Using Advanced Modeling Techniques for Improving the Existing Airfield Pavement Management System Considering Structural and Functional Condition Indices

Airport authorities constantly collect pavement condition data and utilize life-cycle cost analysis to select construction and maintenance alternatives. The current Federal Aviation Administration (FAA) Advisory Circular 150/5380-7B recommends using pavement condition index (PCI) to assess airfield pavement condition for planning of maintenance and rehabilitation (M&R) treatments. However, structural and functional distresses initiate different pavement deterioration behavior that might not be fairly represented by one indicator, the PCI. The latter might mask the root cause of the airfield pavement deterioration by limiting the ability to focus on one failure type over the other. This results in inadequate M&R recommendations. Furthermore, the regression nature of the existing airfield pavement assessment models is not adequate for pavement performance prediction as a function of multitude features. Hence, the objective of this study was to evaluate the airfield pavement condition of key airport components (runway, taxiway, and apron) using three indices; PCI, structural condition index (SCI), and foreign object debris/damage (FOD). To achieve this objective, three machine learning algorithms, namely, ensemble-learning method, CatBoost, and LightGBM, were used to predict the SCI and FOD based on the corresponding PCI value and key project conditions such as pavement age, branch use, pavement surface type, inspection year, aircraft average operation per day, air traffic composition (%General, %Transient, %Military, and %Air Taxi aviation), mean annual temperature, annual cumulative rainfall, annual cumulative rainfall days, and annual cumulative snowfall. A total of 2,505 pavement sections obtained from 89 airport networks in seven states were included in the analysis. Based on the collected data, two models were trained (for each algorithm) and validated to predict the SCI and FOD of airport sections from the existing PCI database and other key inputs. Results indicated that the CatBoost algorithm yielded the highest accuracy; therefore, the CatBoost-based models were considered in the proposed decision-making framework. This framework is expected to advance the existing FAA PCI-based approach and expand the current airfield performance models beyond pavement age using robust techniques.