AI in Russia Is to Predict Repair Timelines for Uranium Wells
The system helps plan maintenance work on wells while keeping production levels steady.
Scientists at the Seversk Technological Institute of the National Research Nuclear University MEPhI have developed an AI-powered decision-support system for a “Smart Mine” project. The system will be used at mining enterprises within the mining division of Rosatom. Previously, geotechnologists decided when and how to service wells based on past performance. Now, artificial intelligence can factor in a much wider range of variables.
Testing Prediction Accuracy
The information system, integrated with the Smart Mine platform, is currently undergoing pilot industrial operation at the company Dalur. The solution helps determine how much uranium each well is producing. Wells that begin to show declining output are prioritized for treatment.
The neural network model was trained on data from more than 50,000 repair and restoration operations obtained from digital twins of uranium deposits. The AI was then tested using datasets that were not included in training. Its forecasts closely matched actual results.
The development by the Seversk researchers is expected to help maintain uranium production levels, reduce the workload on personnel, and improve labor productivity.
