaporative water loss at a heap leach operation using wobbler-type sprinklers and forced aeration are expected to be evaporation from irrigated solution droplets as they travel through air between the sprinkler and the heap, and evaporation to warm air that is forced through the hot, wetted heap interior.3 In general, the main drivers of evaporation of water from liquid bodies exposed to the environment are solar irradiance, wind speed and relative humidity. Since relative humidity at the Project is almost constant year-around, and in the case of sprinkler droplets wind speed is given by the essentially constant speed of liquid ejection from the sprinkler head, the only remaining driver of variation in evaporative loss from liquid droplets over a year should be solar irradiance. Due to the tortuous path of air forced through the wetted heaps it was expected that complete or almost complete air saturation would be achieved. At the average operating temperatures of 55°C to 65°C seen in Project heaps the saturated loading of water in air could be expected to range from 0.11 to 0.20 kg/m3 dry air. The following relationship summarises the evaporative loss calculation:

𝑉 = 𝐼 × 𝑎 𝑆 + 𝑏 𝐴 Where in any given period of time:

V = water evaporated (t)

I = total solution irrigated (m3)

S = solar irradiance (kWh/m2/d)

A = total air forced through the heaps (m3)

a = solar irradiance correlation coefficient

b = forced aeration correlation coefficient

Water is removed from the top lift volume element of each model heap at the required rate in a given leach period. Leaching Leached species are added to each lift volume element that is active in a given leach period, at a rate calculated by the leach model. Each volume element of the model lift is fitted with an input stream for transporting leached species (Cu, Fe, Zn and H2SO4) into solution. The following figure illustrates a simplified version of the model heap, showing only one lift:

Figure 5: Schematic of the model heap

Solvent Extraction

The solvent extraction plant was modelled as a single volume corresponding to the volume of aqueous in the mixer-settlers. The instantaneous copper extraction extent is calculated as a function of PLS free acid concentration according to a relationship determined from site data:

𝐸𝐶𝑢 = 1 −1

[1 + 𝑒−0.074(𝑐−44)]

Where:

ECu = fractional copper extraction

c = PLS sulphuric acid concentration (g/L)

The following figure plots this relationship over plant data:

Irrigation

Evaporative Loss

Rain Capture

Lift

Model

Volume Element

Drainage

Model

Off-Flow

Leaching

Figure 6: Copper extraction versus PLS acid concentration – actual and modelled

Copper exchange acid is added to the solvent extraction volume element in a 1-to-1 molar proportion to the rate of copper extraction, according to the following extraction equation:

CuSO4 + 2R-H Cu-R2 + H2SO4

Neutralisation Plant

The neutralisation plant is modelled as a single volume corresponding to the combined volume of the NP circuit, including staging pond. For each leach period a user-specified split of SX raffinate is sent to NP, where user-specified fractions of acid, iron, copper and zinc are removed from the feed stream. Water is added to NP as a ratio of metal removal, corresponding to the water added in limestone slurry plus additional inputs from filter cloth washing, gland water, etc. Water is removed from NP as filter cake moisture, calculated from user-input values for limestone utilisation and cake moisture fraction, and metal removal rates calculated by the solution model.

MODEL CALIBRATION A dataset of validated monthly plant data covering 2019 was compiled against which to calibrate and validate solution model output. Most model inputs (for example stacked ore tonnes, stacked ore copper grade, irrigation flux and percent irrigated) had a high degree of certainty and were therefore fixed. A number of parameters such as detailed solution management procedures and ILS-irrigated pad off-flow destinations were uncertain due to lack of records and as such had to be inferred during the early stages of model calibration. Finally, the following factors were iteratively adjusted within limits until all simulation outputs aligned with actual data to an acceptable degree:

solution saturation leach reduction factor (LRF)

evaporative loss correlation coefficients (a, b)

pyrite oxidation free acid production factor (APF)

SX plant utilisation (SXU)

The following table summarises calibrated factor values:

Table 1: Calibrated correlation coefficient values

The following figures illustrate calibrated solution model outputs against respective actual values:

Figure 7: Actual and simulated PLS pond copper grades

Figure 8: Actual and simulated raffinate pond copper grades

Factor Units Value Note

LRF (%) 1.35x10-4

[PLS Fe g/L]2 -

4.74x10-3

[PLS Fe g/L]

Fit to 2019 values for

36<[PLS Fe g/L]<106

a (t.d.m-1.kWh

-1) 6.6x10

-3Single value

b (t.m-3) 1.5x10

-4Single value

APF (%) 88.3 Average of 2019 values

SXU (%) 91.3 Average of 2019 values

Figure 9: Actual and simulated PLS pond iron grades

Figure 10: Actual and simulated raffinate pond iron grades

Figure 11: Actual and simulated PLS pond zinc grades

Figure 12: Actual and simulated raffinate pond zinc grades

Figure 13: Actual and simulated total ponds solution volume

EXAMPLE FORECAST RESULTS A combined leach-solution simulation of Project operations from January 2020 through September 2023 was run. Due to the sensitive nature of the results only a small sample is reproduced here and some values are redacted. The simulation setup reflected a nominal base case where no drastic changes to the process were introduced except for the installation and ramp-up of an expanded neutralisation plant. The ore stacking schedule was supplied by the client.

Copper Leach Performance Iron build-up in solution continues until the NP expansion ramp-up is complete, after which copper leaching and PLS copper grade recover quickly:

Figure 14: PLS iron and period leached copper

Figure 15: PLS pond copper grade

Water Balance From 2020 onwards the water balance is net negative, meaning that make-up water is required. This is due to the continued application of high rates of aeration, heap area utilisation, and irrigation flux:

Figure 16: Total ponds solution volume

The following figure plots the fractional contribution of each input water source to the total, showing that while rain capture remains significant, water input to an expanded neutralisation plant (via water in limestone slurry, filter cloth washing, etc) is significant.

Figure 17: Water inputs breakdown

The main contributor to water loss is evaporation, as illustrated below, however water exiting with neutralisation filter cake is a significant and growing contributor.

Figure 18: Water outputs breakdown

CONCLUSIONS A leach and solution flow simulation tool has been developed for use in predicting process performance under time-varying conditions to assist a client in planning for the future at an overseas copper sulphide heap leach operation. The model tracks water, copper, iron, zinc and free acid and has been calibrated against one year of historical data (2019). A simulation of years 2020 through September 2023 was run assuming implementation of a neutralisation plant expansion to reduce solution TDS, resulting in a reduction of solution iron and rebounding of copper leaching in 2021 to design rates. The simulation tool continues to be developed to accommodate planning needs, including future heap flushing and closure operations.

ACKNOWLEDGEMENTS We are grateful to our client for allowing this project to be presented and for the feedback received.

REFERENCES 1. Bouffard, S.C., West-Sells, P.G. Hydrodynamic behavior of heap leach piles: Influence of testing scale

and material properties. Hydrometallurgy Volume 98 Issues 1-2, 136-142 (2009)

2. Bouffard, S.C., Dixon, D.G. Investigative study into the hydrodynamics of heap leaching processes. Metallurgical and Material Transactions B 32, 763–776 (2001)

3. Kappes, D. W. Precious Metal Heap Leach Design and Practice, Mineral processing plant design, practice, and control, 1606-1630 (2002)