Instability detection for high-speed weigh-in-motion systems towards direct enforcement
Last modified: 2024-04-10
Abstract
High Speed Weigh-In-Motion (HS-WIM) systems are used to automatically enforce weight restrictions on heavy vehicles operating within specified maximum weight limits. In Belgium, these systems have recently been implemented into the road surfaces of highways at several locations. They accurately estimate the weight of semi-trailers and vans moving at high speeds, meeting Class I standards according to the E1318-94 criterion.
Despite the initial precision in weight measurement, it is crucial for the system to maintain calibration for reliable estimations over an extended period. Currently, periodic tests of estimation errors are conducted, and the parameters of the system are calibrated when necessary. The main objective of this research is to develop a complete automatic monitoring of this HS-WIM system, operating without the need for constant supervision. In the field of changepoint detection methods, papers have laid the theoretical background for developing such monitoring, assuming strong mathematical characteristics on the nature of the analyzed data and the changes. More recently, a breakthrough research on locally stationary time series extended the previous theory avoiding strong assumptions on the dataset. However, extensive calculations are required to perform these changepoint detections. Despite lower bounds of thresholds are suggested, no exhaustive solution is provided in the initial research.
This paper then aims to present the complete practical implementation of unsupervised monitoring used for the HS-WIM. First, it especially addresses the issue of natural dependence between the HS-WIM data and external variables to the system by performing a pre-processing stage.
Second, we propose parametrized approximations of the theoretical background thanks to characteristics of the HS-WIM data. These parameters are user-defined, considering the available time and memory capabilities.
Third, we show that we can enhance the sensitivity of the monitoring by using our procedure compared to the usual parametric method.
Despite the initial precision in weight measurement, it is crucial for the system to maintain calibration for reliable estimations over an extended period. Currently, periodic tests of estimation errors are conducted, and the parameters of the system are calibrated when necessary. The main objective of this research is to develop a complete automatic monitoring of this HS-WIM system, operating without the need for constant supervision. In the field of changepoint detection methods, papers have laid the theoretical background for developing such monitoring, assuming strong mathematical characteristics on the nature of the analyzed data and the changes. More recently, a breakthrough research on locally stationary time series extended the previous theory avoiding strong assumptions on the dataset. However, extensive calculations are required to perform these changepoint detections. Despite lower bounds of thresholds are suggested, no exhaustive solution is provided in the initial research.
This paper then aims to present the complete practical implementation of unsupervised monitoring used for the HS-WIM. First, it especially addresses the issue of natural dependence between the HS-WIM data and external variables to the system by performing a pre-processing stage.
Second, we propose parametrized approximations of the theoretical background thanks to characteristics of the HS-WIM data. These parameters are user-defined, considering the available time and memory capabilities.
Third, we show that we can enhance the sensitivity of the monitoring by using our procedure compared to the usual parametric method.
Keywords
Weigh-In-Motion;Monitoring;Non-parametric