Machine Learning Predictive Models for Load Identification of Railroad Bridges Subjected to Train-Induced Vibration

Aug 5, 2026

Upload by Noah

The U.S. rail network relies on numerous railroad bridges that are often over a century old, experiencing structural deterioration beyond their practical service life. Rising train weight and speed increase the risk to these vital assets, while traditional load identification and monitoring approaches such as Weigh-in-Motion (WIM) are seldom used due to cost and operational limitations. This thesis develops a cost-effective strategy using vibration sensors and machine learning (LSTM, RNN) with finite element modeling to estimate axle load and train speed from vibration data, achieving acceptable time-series prediction performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages98

Reading Time4 h 7 min

AccessFree to download

Abstract

Background and Motivation

Limitations of Traditional Load Identification

Proposed Approach

Finite Element Modeling

Vibration-Sensor Data and ML Algorithms

Model Training and Evaluation

Time-Series Load and Speed Prediction

Results and Findings

University of Texas Rio Grande Valley ScholarWorks @ UTRGV Theses and Dissertations 5-2024 Machine Learning Predictive Models for Load Identification of Railroad Bridges Subjected to Train-Induced Vibration Md Masnun Rahman The University of Texas Rio Grande Valley Follow this and additional works at: [https://scholarworks.utrgv.edu/etd](https://scholarworks.utrgv.edu/etd) Part of the Civil Engineering Commons Recommended Citation Rahman, Md Masnun, "Machine Learning Predictive Models for Load Identification of Railroad Bridges Subjected to Train-Induced Vibration" (2024) . Theses and Dissertations. 1488. [https://scholarworks.utrgv.edu/etd/1488](https://scholarworks.utrgv.edu/etd/1488) This Thesis is brought to you for free and open access by ScholarWorks @ UTRGV. It has been accepted for inclusion in Theses and Dissertations by an authorized administrator of ScholarWorks @ UTRGV. For more information, [please contact](please contact justin.white@utrgv.edu)[ justin.white@utrgv.edu](please contact justin.white@utrgv.edu), [william.flores01@utrgv.edu](william.flores01@utrgv.edu). MACHINE LEARNING PREDICTIVE MODELS FOR LOAD IDENTIFICATION OF RAILROAD BRIDGES SUBJECTED TO TRAIN-INDUCED VIBRATION A Thesis by MD MASNUN RAHMAN Submitted in Partial Fulfillment of the Requirements for the Degree of MASTER OF SCIENCE Major Subject: Civil Engineering The University of Texas Rio Grande Valley May 2024 MACHINE LEARNING PREDICTIVE MODELS FOR LOAD IDENTIFICATION OF RAILROAD BRIDGES SUBJECTED TO TRAIN-INDUCED VIBRATION A Thesis by MD MASNUN RAHMAN COMMITTEE MEMBERS Dr. Mohsen Amjadian Chair of Committee Dr. Philip Park Committee Member Dr. Constantine M. Tarawneh Committee Member Dr. Fatemeh Nazari Committee Member Copyright 2024 Md Masnun Rahman All Rights Reserved ABSTRACT Rahman, Md Masnun, ML Predictive Models for Load Identification of Railroad Bridges Subjected to Train-Induced Vibration. Master of Science (MS), May 2024, 77 pp., 2 tables, 29 figures, 44 references, 6 titles. The U.S. rail network is among the world's largest, safest, and most efficient. However, many its railway bridges, which play an essential role in its connectivity, are over 100 years old, showing signs of structural deterioration exceeding their practical service life. The increased weight and speed of modern trains also pose a risk to these bridges. Traditional load identification and monitoring technologies like Weigh-in-Motion (WIM) systems are rarely used on railway bridges due to their cost and practical limits. This thesis explores a cost-effective approach to utilize vibration sensors on railway bridges to identify and predict train load features (speed and axle load), using machine learning (ML). The study involves finite element modeling of railway bridges, followed by applying ML algorithms like LSTM and RNN to estimate axle loads and speeds from vibration data. The results show the acceptable performance of developed time series ML models in predicting these parameters. DEDICATION This master's thesis is dedicated to my parents and siblings, whose unwavering love and encouragement have been my driving force. My mother, Laila Parvin, my father, Md Mobassar Ali, and my siblings, Md Manjurul Alam and Md Mehesum Rahman, are the reasons I am here today. Their belief in my abilities and the sacrifices made for my education inspire me every day. I dedicate this work to all individuals striving for knowledge and positive change. May this thesis contribute, even in a small way, to the collective pursuit of understanding and progress. To my family, friends, and mentors, thank you for your unwavering support and belief in me.

FAQs

Why is load identification for railroad bridges important in this thesis?

Railroad bridges form critical connectivity in the U.S. rail network, yet many are aged and deteriorating. Increased train weight and speed further elevate structural risk, making reliable load identification essential.

What alternative to Weigh-in-Motion (WIM) is proposed?

The thesis proposes using vibration sensors on railway bridges as a more cost-effective monitoring approach. Machine learning then derives train load features such as speed and axle load from vibration data.

Which machine learning models are used to estimate axle load and speed?

The study applies time-series machine learning algorithms, including LSTM and RNN, to estimate axle loads and train speeds from the measured vibration signals.

How are the predictive models developed and assessed?

Finite element modeling supports the study, followed by training the ML models on vibration data. Results report acceptable performance for predicting the target parameters over time series.

Aug 5, 2026

Upload by Noah