An integrated intelligent framework for maximising SAG mill throughput - Incorporating expert knowledge, machine learning and evolutionary algorithms for parameter optimisation

Aug 5, 2026

In mineral processing plants, grinding consumes about half of total processing costs, making semi-autogenous grinding (SAG) mills central to operational efficiency. Maximising SAG mill throughput has major financial impact, yet optimal parameter settings for highest throughput remain underexplored in prior research. This study presents an intelligent framework combining expert knowledge, machine learning and evolutionary algorithms. It uses an industrial dataset with 36,743 records, selects and refines features, and evaluates 17 predictive models with outlier detection and feature selection. CatBoost proves the most accurate ensemble predictor, while differential evolution delivers robust optimisation under input constraints.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Introduction

Grinding and energy-intensive cost context

SAG mills and throughput as key performance metric

Complex nonlinear relationships affecting throughput

Proposed intelligent framework

Expert knowledge and feature selection

Machine learning model evaluation

Outlier detection and refined predictive objective

Evolutionary optimisation

Differential evolution as optimisation algorithm

Constraint handling for input parameters

Key findings

Minerals Engineering 212 (2024) 108733 Contents lists available at ScienceDirect Minerals Engineering journal [homepage:](homepage: www.elsevier.com/locate/mineng)[ www.elsevier.com/locate/mineng](homepage: www.elsevier.com/locate/mineng) | An integrated intelligent framework for maximising SAG mill throughput: Incorporating expert knowledge, machine learning and evolutionary algorithms for parameter optimisation
Zahra Ghasemi a, *, Mehdi Neshat a, g, Chris Aldrich b, John Karageorgos c, Max Zanind, e, Frank Neumann f, Lei Chena
a School of Electrical and Mechanical Engineering, The University of Adelaide, North Terrace, Adelaide, SA 5005, Australia b Western Australian School of Mines, Curtin University, Perth, Western Australia 6845, Australia
c Manta Controls Pty Ltd, 1 Sharon Pl, Grange, SA 5022, Australia
d School of Chemical Engineering, The University of Adelaide, North Terrace, Adelaide, SA 5005, Australia
e School of Civil, Environmental and Mining Engineering, The University of South Australia, Adelaide, SA 5095, Australia f School of Computer and Mathematical Sciences, The University of Adelaide, North Terrace, Adelaide, SA 5005, Australia g Faculty of Engineering and Information Technology, The University of Technology Sydney, Ultimo, NSW 2007, Australia | | | | --- | --- | --- | | A R T I C L E I N F O | | A B S T R A C T | | Keywords:
Semi-autogenous grinding (SAG) Throughput
Machine learning (ML) Ensemble models
Meta-heuristic algorithm Evolutionary algorithm (EA) | | In mineral processing plants, grinding is a crucial step, accounting for approximately 50% of the total mineral processing costs. Semi-autogenous grinding (SAG) mills are extensively employed in the grinding circuit of mineral processing plants. Maximising SAG mill throughput is of significant importance considering its profound financial outcomes. However, the optimum process parameter setting aimed at achieving maximum mill throughput remains an uninvestigated domain in prior research. This study introduces an intelligent framework leveraging expert knowledge, machine learning techniques and evolutionary algorithms to address this research need. In this study, an extensive industrial dataset comprising 36,743 records is utilised and relevant features are selected based on the insights of industry experts. Following the removal of erroneous data, an evaluation of 17 machine learning models is undertaken to identify the most accurate predictive model. To improve the performance of the model, feature selection and outlier detection are executed. The resultant optimal model, trained with refined features, serves as the objective function within three distinct evolutionary algorithms. These algorithms are employed to identify parameter configurations that maximise SAG mill throughput while adhering to the working limits of input parameters as constraints. Notably, analysis revealed that CatBoost, as an ensemble model, stands out as the most accurate predictor. Furthermore, differential evolution emerges as the preferred optimisation algorithm, exhibiting superior performance in both achieving the highest mill throughput predictions and ensuring robustness in predictions, surpassing alternative methods. | 1. Introduction The process of separating valuable minerals from waste materials in mineral processing plants involves a series of intricate procedures. These procedures can be categorised generally into crushing, grinding, and concentration. Grinding is one of the most important procedures, where unconnected media such as balls, rods, or pebbles are utilised for particle size reduction. This process is usually performed wet to produce slurry for the concentration step. This operation is the most energyintensive stage, constituting around 50 percent of all mineral processing costs (Wills and Napier-Munn, 2006). Mineral processing plants commonly employ SAG mills for size reduction within their grinding circ

FAQs

Why is maximising SAG mill throughput important in mineral processing plants?

SAG grinding is a crucial, energy-intensive step that accounts for around 50% of total processing costs. Higher throughput directly improves the financial outcomes of the plant.

What components does the proposed framework combine to optimise SAG mill throughput?

The framework integrates expert-knowledge-guided feature selection, machine learning predictive modelling, and evolutionary algorithms for parameter optimisation.

Which machine learning and optimisation methods performed best in the study?

CatBoost was identified as the most accurate predictive ensemble model. Differential evolution produced the strongest and most robust parameter optimisation results while respecting input parameter constraints.

Aug 5, 2026

Upload by Seraphina