Designing High Speed Weigh-in-Motion System With Principal Component Regression in Wallonia (Belgium) Toward Direct Weight Enforcement

The use of High Speed Weigh-In-Motion (HS-WIM) systems towards direct enforcement is growing worldwide. Sensors can be implemented inside the roadway and directly weigh the moving vehicle. Regarding the specific system in Wallonia (Belgium), the class I is reached for the semi-trailers and for the vans according to the E1318-09 criterion. However, this accuracy is only obtained for a small number of vehicles, which are filtered by using an algorithm of selection as detailed in Antofie et al. (2019). The purpose of this paper is to present an algorithm with the capability of increasing the number of selected vehicles and improving the accuracy of the raw system. This work aims to analyze the mutli-colinearity of some features of a passing vehicle and propose a mathematical model based on the principal component analysis to estimate its weight with high accuracy. A cross-validation of the algorithm is performed in order to choose the best hyper-parameters. We especially bring up the topic of the accuracy of the HS-WIM regarding the number of needed vehicles to train our model, for data sets recorded under full reproducibility (R2) and limited reproducibility (R1) conditions.