@article{01HXHF6VZSMH1ZXX46XTTQN7K6,
  abstract     = {{There are multibody systems whose physics are partially known owing to their complexity and nonlinearity. Therefore, motion equations are not utterly available to be utilized for the prediction, control, design, and monitoring of these systems. To alleviate this issue, this study aims at developing a hybrid modeling procedure to discover respective unidentified physics and, subsequently, provide a holistic governing model of the original mechanism. For approach development, a vehicle with unmodeled tires is thoroughly considered in this research work. Tires profoundly impact the dynamics of vehicles, influencing their handling, drivability, and ride comfort. Advanced chassis control systems used to improve vehicles' safety, performance, and reliability also require knowledge of tire behavior. Nevertheless, tires are very challenging to model as they are very complex and nonlinear components. Although simplified models are often employed, they are incapable of fully capturing tire behaviors. Using neural networks, i.e., black-box models, of the tire represents a common alternative. However, these approaches do not work outside the training data distribution, and they need costly and hard-to-measure experimental data for training purposes. Thus, this research study proposes a hybrid method by combining partially known physics of vehicle dynamics and a neural network to compensate for the unknown physics of tires. The developed approach learns the tire dynamics automatically from vehicle responses without requiring costly measured tire forces but solely relying on signals from an inertial measuring unit. The suggested methodology is validated experimentally, providing accurate and stable results. The time-depending behaviors of tires during cornering are also discovered and reported. The developed model is generic and can handle either linear or nonlinear physics-based models. However, the linear tire model integrated into the hybrid procedure in this study limits the simulation to stationary trajectories and cannot address the physics of tires when a vehicle undergoes nonstationary maneuvers.}},
  author       = {{Askari Renani, Ehsan and Gorgoretti, Davide and Crevecoeur, Guillaume}},
  issn         = {{1384-5640}},
  journal      = {{MULTIBODY SYSTEM DYNAMICS}},
  keywords     = {{Hybrid modeling,Multibody systems,Tire lateral forces,Physics-based,modeling,Data-driven modeling,Cornering stiffness,SIDESLIP ANGLE,FORCE ESTIMATION,NEURAL-NETWORKS,KALMAN FILTER,DYNAMICS,IDENTIFICATION,SYSTEM,ESTIMATOR,EQUATIONS,FRAMEWORK}},
  language     = {{eng}},
  title        = {{Hybrid modeling of multibody vehicles with partially known physics : discovering complex behaviors of tires}},
  url          = {{http://doi.org/10.1007/s11044-024-09983-3}},
  year         = {{2024}},
}

Altmetric

View in Altmetric

Web of Science

Times cited: