Parameter identification of multi-body railway vehicle models – Application of the adjoint state approach

• Parameter identification for nonlinear multi-body railway vehicle model is proposed.
• The misfit function gradient is computed from the adjoint state approach.
• Efficient and accurate gradient computation relative to many vehicle parameters.
• Different signal types in the misfit function are analysed.
• Measurement errors are considered in the gradient computation and parameter identification.

For the calibration of multi-body models of railway vehicles, the identification of the model parameters from on-track measurement is required. This involves the solution of an inverse problem by minimising the misfit function which describes the distance between model and measurement using optimisation methods. The application of gradient-based optimisation methods is advantageous but necessitates an efficient approach for the computation of the gradients considering the large number of model parameters and the costly evaluation of the forward model.This work shows that the application of the adjoint state approach to the nonlinear vehicle–track multi-body system is suitable, reducing on the one hand the computational cost and increasing on the other hand the precision of the gradients. Gradients from the adjoint state method are computed for vehicle models and validated taking into account measurement noise.