Traffic Data Quality Verification and Sensor Calibration for Weigh-In-Motion (WIM) Systems

Abstract

Many state departments of transportation have been collecting various traffic data through the Automatic Traffic Recorder (ATR) and Weigh-in-Motion (WIM) systems as outlined in the Traffic Monitoring Guide (TMG) published by USDOT. A pooled fund study led by MnDOT was conducted in 2002 to determine traffic data editing procedures. It is challenging to identify potential problems associated with the collected data and ensure data quality. The WIM system itself presents difficulty in obtaining accurate data due to sensor characteristics, complex vehicle dynamics, and the pavement changes surrounding the sensor over time. To overcome these limitations, calibration procedures and other monitoring activities are essential to data reliability and accuracy. Current practice of WIM calibration procedures varies from organization to organization. This project aims to understand the characteristics of WIM measurements, identify different WIM operational modes, and develop mixture models for each operation period. Several statistical data analysis methodologies were explored to detect measurement drifts and support sensor calibration. A mixture modeling technique using Expectation Maximization (EM) algorithm and cumulative sum (CUSUM) methodologies were explored for data quality assurance. An adjusting CUSUM methodology was used to detect data anomaly. The results indicated that the adjusting CUSUM methodology was able to detect the sensor drifts. The CUSUM curves can trigger a potential drifting alert to the WIM manager. Further investigation was performed to compare the CUSUM deviation and the calibration adjustment. However, the analysis results did not indicate any relationship between the computed CUSUM deviation and the calibration adjustment.