Traffic Inputs for Pavement ME Design Using Oklahoma Data

Abstract

Mechanistic-empirical pavement design guide (MEPDG) requires traffic inputs in three levels based on the availability of data and scale of the project. Site-specific (Level 1) data is high quality and can be obtained by automated traffic data collection techniques like Automatic Vehicle Classification (AVC) and Weigh-in-motion (WIM) data. However, available of Level 1 data is limited and even it is very expensive to obtain data. On the other side, statewide default (Level 3) data has the lowest quality. So, regional-specific (Level 2) with medium quality need to be developed. However, automatic data have errors; this happens more with WIM data. To ensure the quality of data, this research is started with developing QC metrics for the Oklahoma state WIM data and then generating site-specific (Level 1), region-specific (Level 2), and statewide average (Level 3) traffic inputs that are required for the Pavement ME Design in Oklahoma. This process includes performing a comprehensive check for the quality of data by using a software Prep-ME followed by manual review. Developed and presented homogeneous groups for each traffic input by analyzing data with K-means cluster analysis techniques for regional specific (Level 2) inputs. Investigated and identified the available independent variables that are influencing the traffic cluster groups. Decision tree model and Multinomial regression model are developed by training them with available data from multiple stations and multiple years. These models can identify the suitable cluster group for the given site conditions. To evaluate the variation in pavement performance for Level 2 and Level 3 traffic inputs, case study is included. This study can provide a set of procedures and methodology to assist design engineers in developing regional-specific (Level 2) traffic inputs for pavement ME Design in the Oklahoma.