Mineralogy informed flowsheet modelling

Project Highlights

• Model the impact of mineralogy on circuit performance
• Predict liberation distributions from breakage behaviour
• Directly coupled comminution and separation predictions

Overview

In minerals processing the mineralogy and texture of an ore plays a key role in the link between comminution and subsequent separation. In current circuit simulation software this link is usually lost, with comminution simulations usually only predicting the size distribution of the resultant particles, with important considerations for the subsequent separation, such as volumetric and surface liberation not directly being predicted.

In this work we propose to develop a methodology for using SEM-EDX and/or microCT images of unbroken ore particles as an input to the flowsheet modelling. These will then be virtually broken, with mass balances based on populations of these virtual particles. In order to do this efficiently, a balance will need to be struck between the number of particles being tracked and the accuracy of the resultant distributions. As the size of the particles can change by orders of magnitude over a comminution circuit, which particles are being tracked will need to automatically evolve over the flowsheet. Other than the unit specific models, the key algorithms that will need to be developed are those for virtually breaking particles and those for deciding which are key particle types to include in the mass balance and which to discard, as well as how to ensure that the macroscopic mass balances remain accurate as particles are discarded. The discarding of particles from the simulations will need to be based on a number factors, including their importance to mass balance for the various minerals and particle size classes, as well as their uniqueness. The ultimate aim of the simulator will be to allow comminution and separation to be jointly optimised based on the actual ore textures and mineralogy. It will also allow circuits where separation and comminution are directly integrated with one another, such as flash flotation cells in the milling circuit or regrinding within the flotation circuit, to be far more accurately simulated than is currently possible.

Key research questions

This project aims to answer the following key research questions:

1. How does mineralogy, comminution and separation interact in determining a circuit performance and efficiency
2. Can statistical breakage algorithms be developed that allow the resultant distributions of particle properties, such as liberation, to be accurately modelled
3. Can algorithms be developed that automatically balance the accuracy of the resultant simulation with the computation cost of simulating a large number of particle classes

Methodology

The PI has already developed a framework into which these algorithms can be inserted. The key task of this project will therefore be to develop the algorithms for virtually breaking microCT or SEM-EDX images of ores. This will then need to be coupled with additional algorithms for determining the representative particles to carry forward at each point in the simulation and how to redistribute those that are rejected. To test these algorithms some lab based experiments will need to be carried out in which a piece of ore is milled and separated, with both the intact ore and the resultant particles at each stage of the separation imaged. This will also require the adaption and improvement of our existed imaging algorithms to allow comparisons to be made between the actual and virtual particle distributions.

Possible timeline

Year 1: Develop initial algorithms together with simple unit models. Obtain initial 3D and 2D images of ore particles to use in algorithm development

Year 2: Obtain images of processed ores to compare with the virtual particles predicted by the simulator with the actual ore particles produced. This will be coupled to algorithm improvement and the development of initial unit models. These will be based on literature models, but adapted to this new framework.

Year 3: As well as continued improvement of the simulator, in this year it will be used to investigate different comminution and separation configurations and their impact on the overall circuit performance.

Training and skills

TARGET researchers will participate in a minimum of 40 days training over the 3.5 years of study composed of:

• an annual one-week workshop dedicated to their year group, and tailored to that cohort’s needs in terms of skills development – for the first three years of their study;
• an annual all-TARGET workshop with cross-year interactions, advanced training and opportunities to specialise in particular areas – all years of study;
• a number of one-day workshops;
• additional online events and in-person workshops attached to relevant conferences.

Partners and collaboration (including CASE):

Our main CASE partner will be FLSmidth. They are an equipment manufacturer who also designs and implements minerals processing circuits for their clients. They will therefore have a strong interest in the resultant flowsheet simulation framework, as well as having case studies that could be used as the basis for the testing of the simulator and the evaluation of different circuit designs.

The NHM, as the second TARGET partner, will provide mineralogical expertise, as well as access to imaging equipment such as SEM EDX.

Requirements

Applicants should ideally have a degree in engineering, with an interest ideally in both minerals processing and mineralogy. They should also have good coding skills in C++ ideally with some experience of parallel computing.

Further details

Further details: Please visit https://www.imperial.ac.uk/people/s.neethling and https://target.le.ac.uk/