AsphaltMine: The asphalt test result database for performance prediction

Abstract

The lack of a unified, accessible platform for collecting and sharing standardized asphalt test result data has limited progress in data-driven pavement research. To address this challenge, AsphaltMine (https://asphaltmine.org) is being developed as a comprehensive web application platform dedicated to advancing data-driven pavement engineering. It allows users to collect asphalt test results in an organized online database, which can later be used to train machine learning models for predicting asphalt performance. The platform enables laboratories to contribute and retrieve unified laboratory test data for different tests, currently including the Rutting test, Marshall test, ITS, TSRST, and dynamic stiffness test. In addition to mixture test results, users are requested to input the corresponding component material test results and material proportions. The schema for these tests follows the European test standards, allowing standardized parameters to be entered. Users can upload data via an interactive web form or through a unified Excel sheet using a bulk import tool. The results are ultimately stored in a NoSQL database. In this paper, we discuss the potential benefits and challenges of creating such a database as well as describe the system architecture and data model of AsphaltMine, including the curation process and record storage, record structures, and user record management functionalities. We then demonstrate the data analysis and prediction features using the first test result prediction tool that was developed using the collected test results.