Parameter identification for a TGV model

This work investigates the applicability of identification methods to the suspension parameters of a TGV multi-body model. The aim is to adjust the model to the real system by estimating the suspension parameters from measured vehicle response data. Due to the nonlinear behavior of the system the time-domain based model updating has been chosen. It requires the definition and minimization of a misfit function in the time domain describing the distance between model and measurement. The fastest convergence is obtained by the use of gradient methods requiring the calculation of the derivatives of the misfit function relative to every parameter. Since the calculation from finite differences is time consuming and less accurate the gradients are calculated from the adjoint method. The application to a simplified bogie model with known mathematical description allows the identification of its suspension parameters. The presence of local minima in the misfit function of the TGV model requires the use of global optimization methods. The simulated annealing and the genetic algorithm method give important reductions of the misfit function and improved parameter estimations. In following work this information could be used for further applications like the condition monitoring.