European project BLOOM has completed its first 18 months, with a containerised pilot taking shape to bring near-real-time mineral liberation analysis and AI-based process control into grinding and flotation circuits. BLOOM said the technology is intended to help mines adjust processing to the mineralogy of incoming ore rather than rely only on slower laboratory testing.
The Horizon Europe Innovation Action is developing a modular online mineral-analysis system for mines. It has an EU contribution of about 7.49 million euros and is scheduled to run from December 2024 to November 2028. Its consortium includes European partners and partners from Canada, Ukraine and Brazil.
During the first reporting period, consortium members completed mineral characterisation and laboratory work on representative copper-zinc and tantalum-niobium ores. The tests are examining how minerals respond during grinding and flotation, while supporting development of the mineralogical database, software architecture and communications interfaces needed for later industrial deployment.
The system’s central concept is online Mineral Liberation Analysis supported by machine learning and artificial intelligence. Mineral liberation describes whether valuable mineral grains have been separated sufficiently from surrounding gangue for downstream recovery. Insufficient grinding can reduce metal recovery, while excessive grinding wastes energy and can create fine particles that are harder to float. Better liberation data could improve recovery and energy efficiency.
BLOOM has developed the engineering design of a containerised pilot plant combining sensing, automation and intelligent process control. Its modular format is intended to make realistic testing easier before equipment is installed at commercial mining operations. The consortium is working with mining companies and projects in Europe, Canada and Ukraine, and has access to industrial samples and operating expertise for future validation.
The technology is relevant to Europe’s raw-material ambitions, which increasingly include lower-grade and more complex deposits. Such ores can be harder to develop economically as processing costs, recovery losses and energy intensity rise when ore quality falls. More precise adjustments to operating conditions could help reduce unnecessary grinding or reagent use.
The next phase will assess whether the system can provide sufficiently fast, robust and actionable measurements under plant conditions. If it proves capable, continuous measurement of ore characteristics could allow plants to respond automatically and move process control toward a closed-loop model, with AI integrated into the core process-control architecture.
