A Machine Learning–Based Framework for Predicting Pavement Roughness and Aggregate Segregation during Construction

Abstract

Pavement construction monitoring and quality assurance (QA) practices are based mostly on costly, discrete, and destructive methods. Most quality assurance programs are based on pavement construction procedures encompassing in situ coring for layer thickness determination, density measurements, laboratory testing to measure volumetric properties, and smoothness measurements in the case of the availability of an inertial profiler. However, most of these practices are costly and/or destructive. Therefore, the key objective of this study was to develop a quality assurance decision-making tool that can predict pavement roughness, in terms of the International Roughness Index (IRI), and aggregate segregation based on digital image analysis, image recognition, and deep machine learning models. The developed models were trained, tested, and validated using 600 pavement surface images extracted from the Louisiana Department of Transportation and Development (LaDOTD) Pavement Management System (PMS). Furthermore, the effectiveness of the convolution neural network (CNN) model was validated using pavement surface images collected at construction sites in Louisiana a few days after paving. The roughness model predicted the International Roughness Index values with a coefficient of determination R2 of 0.98 and a RMS error (RMSE) of 3.5%. In addition, the developed image-processing model for the detection of aggregate segregation achieved acceptable accuracy. To support the implementation of these results, the models were incorporated into a computer application that can be used by site engineers for quality assurance without the need for coding software on their device.