A multi-scale framework for the design of lead-free potassium sodium niobate pavement sensor arrangements for vehicle classification

• A simulation-based framework is proposed for evaluating pavement sensor arrangement strategies. • FEM-derived attenuation functions are used to compare different sensor layouts without repeated full-scale simulations. • The influence of vehicle paths and wheel-track constraints on sensor response distribution is examined. • The framework supports comparative assessment of sensor arrangements under defined modeling assumptions. An efficient sensor arrangement is critical for intelligent pavement monitoring, as dense sensor arrangements increase construction complexity and pavement disturbance. Finite element modeling (FEM) can accurately represent vehicle-pavement-sensor interactions, but repeated simulations limit its efficiency for large-scale layout evaluation. This study proposes a simulation-based method (ESAM) for evaluating pavement sensor arrangements using lead-free potassium sodium niobate (KNN) piezoelectric sensors with low piezoelectric response. Finite element outputs are used to describe sensor voltage responses under vehicle loading, and a voltage-distance relationship is used to estimate responses without repeated FEM simulations. Multiple sensor arrangements are examined under representative vehicle passage scenarios and lane configurations. Sensor arrangements are evaluated using a rule-based confusion matrix, from which a classification accuracy ( AC ) is derived as an indicator of signal separability. The results show that, under the modeled conditions, several layouts can reduce sensor numbers compared with dense deployments while maintaining high AC under the simulated conditions. Uniform Arrangement demonstrated high performance across the single-lane scenarios. Centrally Weighted Arrangement had more advantages in handling the classification of large vehicles. Minimum Coverage Arrangement significantly reduced the number of sensors required when high accuracy was not essential. The results clarify how sensor number and spatial arrangement influence classification performance in the simulated cases. The proposed method enables rapid and consistent comparison of sensor arrangement strategies in KNN-based pavement sensing systems.