The integration of Artificial Intelligence (AI) in mineral processing automation and the mining industry holds transformative potential, enhancing various operational facets. One critical application of AI is in the size reduction of ores, an essential process for achieving uniform mineral distribution. By employing AI to monitor and adjust the size of rocks entering the mill, the process becomes more efficient and effective, leading to improved downstream processing and higher quality outputs.

In the realm of equipment maintenance, AI's role in predictive maintenance is invaluable. Traditional maintenance schedules often lead to unnecessary downtime or catastrophic equipment failures. AI mitigates these issues by continuously monitoring equipment health and performance. By analysing sensor data and identifying patterns indicative of wear and tear, AI can predict when maintenance is genuinely needed, ensuring that assets are taken offline only when necessary. This approach minimizes downtime, reduces maintenance costs, and extends the lifespan of critical machinery.

AI also excels in autonomous decision-making, which significantly reduces the need for human intervention. AI's ability to process and evaluate large datasets rapidly is a gamechanger as it can adjust equipment settings or reroute materials in real-time to optimize efficiency. This capability extends to analysing data from multiple sources to make informed decisions about ore extraction, resource allocation, and energy consumption. For instance, AI can dynamically adjust the operational parameters of crushers and mills to optimize throughput and energy usage, ensuring that the process remains both efficient and cost-effective.

Overall, the adoption of AI in mineral processing and mining stands to revolutionize the industry. AI-driven automation and decision-making enhance operational efficiency, reduce costs, and improve safety by minimizing human exposure to hazardous conditions. Furthermore, AI's advanced data analysis capabilities facilitate the discovery of new resources, ensuring the industry's sustainable growth. By leveraging AI, mining companies can achieve significant competitive advantages, leading to a more innovative and productive future for the sector.

The motivation behind this research stems from the growing distrust and criticism that the Australian mining sector receives across media outlets. Despite technological advancements aimed at enhancing efficiency and safety there is a persistent and notable disconnect in public perceptions and community engagement efforts. This research seeks to explore alternative approaches that can bridge this gap and improve the industry's public image to engineer multiple positive outcomes.

The research problem exposes the challenges that a technology-driven approach brings, in marketing and public relations’ content, in the face of building positive perceptions and community relationships. Findings suggest that traditional mining practices, like manual techniques and local employment, can potentially be effective in driving community engagement and advancing national pride.

The implications of this research are twofold. Firstly, it underscores the importance of diversifying approaches to community engagement beyond technological solutions. Secondly, it suggests that mining companies should prioritize the preservation and promotion of traditional practices in marketing and community outreach efforts to build a better sense of trust to improve public perceptions.

In mineral exploration, accurate lithology classification of drillhole data is paramount for assessing geological features and enhancing the profitability of mining operations. Traditional methods rely heavily on a geologists’ visual inspection and subjective interpretation of core samples, which can be time-consuming and inconsistent. Machine learning algorithms applied in this classification context, have the capability to assist in these challenges by producing rapid first-pass interpretations even when the core samples are missing or damaged (Horrocks, Holden and Wedge, 2015).

The main objective of this study is to apply a variety of machine learning methods to conduct supervised classification of lithology types using multivariate geophysical data that is recorded during drilling operations. A total of 11 drillholes and four distinct lithology types were utilised in the application of each method. The data was submitted to training, cross-validation, and testing in three different scenarios using Decision Tree (DT), Random Forest (RF), Multi-layer Perceptron (MLP), Support Vector Classifier (SVC), XGBoost and LightGBM algorithms. A novel approach to crossvalidation was employed during the optimisation of each machine learning algorithm to preserve the sequential integrity of each drillhole. Balanced accuracy or macro-average recall was the primary evaluation metric for each algorithm.

The analysis revealed that while most algorithms can easily achieve over 90% balanced accuracy within the same drillhole, their performance declined by approximately 30% when applied to independent and previously unseen drillholes. A small increase in performance was observed in the same context when the models were trained with an expanded set of four drillholes. Among the tested algorithms, RF and SVC consistently outperformed the field on average across all testing scenarios. These results imply that access to a more extensive dataset, featuring consistent geological assessments with higher confidence intervals, may significantly improve each algorithms’ capability to classify lithology accurately and rapidly.