Runway occupancy time is an essential parameter to estimate the performance of airport operations. With improvements in airport surface radar surveillance technology, estimating runway occupancy time and aircraft exit distance on runways is possible. Past studies have predicted runway occupancy times using traditional simulation-based methods aided with airport observations. However, there are not many attempts to use deep learning and more recent data science algorithms to predict runway occupancy times. This paper describes a neural network algorithm for predicting runway occupancy times for arrival flights at airports. The algorithms used to predict runway occupancy time are data-driven. The data employed in this study is extracted from two years of the Airport Surface Detection Equipment Model-X deployed at 37 airports in the United States. The algorithm's input layer is defined using estimated speed and acceleration parameters for individual aircraft operating at different airports. We studied the performance of our model for fourteen distinct aircraft types at eight different airports and the weighted average R-squared values of the regression analysis between observed and estimated values for our predicted runway occupancy time model was 0.9. The R-squared value for predicted exiting distances was 0.94.
展开▼
