Date of Graduation
Master of Science in Civil Engineering (MSCE)
Sarah V. Hernandez
Kevin D. Hall
Second Committee Member
Clinton M. Wood
Algorithm, Auto-Calibration, Weigh-in-Motion, WIM
This project evaluates the performance of Weigh-in-Motion (WIM) auto-calibration methods used by the Arkansas Department of Transportation (ARDOT). Typical auto-calibration algorithms compare the WIM measured weights of vehicles from the traffic stream to reference values, using five-axle tractor-trailer configured trucks for comparisons, e.g. Federal Highway Administration (FHWA) Class 9. Parameters of the existing algorithms including the Front Axle Weight (FAW) reference value, the sampling frequency required to update the calibration factor, and thresholds for Gross Vehicle Weight (GVW) bins were evaluated. The primary metric used to estimate algorithm performance was Mean Absolute Percent Error (MAPE) between the WIM and static scale GVW measurements.
Two altered auto-calibration algorithms based on methodologies utilized by ARDOT and the Minnesota DOT (MNDOT) were developed. Parameters for the algorithms are based on combinations that produced minimal MAPE at the study sites. WIM data from two sites (Lamar and Lonoke) and static scale data from one site (Alma) were collected along Eastbound I-40 in Arkansas during March 2018. The updated MNDOT auto-calibration algorithm reduced the MAPE by 2.5% compared to the baseline method at the Lamar site (n = 77 trucks) and by 1.1% for the Lonoke site (n = 44 trucks). The updated ARDOT algorithm reduced MAPE by 1.6% at the Lamar site and 0.6% at the Lonoke. Due to limitations of the field data collection methodology, the thresholds defining FAW reference values and the FAW reference values themselves were not able to be tested for spatial transferability, e.g. the samples of trucks at the Lonoke WIM site were a subsample of the trucks at the Lamar WIM site. Improvements in auto-calibration accuracy at low volume sites but was not tested due to the small number of confirmed vehicle matches at a third WIM site (Bald Knob, n = 2 trucks).
Overall, site specific tuning of auto-calibration algorithms will improve the accuracy of WIM data which is used for pavement design, maintenance programming, and traffic forecasting. For example, improvements of 2.5% MAPE of WIM measured GVW results in a 39% difference in the estimated Equivalent Single Axle Load (ESAL) factors used for pavement design.
Baker, J. (2019). Auto-Calibration of WIM Using Traffic Stream Characteristics. Graduate Theses and Dissertations Retrieved from https://scholarworks.uark.edu/etd/3163