Russian AI Protects Norilsk Nickel’s Conveyor Systems
Ore moves continuously along a metallurgical plant’s conveyor belt. Sometimes it carries debris – pieces of wood, broken metal and other foreign objects. If these reach a crusher, they can damage or destroy it. Metal detectors used in the past can identify metal, but they cannot detect other unwanted materials.
Norilsk Nickel, in partnership with Russian technology company Reksoft, has deployed a video analytics system that identifies foreign objects contaminating the ore stream. Cameras connected to a computer running a neural network scan the belt and detect items that should not be there.
“Smart Eyes” on the Production Line
When the system sees a hazardous object, it sends an alert and stops the conveyor before the debris reaches the crusher. This helps prevent production shutdowns, repairs to expensive equipment and potentially serious accidents. The plant’s “smart eyes” monitor the ore stream, identify unwanted objects and keep them from damaging machinery. The cameras operate in harsh industrial conditions marked by dust, dirt and vibration. Special protective housings were therefore developed for them, with built-in mechanisms that clean the optics automatically.
The system is already operating at the Mayak and Komsomolsky mines, as well as at the Norilsk Concentrator. Its F1 score, a metric that measures classification performance by balancing precision and recall, reached 0.96 to 0.98, with some sites recording a perfect 100%. The system does more than detect foreign objects: it does so with high reliability while avoiding false alarms.
Complex, Yet Reliable Solutions
These developments further demonstrate Russia’s technological capabilities in industrial AI, one of the most promising areas of digital transformation. Heavy industry is emerging as a major customer for sophisticated, reliable solutions.
In 2021, Norilsk Nickel introduced video analytics at its Kola Mining and Metallurgical Company facilities to monitor the use of personal protective equipment (PPE). If an employee failed to wear a hard hat or safety vest, the system immediately flagged the violation. That same year, the company completed a major project to build a 3D model of the Kola operation using laser scanning and drones. Norilsk Nickel subsequently began developing virtual replicas not only of buildings and structures but also of complex physical and chemical processes. Operators could see on a monitor what was happening inside industrial equipment.
Norilsk Nickel and Reksoft also developed a flotation control system. Machine-learning algorithms and a neural network analyze camera footage of the separation of valuable metals from waste rock. The system recommends process adjustments and can even make them automatically. Another application weighs bulk materials on conveyor belts without physical contact: software calculates the volume of material moving along the belt. As the IT Russia media platform recently reported, a system for monitoring converter tapping was introduced at a metallurgical facility in Norilsk. Norilsk Nickel has also been successfully using video analysis to monitor water turbidity after ore processing for several years, significantly reducing copper concentrate losses.
These digital technologies help keep production of metals used in energy, construction, transportation and electronics stable while reducing the risk of accidents at hazardous industrial facilities. Just as importantly, Russian companies are increasingly developing their own industrial software.
Following the Flow
Foreign objects entering industrial process streams are a challenge for mining operations around the world. The Russian technology is not tied to a specific raw material and can be adapted to different types of conveyor systems. In Russia, it could find applications at other mining and processing plants, in metallurgy, in the coal industry and at raw-material processing facilities. Reksoft is already considering the system as a universal platform that could also be used in petrochemicals and other continuous-process industries.
The next likely step is to move from detecting problems to predicting them. Future systems will do more than identify foreign objects entering a process stream: they will warn operators in advance about potential production disruptions.
