Investigating Axle Load Data Quality from Portable Weigh-in-Motion Systems on Canadian Roadways: Practical Lessons for Deployment and Postprocessing

This article investigates the quality of axle load data collected from a portable weigh-in-motion (PWIM) system installed near a permanent WIM in Manitoba, Canada. Using the permanent WIM as the reference dataset, the investigation examined record pairing, calibration drift, the impact of temperature and speed on data quality, per-vehicle data validity, and aggregated data validity. The findings showed that the PWIM system was unable to produce sufficiently accurate data for most applications. This was true even when the system was precalibrated and subject to temperature corrections and autocalibration postprocessing. Error tolerances two to three times higher than those specified by ASTM (Type II) would be required to meet the standard. The results corroborated previous findings in the literature and reinforced the skepticism practitioners have about PWIM data validity. However, the investigation offered practical lessons concerning PWIM system deployment and postprocessing that could support further research and development. Specifically, PWIMs exhibit limited sensitivity to highway speeds and may serve as a cost-effective alternative for collecting traffic and axle load data on secondary highways. The postprocessing methods were found to be effective at stabilizing measurements for fully loaded vehicles. Further, the Gaussian mixture modeling technique proposed in this article offers an approach for using PWIM data to attribute higher quality load data to the PWIM site.