Sessions & Events

AI and Data Analytics for Structural Monitoring and Evaluation

Sunday, October 26, 2025  3:30 PM - 5:30 PM, H - Key ballroom 12

This session will explore cutting-edge advancements in using Artificial Intelligence (AI) and data analytics for monitoring and evaluating concrete structures. As concrete remains a critical component of global infrastructure, effective monitoring methods are essential to ensure safety, performance, and longevity. Presentations will delve into AI-driven approaches, such as machine learning models for real-time damage detection, computer vision techniques for crack analysis, and data-driven predictive maintenance strategies tailored for concrete assets. The session will highlight practical applications, including case studies where AI has significantly improved the assessment and management of concrete structures. Attendees will gain insights into overcoming the unique challenges of concrete monitoring, leveraging AI tools to enhance reliability, reduce maintenance costs, and ensure the structural integrity of these essential assets.

Learning Objectives:
(1) Apply machine learning to field observation data to facilitate condition assessment of concrete bridge elements;
(2) Demonstrate the integration of AI with physics-based models for structural performance assessment;
(3) Present case studies showcasing cost-effective, non-destructive approaches to bridge evaluation;
(4) Integrate machine learning into non-destructive evaluations techniques to improve their accuracy.

Advancing In-Situ Concrete Strength Estimation Using Machine Learning and Hybrid NDT Techniques

Presented By: Hamed Layssi
Affiliation: Fprimec Solutions Inc.
Description: This presentation explores the application of machine learning (ML) techniques to enhance the estimation of concrete compressive strength using in-situ non-destructive tests—specifically Ultrasonic Pulse Velocity (UPV) and Schmidt Rebound Hammer (RN). Traditional SonReb methods often depend on linear or polynomial regression models, which are constrained by dataset-specific calibrations and lack flexibility across diverse concrete types and field conditions. To address these limitations, a novel machine learning toolbox has been developed, leveraging large and diverse datasets to significantly improve prediction accuracy. At the core of this toolbox is an Adaptive Neuro-Fuzzy Inference System (ANFIS) model, trained and validated using independent datasets collected from extensive laboratory experiments. The model was further refined by incorporating key influencing factors such as specimen geometry and testing conditions, leading to a robust and adaptive framework capable of delivering accurate strength estimates across a broad range of structural configurations. The presentation will highlight two real-world case studies where this ML-driven approach has been implemented and validated, demonstrating its practical value. The results underline the transformative potential of machine learning in condition assessment and structural evaluation—offering enhanced accuracy, reduced dependence on intrusive testing (e.g., core extraction), and improved efficiency for field engineers working on complex infrastructure projects.

Machine Learning for Associating Field-Measured Crack Information with Beam Condition

Presented By: Pinar Okumus
Affiliation: University at Buffalo
Description: Shear cracking of prestressed concrete beams in service may be a cause for concern that requires evaluation of the beam condition. Conventional evaluation methods include load testing or detailed finite element analyses, which may be costly or time consuming. This presentation focuses on a new method that utilizes machine learning algorithms to associate crack widths measured in the field with beam condition indicators such as shear loading and stiffness. The algorithms are trained using laboratory test data of prestressed concrete beams that failed under shear with crack widths documented. Material properties, geometric properties and crack widths are used as input for a Gaussian Process Regression algorithm to predict shear capacity, load and stiffness. The results show that machine learning can provide reasonable predictions of shear load corresponding to a crack width and the accuracy of the predictions depend on the size and quality of the available training data.

Data Analytics of Camber in Constructed Prestressed Concrete Bridges

Presented By: Mohamed Issa
Affiliation:
Description: This presentation discusses the data analytics of camber in constructed prestressed concrete bridges. Over 1,200 datasets are collected in the field and computationally processed to generate meaningful information. A comparative study is conducted between the measured and predicted values, with their differences statistically characterized based on the geometry of the bridges. In addition, the influence of service years is examined to determine its potential correlation with the magnitude of in-situ camber. Advanced mathematical models are formulated to predict the camber of the bridges, which are integrated with a machine learning approach.

Field-Collected Data Analytics for Bridge Monitoring and Evaluation Using AI

Presented By: Mi-Geum Chorzepa
Affiliation: University of Georgia
Description: The performance of bridge girders in service bridges can be effectively assessed using field live load and strain data measured non-destructively. This study presents an innovative approach that combines real-time live load data with physics-based models to evaluate bridge girder performance. By integrating a generative model with mode shapes observed in ambient vibration tests, we enable a more accurate assessment of structural health and dynamic performance. The physics-based model incorporates material properties and mode shapes to simulate realistic performance under varying conditions. Using practical bridge case studies, this methodology provides valuable insights for bridge improvements, offering a cost-effective, non-destructive alternative to traditional evaluation techniques while advancing bridge monitoring, evaluation, and asset management.