Mining / Iron ore
Vale: Real-time AI adjustment of process variables at an iron ore processing plant (Conceição 2 model plant)
- Company
- Vale
- Country
- Brazil
- Adoption stage
- In operation
- Source published
- Date basis
- The date the source was published. It can differ from the date adoption started.
- How the source was checked
- Read the full source text
The work problem
According to a press release Vale issued on June 10, 2026, Vale modernized its Conceição 2 plant in Itabira, Brazil, to eliminate process bottlenecks, reduce people's exposure to hazardous activities and raise quality and efficiency. The plant produced 9 million tonnes in 2024.
Technology and data
Over 1.5 years, Vale applied 51 solutions to eliminate bottlenecks and improve operations. It installed more than 100 monitoring cameras across the plant complex and automated about 7,300 instruments, including new measurement devices and sensors, and uses data intelligence to control and optimize more than 400 variables across all stages of ore processing. Vale's Vice President of Operations explained that an artificial intelligence layer supervises hundreds of variables and adjusts the production process in real time according to the characteristics of the ore and the product. Online ore grade analysis during processing allows immediate correction of the processing route, and operators monitor equipment remotely from control rooms and phones, going into the field only when needed. ABB acted as technology integrator, and 122 operators, technicians and leaders received more than 2,800 hours of training.
Results
According to Vale, productivity at the plant rose 25% in less than two years of the pilot, reaching planned capacity, and this year the plant is capable of producing 11.2 million tonnes. The share of direct reduction pellet feed used in decarbonized steelmaking increased by 40%, and the average iron content in processing waste fell 26% in 2026. The company said 100% of the plant's critical operational decisions are supported by expert systems.
Limits and open questions
The source does not break down the productivity gain into the shares from automation, equipment improvements and the AI layer.
Sources
Compiled from public sources. These are not results from ATF Works customers.