This report describes a method that aims to use low-cost vehicle classifiers to provide some indication of pavement loading or gross vehicle mass (GVM). The proposed methodology in this report aims to identify, from a list of candidate WIM sites (therefore with known GVM frequency distributions), the one that can give the best indication of the GVM distribution at a classifier site. The method consists of two stages. The first stage determines whether the loading characteristics for a vehicle class in a jurisdiction are suitable for correlating classified counts with WIM data. It is based on the analysis of GVM cumulative frequency distributions of WIM sites and the use of the Kolmogorov-Smirnov Statistic (KSS). The second stage is to identify the best site from a list of candidate WIM sites to match the data at an intelligent classifier site, if the loading characteristic of that jurisdiction is found suitable. The method has been found robust and all the analyses in this report appear to have produced the right matches. More analyses would be useful and it is likely that more WIM site data will provide a larger range of sites for matching. It is recommended that each jurisdiction evaluate the method with its own WIM data in the same manner as employed in this report. When the method is found satisfactory, a jurisdiction should consider implementing intelligent classifier sites with piezo-cables to determine the loading frequency distribution at any site in its road network with information (or assumed information) on the level of unladenness. The correlation methodology investigated in this report is quite general and can be employed for other applications in road use data collection. For example, some further research can extend the framework for identifying the permanent count site that provides the best count statistics for a temporary count site, or any site that is in need of comprehensive count data. (Author/publisher)
Abstract