Proposal on Implementing Machine Learning with Highway Datasets

Abstract

Every year State Highway Agencies (SHA) invests millions of dollars into testing materials to optimize engineering designs, and challenges are bound to be encountered in the areas of projects cost, rehabilitation process, delays in construction activities, rate of materials, maintenance costs, risk analysis etc. which are highly complicated in nature. Many Departments of Transportation regularly evaluate the condition of infrastructure through visual inspections, nondestructive evaluations, image recognition models and learning algorithms. State Highway Agencies (SHAs) across the United States are now able to collect a large amount of pavement condition information because of these advanced technological data collection methods. SHA and many other agencies collect pavement performance data, which encompass measurements of the international roughness index (IRI) and rut depth using electronic sensing devices that utilize laser, acoustic, and infrared technologies. These agencies also use imaging technologies and automated image processing techniques to estimate the levels of severity and extent of surface distresses. These effort by SHA result in high-density data that is used to support a variety of decision-making. The challenge is how these historical data can be managed and analyzed while extracting these data to improve practices and decision-making processes. With the advent of big data, from large-scale databases to data mining applications, there has been a tremendous progress in machine learning intelligence for understanding and optimizing engineering processes. This provides an opportunity for SHA to implement machine learning (ML) for large datasets in materials and testing including pavement data, construction history, slope stability, and geologic risk. There can be significant cost savings in many SHA transportation and infrastructural projects geared from supervised learning which leverages on historic data sets. Artificial Intelligence (AI) techniques like fuzzy logic, case-based reasoning, probabilistic methods for uncertain reasoning, classifiers and learning methods, Artificial Neural Networks (ANN), Genetic Algorithms and hybrid techniques have been widely used in the many applications in the engineering field. Thus, a dynamic pavement condition algorithm that allows agencies to detect pavement segments in need of rehabilitation would be beneficial for a variety of decision making including pavement maintenance performance evaluation, pavement deterioration model development, and budgeting. This project develops a pavement test condition-prediction algorithm using ANN (Artificial Neural Network) to dynamically detect untested road segments for SHA programming. This project uses an artificial neural network simulator to suggest locations for pavement testing program from historical datasets. INTRODUCTION Every year State Highway Agencies (SHA) and many other agencies collect pavement performance data, which