A data and model hybrid-driven three-directional forces and friction estimation framework based on intelligent tires

Intelligent tires with embedded sensors are crucial for vehicle safety but lack quantitative placement criteria and physical interpretability. This article proposes a physics-informed hybrid framework to address these challenges. First, a feature-based global sensitivity analysis optimizes piezoelectric sensor placement on a high-fidelity finite element tire model. Second, a 1D CNN-LSTM network extracts spatio-temporal strain features to decouple three-directional forces. Third, a mechanistic brush model uses differential evolution to identify the friction coefficient, enforcing physical consistency. Simulations and test rig measurements validate the framework, achieving low root-mean-square errors below 0.04 kN, providing a real-time, hardware-efficient solution for autonomous driving safety.