Machine Learning Innovations for Improving Mineral Recovery and Processing: A Comprehensive Review*

Abstract

To overcome the limitations of traditional mineral processing and recovery methods, cutting-edge technologies, including Machine learning (ML), emerge as a paradigm shift in this sector, offering predictive insights, data analysis, and real-time monitoring capabilities. The emergence of ML algorithms, such as Artificial Neural Networks (ANN), Support Vector Machines, and others, trigger this paradigm. This review explores real-world examples and case studies to unveil the transformative potential of ML in mineral processing and recovery (exploration, mining, production). This attempt unveils that ML algorithms are extensively utilized in enhanced ore sorting and classification, predictive modeling, real-time process control and fault diagnosis, and automated mineral identification. Among these applications, predictive modeling for process optimization and enhanced ore sorting and classification stand out, with ANN being the most frequently employed algorithm. While challenges persist, such as limited data availability, non-normally distributed and non-linear data, and varying data dimensions and rates, the advantages of employing ML algorithms are undeniable. These advantages include enhanced operational efficiency, waste reduction, increased recovery rates, real-time monitoring, cost-effectiveness, time efficiency, and reduced energy consumption. This article aims to catalyze further research and promote the widespread adoption of ML for more efficient and sustainable mineral processing and recovery practices.