Embedded GFRP-packaged sensor system for real-time and high-precision vehicle load estimation in weigh-in-motion (WIM) applications

Abstract Accurate weigh-in-motion (WIM) systems are critical for effective transportation network planning, pavement asset management, and axle-load enforcement. However, conventional surface-mounted or cut-in WIM sensors often suffer from poor durability, temperature sensitivity, and limited awareness of lateral wheel offset. This study presents an embedded WIM approach that integrates glass fiber-reinforced polymer-packaged fiber Bragg grating (FBG) sensors with physics-based simulation and data-driven estimation. Six FBG sensors were installed at two depths (2.5” and 4.5”) within the flexible pavement to enhance strain transfer efficiency and capture multilayer mechanical responses. Vertical strain fields under varying wheel loads and transverse offsets were modeled in KENPAVE, and corresponding Bragg wavelength shifts were correlated with radial distance r from the wheel path. A Gaussian inverse model was developed to estimate the lateral wheel offset and compensate for wander-induced variations before load inference. Field validation was conducted using data from nineteen trucks at a North Dakota weigh station, where Bragg shifts and independently verified axle weights were collected for calibration and testing. Among several machine-learning regressors trained on extracted features, XGBoost achieved the highest accuracy, yielding a mean absolute percentage error of 9.9%. These results demonstrate the field-scale proof-of-concept feasibility of the integrated sensing–simulation–machine learning framework under live traffic conditions. The proposed system offers a non-intrusive, robust, and wander-resilient solution, establishing a methodological foundation for scalable high-level WIM implementation, subject to expanded statistical validation and controlled certification testing.