Date Approved
6-20-2024
Embargo Period
6-24-2025
Document Type
Thesis
Degree Name
Master of Science (M.S.) in Data Science
Department
Computer Science
College
College of Science & Mathematics
Sponsor
U.S. Department of Transportation, Accelerated Bridge Construction University Transportation Center,
Advisor
Islam Mantawy, Ph.D.
Committee Member 1
Adriana Trias, Ph.D.
Committee Member 2
Anthony Brietzman, Ph.D.
Keywords
machine learning, bridge maintenance
Subject(s)
Structural health monitoring
Disciplines
Civil and Environmental Engineering | Computer Sciences
Abstract
Machine learning-based structural health monitoring (ML-SHM) plays a pivotal role in enhancing structural resilience. By recognizing potential hazards, implementing resistance measures, facilitating swift recovery, and continuously monitoring structural health, ML-SHM ensures proactive maintenance and minimizes recovery delays post-events. Leveraging machine learning algorithms and sensor data, ML-SHM enables early detection of anomalies, prediction of failures, and adaptive responses, enhancing the structure's ability to withstand and recover from adverse conditions. This integrated approach not only improves the structure's performance and adaptability but also contributes to overall safety and longevity. This thesis presents a comprehensive exploration of structural health monitoring (SHM) techniques for bridges, employing varied approaches to assess damages, predict responses, and enhance overall resilience. In this study, I encompassed four distinct applications, each harnessing innovative methodologies and machine learning models. In the first application, I developed a convolutional neural network (CNN)-based approach for detecting spalling of columns. Leveraging image data from a shaking table experiment on a quarter-scale two-span reinforced bridge, I employed diverse CNN models, including transfer learning and data augmentation. Notably, the ResNet50 model achieved exceptional accuracy, scoring 100% in training and 97% in testing. In the second application, I employed machine learning algorithms for detecting damage in reinforcing bars within columns, utilizing strain cycles. A hybrid model combining a Gaussian naive Bayes base estimator and “ada” boost classifier is implemented, achieving consistent high accuracy with a mean test accuracy of 95%. This study provides significant insights into machine learning integration for structural health monitoring. In the third application, the second application was innovated by addressing challenges in monitoring low-cycle fatigue, specifically rocking columns in bridges. Utilizing CNNs, I directly encoded strain time series data into images through the Markov transition field technique. This advancement bypasses the need for calculating strain cycles, achieving exceptional accuracy of 100% in training and 97.8% in testing. The project highlights the potential of CNNs in predicting reinforcing bar fractures due to low-cycle fatigue effects. In the fourth application, I focused on predicting response (accelerations) in rocking column bridges during seismic events. Employing gated recurrent unit (GRU) machine learning models, I evaluated lateral displacements and corresponding accelerations. The findings improve the accuracy of seismic response predictions achieving a test output of over 0.9 R2, offering important insights into the responses of rocking column bridges under various seismic conditions. Collectively, by conducting these applications, I demonstrated the effectiveness of advanced machine learning techniques, notably CNNs and GRUs, in comprehensive structural health monitoring for bridges. This research not only advances the understanding of damage detection and prediction but also offers practical applications to enhance critical infrastructure resilience and safety.
Recommended Citation
Ravuri, Naga Lakshmi Chittitalli, "INTEGRATION OF MACHINE LEARNING IN STRUCTURAL HEALTH MONITORING FOR DAMAGE IDENTIFICATION AND RESPONSE PREDICTION IN BRIDGES" (2024). Theses and Dissertations. 3252.
https://rdw.rowan.edu/etd/3252
