Multiway-Based Weigh-in-Motion Data-Clustering Analysis for Pavement ME Design

Abstract

Weigh-in-motion (WIM) systems are widely used to study traffic patterns and generate traffic inputs for Pavement Mechanistic-Empirical (ME) Design. Clustering analysis based on individual traffic parameters has been implemented in several studies to prepare Level 2 traffic data for Pavement ME Design where site-specific WIM stations are not available. Recognizing that an individual traffic input cannot fully represent the multiattributes of the traffic pattern of a WIM station, this paper introduces a multiway-based cluster approach to grouping available WIM data sets. Four-way WIM data, including the truck volumes of 10 vehicle classes for their corresponding load bins in 12 months of a year at the 31 WIM stations in Michigan, are prepared in this paper. A parallel factor (PARAFAC) model is applied to decompose the multiway WIM data and correlate the multiple traffic characteristics using component scores between each mode. The component scores of the WIM stations are further used as the input data for a hierarchy clustering analysis to generate traffic groups and examine traffic patterns. A case study is performed to demonstrate the advantage of proposed methodology to prepare Level 2 traffic inputs for Pavement ME Design.