Generative AI Comes to U.S. Steel’s Iron Ore Facilities

United States Steel Corporation (U.S. Steel) announced in August that the company is partnering with Google Cloud to develop applications based on generative Artificial Intelligence (AI). In addition to producing steel sheets, the company mines coke and iron ore. The first application, MineMind, will go live during September at U.S. Steel’s largest iron ore facilities (Minntac and Keetac) in Minnesota.

Iron ore mines are often located in remote areas and mining firms struggle to attract new recruits to work at these facilities. These issues, among others, have contributed to mining firms heavily investing in innovative technologies to support their operations, such as using drones to capture images of open-cast mines and autonomous haulage trucks to transport extracted minerals (see ABI Research’s Digital Transformation in the Mining Industry presentation (PT-2748)). MineMind continues this trend and will focus on improving the maintenance processes that support the haulage trucks operating at the sites.

The initial focus will be supporting and improving the productivity of the maintenance teams responsible for the 60 haulage trucks operating at the sites. A member of the maintenance team might have a question related to repairing the suspension system, for example. Rather than having to refer to a lengthy technical manual, the employee outlines the issues in MineMind and then receives a summary of the applicable repair instructions (including diagrams), as well as the ability to order spare parts.

MineMind is underpinned by the Document AI and Vertex AI solutions from Google Cloud. Document AI extracts and classifies data via a suite of pretrained models, and stores the information. Vertex AI uses a Machine Learning (ML) workflow that includes data preparation, model training, model evaluation and iteration, model deployment and serving, and model monitoring.

U.S. Steel anticipates that MineMind will reduce the time needed to assess a problem and issue a work order by 20%. However, to be successful, the solution must be trusted by the users. The results provided by MineMind will also include references and the source materials that inform the instructions. Furthermore, a validity score will be included along with the results.

The companies are now working to expand the partnership to improve other areas of U.S. Steel’s operations with generative AI applications.

ABI Research’s Generative AI Use Cases in Manufacturing presentation (PT-2763) outlines 25 use cases across a manufacturer or industrial firm’s design, engineering, production, and operational domains. The current iteration of the MineMind solution addresses several use cases, including frontline worker documentation, stock inventory, and purchasing.

Future use cases that U.S. Steel should consider include predictive maintenance and scheduling (bearing in mind the time required to deliver spare parts to remote areas), production root cause analysis, optimizing process sequences and tool paths, and improving worker safety through data analysis. There is enthusiasm about implementing AI initiatives across U.S. Steel’s facilities with the President and Chief Executive Officer (CEO) of U.S. Steel, David Burritt, stating the following during the 2Q 23 Earning Call Transcript on July 28: “At our mines in Minnesota, we are leveraging advanced analytic models for operator recommendations to increase productivity at our concentrator. This has achieved nearly $3 million of value. This is only the beginning of our digital and AI journey.”

Boardroom enthusiasm can provide the thrust for AI and digital transformation initiatives. But in the case of U.S. Steel, enthusiasm might be tempered. The company stated on August 13 that it conformed “Receipt of Unsolicited Proposals from Cleveland-Cliffs and Multiple Other Parties.” U.S. Steel further stated: “The Company previously disclosed it has commenced a formal review process, with the assistance of outside financial and legal advisors, to evaluate strategic alternatives for the Company after receiving multiple unsolicited proposals that ranged from the acquisition of certain production assets to consideration for the whole Company.” Uncertainty surrounding a firm’s future shape and direction will test commitment to the digital and AI journey.