2026-09-19T06:20:59Zhttps://api.drum.lib.umd.edu/server/oai/request

oai:drum.lib.umd.edu:1903/340662025-08-09T07:50:00Zcom_1903_2221com_1903_1654com_1903_2col_1903_2753col_1903_3

Zhang, Yunfeng Xu, Jianshu Digital Repository at the University of Maryland University of Maryland (College Park, Md.) Civil Engineering 2025-08-08T11:39:32Z 2025 https://doi.org/10.13016/llji-bwhl http://hdl.handle.net/1903/34066 This dissertation explores the application of an AI-powered digital twin framework to address challenges in civil infrastructure management, which focuses on three applications: slope stability risk monitoring, moving train-induced ground vibration analysis, and pavement strength evaluation. By Leveraging machine learning (ML) and finite element modeling (FEM), the study develops innovative tools for geotechnical and structural condition monitoring and response analysis. Key contributions include the use of neural networks for subsurface property prediction, LiDAR-based landslide risk assessment, earth surface feature detection by instance segmentation and dynamic FEM simulations for structural behavior under environmental and operational loads.In this study, machine learning models were trained on extensive historical highway datasets to predict geotechnical properties, including grain size distribution, SPT-N values, and ground water depth, and achieved acceptable accuracy scores. Feature importance was interpreted through SHAP analysis. The predictions from these ML models were then integrated into FEM as material and geometry voxel input for time-dependent slope stability, train-induced ground vibration response, and pavement falling weight deflectometer (FWD) response analysis. The proposed digital twins enabled the real-time monitoring and analysis of infrastructure conditions which could enhance the decision-making processes, by providing useful insights in risk mitigation and maintenance planning. The research highlights the potential for AI-embedded digital twins to optimize infrastructure management, offering scalable and cost-effective solutions for assessing and maintaining critical assets. Future studies are proposed to expand dataset diversity, integrate real-time input from sensors, and refine hybrid ML-FEM methodologies, which could further improve the resilience and adaptivity of infrastructure systems in the future. en AI-POWERED DIGITAL TWIN MODELING USING LARGE HIGHWAY INFRASTRUCTURE DATASETS Dissertation Civil engineering Digital Twin Finite Element Analysis Highway Dataset Machine Learning embargo