Electronic Thesis/Dissertation
 

AI-Driven Prediction of Alligator Cracking in Asphalt Pavements Using Machine and Deep Learning Models

Open Access Deposited

Prior to the formation of potholes, alligator or fatigue cracking of asphalt concrete in pavement is a maintenance problem for transportation agencies. Alligator cracking, despite being a cheaper repair option, is often a precursor to serious road damage, including potholes. (Hall et al., 2001, p. A-1). Successful infrastructure management requires better resource distribution and improved pavement life through data-driven predictive maintenance models (Kumar & Shoghli, 2022). This research discusses the need for a strong methodology to forecast pavement deterioration (Alligator Cracking). This research also develops a predictive modeling system to estimate the extent of alligator cracking in asphalt pavements using data from the Long-Term Pavement Performance (LTPP) program. The main goals are identifying the primary causes of pavement cracking, the performance of different machine learning and deep learning algorithms, and creating a stable tool to forecast the pavement performance. The methodology involved multistage data preparation and modeling. Climatic, traffic, and other variables extracted from pavement structure that had been excluded from the LTPP database were combined and filtered. A systematic evaluation was conducted on deep learning architectures (TabNet, MLP, and FT-Transformer) and machine learning models such as LightGBM, XGBoost, and Random Forest. The efficacy of models was determined using the coefficient of determination (R2), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). XGBoost performs the best, obtaining an R2 of 0.957, RMSE of 0.1276, and MAE of 0.0579 in the held-out test set. This outcome greatly exceeded the performance of all other tested models, including the top-performing deep learning model, the FT-Transformer, which achieved an R2 score of 0.615.

Author Language Date created Type of Work License
  • All rights reserved
Rights statement GW Unit Degree Advisor Committee Member(s) Persistent URL

Notice to Authors

If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.

Thumbnail Title Date Uploaded Visibility Actions
Preview of Bangre_gwu_0075A_17900.pdf Bangre_gwu_0075A_17900.pdf 2026-06-24 Open Access