A Deep-Learning Model to Estimate Rigid Pavement Stresses for Top-Down Cracking Airfield Pavement Design

Abstract

The Federal Aviation Administration (FAA) initiated a research project to add the top-down cracking failure mode to the rigid pavement design process in the FAARFIELD program. To this end, FAA developed a fast-running stress prediction model based on machine learning (ML) methods to substitute for the three-dimensional finite element (3D-FE ) model used by FAARFIELD. The model targeted the design of a four-layer airfield rigid pavement serving commercial aircraft heavier than 100,000 pounds gross weight. Recently, FAA extended the model to include relatively thin rigid pavements at facilities serving light-load aircraft (such as general aviation airports), and three-layer rigid pavements designed to support aircraft heavier than 100,000 pounds. This paper presents the results of the extended model. The ML model was trained using a comprehensive database consisting of the results of 250,000 3D-FE simulations of a rigid pavement system with wide ranges of material and thickness parameters, thermal loads, and aircraft configurations. The model is based on a modular deep learning method employing a new artificial neural network (ANN) method that predicts a dynamic functional evaluated over a continuous domain. The training operation was performed using backpropagation and the ADAHESSIAN numerical optimization algorithm. The model quickly estimates stress distribution along slab edges due to aircraft and thermal loads. The end result of this research will be an accurate and computationally efficient ML-based model suitable for incorporating to a new cumulative damage factor (CFD) design for top-down cracking failure in FAARFIELD.