Abstract
The increase in electronic and mechatronic content impacts not only the vehicle product itself, but also the way the design and development process has to change to enable widespread market introduction in standard vehicles. Considering the current implementation of increasingly complex sub-systems the control engineer is facing the challenge of identifying the system under study to implement an accurate model based control. The benefits of a structured modeling technique, founded on conceptual models and using on-line parameter estimation algorithms is therefore crucial, pointing out a key element in the usage of numerical simulations, with particular focus on the conceptual (1D) modeling.
This paper proposes an innovative methodology for parameters identification in vehicle dynamics, aiming at deriving reduced models that can be used for functional analyses. The key challenge is that the models must be both of limited size and sufficiently accurate at the same time. An optimal balance has been achieved by the enhancing a small-sized model with a parallel Neural Network structure, which can be trained to target the simple model’s capability to reproduce the complex dynamics of the real system. A detailed 3D vehicle dynamics benchmark case has been defined, namely a multi-body model in LMS Virtual.Lab Motion. In the 1D multiphysics environment of LMS Imagine.Lab AMESim, a mathematical vehicle representation has then been adopted, tested and functionally correlated with the reference. For the reference case, it has been shown that a small-sized 1D functional simulation model can be achieved by applying the parameter estimation procedure together with a parallel implementation of a Neural Network structure.
KEYWORDS: Vehicle Dynamics, Concept Modeling, Multibody Simulation, Parameter Estimation, System Identification