Dynamic Neural Networks for Multi-‐Body Simulation in Mechanical Systems
Tutor / Supervisor
J. Wiedemann, J. Neubeck
Student
Vives Pons, Esteve
Document type
Master thesis (pre-Bologna period)
Date
2010
rights
Open Access
Publisher
Universität Stuttgart
UPCommons
Abstract
This text deals with the simulation of the tyre/suspension dynamics by using recurrent
dynamic neural networks. Recurrent neural networks are based on the multilayer
feedforward neural networks, by adding feedback connections between output and input
layers. The neural network can be trained with data obtained from the simulation of a
physical model created using a multi-body simulation software (SIMPACK). The results
obtained from the neural network demonstrate a good agreement that could be improved, depending on some factors, with the multi-body model simulation results. The
neural network model can be applied as a part of vehicle system model to predict system dynamic behaviour. Although the neural network model does not provide a good insight of the physical behaviour of the system, it is a useful tool to help in vehicle ride dynamics performance due to its good efficiency and accuracy in computational terms.
Participating teacher
- J. Wiedemann, J. Neubeck