Pavement Data Analytics for Collected Sensor Data

The Federal Aviation Administration instrumented four concrete slabs of a taxiway at the John F. Kennedy International Airport to collect pavement responses under aircraft and environmental loading. The study started with developing preprocessing scripts to organize, structure, and clean the collected data. As a result of the preprocessing step, the data became easier and more intuitive for pavement engineers and researchers to transform and process. After the data were cleaned and organized, they were used to develop two prediction models. The first prediction model employs a Bayesian calibration framework to estimate the unknown material parameters of the concrete pavement. Additionally, the posterior distributions resulting from the calibration process served as a sensitivity analysis by reporting the significance of each parameter for temperature distribution. The second prediction model utilized a machine-learning (ML) algorithm to predict pavement responses under aircraft and environmental loadings. The results demonstrated that ML can predict the responses with high accuracy at a low computational cost. This project highlighted the potential of using ML for future pavement design guidelines as more instrumentation data from future projects are collected to incorporate various material properties and pavement structures.