Luleå research highlights the real challenge for AI in mineral processing

Artificial intelligence can help predict when mining machinery is likely to fail, potentially reducing unplanned downtime, improving equipment availability and lowering maintenance costs. Yet a recent systematic review by researchers at Luleå University of Technology, Sweden, suggests that the mining industry is still some way from realising this potential.

The study, Data-driven prognostics and health management for mining processing machinery, reviewed 74 studies selected from 1,285 research articles. It found that research has focused mainly on fault diagnosis and detection, with much less attention given to predicting remaining useful life and translating AI predictions into practical maintenance decisions. Only around 16% of the reviewed studies used real industrial data, while just 27% explicitly connected their findings to decision-making. The challenge, therefore, is not simply developing better AI models, but making them reliable and useful in operating plants.

These findings are highly pertinent to MEI’s inaugural AI in Mineral Processing '27, which will explore how AI can advance mineral processing through improved monitoring, optimisation and decision support. The Luleå review highlights an important question for the conference: how can promising research be translated into robust, industrially proven solutions that deliver measurable operational benefits?

One of the conference’s keynote speakers, Prof. Saeed Chelgani of Luleå University of Technology, will present Mining the pixels: AI image processing for mineral processing. Although his focus on computer vision is distinct from the machinery-maintenance research, both areas face shared challenges around data quality, model reliability and performance under changing industrial conditions.

The other keynote, Dr. Osvaldo Bascur’s AI, the next giant leap in innovation, may offer an especially complementary perspective. His work on digital twins, integrated plant data and AI-enabled decision-making addresses the broader challenge of connecting information and predictive insights to operational performance. This is precisely the gap highlighted by the Luleå review: moving beyond identifying problems towards using AI to make better decisions.

Together, these perspectives underline a central theme for AI in Mineral Processing '27: the real measure of progress will not be how sophisticated the algorithms become, but how effectively AI is integrated into everyday plant operations.

#AI27
#MEIBlog