Prediction Damage Factor Classification of Airport Pavement using Machine Learning

Abstract

Airport pavements experience continuous stress from operating large aircraft, compromising their structural integrity. Maintaining pavement resilience is crucial for ensuring flight safety. The Damage Factor (DF) is a crucial metric for evaluating the cumulative effect of aircraft loads on surface conditions. Accurate damage assessment and prediction are crucial for strategic maintenance planning and operational stability. This research examines the capacity of advanced mathematical methodologies, particularly machine learning, to identify and predict damage determinants. The researchers created sophisticated models that utilize extensive datasets, encompassing historical pavement performance records, aircraft movements, meteorological data, and material characteristics. The study aimed to enhance prediction capabilities using advanced algorithms, including random forest, logistic regression approaches linear SVM, tree and ensemble models. The findings indicated that the Majority Voting (MV) (random forest, decision tree and logistic regression) model attained the highest test accuracy of 81.82%, and ROC AUC value 0.8461, validating its robust capacity to accurately discern data patterns. Conversely, the naïve Bayes model exhibited markedly inferior performance, with a test accuracy of only 52.27%, underscoring its limited applicability to the dataset compared to alternative methodologies. The results reveal that the exceptional performance of the (MV) classification effectively identify underlying patterns in airport pavement deterioration data. This allows for extrapolation to a substantial training set under novel and unforeseen conditions.