AI Advantage: Reducing Downtime at Yakovlevsky Mine
Yakovlevsky Mining and Processing Plant (part of Severstal, located in Belgorod region) has unveiled an in-house computer vision solution to monitor ore passes. Developed jointly by Severstal Digital, Severstal-Infocom, and plant specialists, the system aims to minimize downtime and improve workplace safety.
Minimizing Risk in a Hazardous Industry
Mining is one of the most dangerous industries in the world, making occupational safety a constant priority. At Yakovlevsky, several digital tools are now being tested alongside ore pass monitoring, including remote-controlled shearer loaders, helmet-mounted video systems, and daily checks of ground support.
The program is broad and resource-intensive. Last year, the mine deployed a machine vision system to monitor personnel in storage chambers. Initially, workers paired with AI systems in a supervised learning mode until the models, trained with input from the plant’s health and safety department, reached maturity.
The plant has also implemented SOVA, an AI-based platform for occupational and industrial safety. Currently, 300 cameras track potential violations, with 150 more being installed in 2025 to expand coverage. This technology automatically identifies unsafe practices during routine operations.
Other tools include a remote-control unit for operating shearer loaders from a safe distance with real-time video of blind spots. Workers are also equipped with helmet-mounted cameras that integrate with machine vision models to detect unsafe behavior. Additionally, 30 underground load-haul-dump machines now feature cameras for real-time monitoring of personnel operations.
Ore’s Journey Under AI Supervision
The Yakovlevsky case represents a clear step toward the “digital mine” concept.
Ore moves through vertical shafts: from extraction levels to haulage levels, into rail cars, through bunkers, and then to the surface. Until recently, detecting blockages or overflows at these ore passes—a critical step in ore transport—was done manually by dispatchers. Human error or delays in spotting problems often caused costly downtime, reduced efficiency, and raised operating costs.
The new AI-driven solution developed by Severstal Digital, mine staff, and Severstal-Infocom tackles this risk. Neural networks analyze real-time video feeds from cameras operating in the harsh conditions of underground mining. When anomalies are detected, the system generates photo reports and sends alerts through Severstal’s “Steel Vision” platform, allowing rapid deployment of clearing equipment.
The system’s uniqueness lies in advanced algorithms that compensate for low lighting, camera shifts, and poor visibility—proving the technology’s maturity.
The Numbers Speak
Between January and June 2025, the system helped avoid 19.3 hours of downtime, generating an economic benefit of RUB 2.9 million (about USD 31,000).
Beyond direct financial gains, the project improved operations: dispatchers were freed from routine visual monitoring, allowing them to focus on strategic tasks, which increased both safety and efficiency.
For a large mining enterprise, savings of nearly USD 31,000 in six months from a single module is only the beginning. Scaling such systems to other critical processes—ventilation, power supply, logistics—could generate hundreds of millions of rubles (millions of dollars) in annual savings.
Crucially, ore pass monitoring is not an isolated initiative but part of Yakovlevsky’s broader digital transformation strategy. The system integrates seamlessly with the plant’s existing digital ecosystem.
