A data-driven suspension kinematics and compliance model considering multi-axis coupling effects

Abstract

The kinematics and compliance characteristics (K&C) significantly impact a vehicle’s dynamic response. While a comprehensive multibody suspension model can accurately express desired suspension K&C characteristics, the complexity of the model limits its use in real-time applications. The K&C characteristics are usually obtained through decoupled bench tests, whereby it is assumed that the suspension K&C characteristics can be orthogonally decomposed and that the function outputs can be added up to restore the original characteristics, which may not be accurate. Therefore, in this research, a novel suspension K&C simulation/test method is proposed to demonstrate the suspension characteristics under coupled force inputs. A multibody dynamic model of an air suspension is constructed and calibrated through bench tests. Then, a series of coupled tire forces with shapes of space spiral curves are applied to the multibody model to produce the coupled K&C data. A novel data-driven model based on Long Short-Term Memory (LSTM) network is proposed to predict suspension K&C. Finally, the proposed data-driven K&C model and the decoupled K&C model are compared under full vehicle simulations. The results suggest that the proposed method is more accurate than the decoupled K&C model and can express the suspension hysteresis characteristics without further modification.