The accelerometer-based intelligent tire has gained focus in recent years for its ability to obtain both kinematics and dynamics-related information of the tire. This paper extends the previous steady-state applications of acceleration signals, which mainly estimate tire force, sideslip, and friction coefficient from the steady-state features of acceleration waveforms, to transient acceleration applications. By using the proposed tire mixed Euler–Lagrange rolling model, it is analytically demonstrated that tire rolling acceleration can be decomposed into steady-state and transient-state components from the perspective of kinematics. It is hard to analyze the transient-state component theoretically or split it from the measured signals on real road surfaces; thus, a learning-based algorithm is developed to automatically extract discriminative features without any physical models. With this method, essential information associated with tire transient acceleration could be inferred to help improve driving safety and performance.
As the application, tire wear identification with an artificial neural network is validated to be feasible based on complete acceleration signals. The prediction accuracy reaches 98.2% under different test conditions. The proposed acceleration formation mechanism is proved to be effective in explaining tire rolling acceleration as well as guiding to acquire vital information about the tire to improve vehicle safety and performance.