Researchers Develop Method To Improve Infrastructure Condition Prediction While Complying With Privacy Regulations

Posted: Oct 31, 2025

Research at the University of Utah is providing a scalable, privacy-preserving solution for improving culvert condition prediction. This method will enable transportation agencies to make more data-driven, proactive maintenance decisions. By leveraging federated learning (FL), a form of machine learning, the Utah Department of Transportation and other state DOTs can collaborate without sharing sensitive data, enhancing predictive accuracy while maintaining compliance with privacy regulations. The implementation of FL can improve infrastructure management, optimize resource allocation, and enhance the overall safety and longevity of transportation networks, potentially saving millions in maintenance costs over time.

Abbas Rashidi, Ph.D.
University of Utah

Data-Driven Inspection Planning for Utah Culverts Using Federated Learning