Calibration of the numerical model of a freight railway vehicle based on experimental modal parameters

• Autor(a) principal: Ribeiro, Diogo
• Data de Publicação: 2022
• Outros Autores: Bragança, C., Costa, C., Jorge, P., Silva, R., Arêde, A., Calçada, R.
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.22/21736
• topic: Freight wagon; FE model; Dynamic tests; Model updating; Genetic algorithm

The simulation of the dynamic behavior of the train-track system is strongly dependent on the accuracy of the numerical models of the train and track subsystems. The use of calibrated numerical models of the railway vehicles, based on experimental data, enhances their ability to correctly reproduce the dynamic responses of the train under operational conditions. In this scope, studies involving the experimental calibration of freight wagon models are still scarce. This article aims to fill this gap by presenting an efficient methodology for the calibration of a numerical model of a freight railway wagon based on experimental modal parameters. A dynamic test was performed during the unloading operation of the train, adopting a dedicated approach which does not interfere with its tight operational schedule. From data collected during the dynamic test, five natural frequencies and mode shapes associated with rigid-body and flexural movements of the wagon platform were identified through the Enhanced Frequency-Domain Decomposition (EFDD) method. A detailed 3D finite-element (FE) model of the loaded freight wagon was developed, requiring precise knowledge of the vehicle design details which, in most situations, are difficult to obtain due to confidentiality reasons of the manufacturers. The model calibration was performed through an iterative method based on a genetic algorithm and allowed to obtain optimal values for seven numerical parameters related to the suspension’s stiffnesses and mass distribution. The stability of the parameters considering different initial populations demonstrated the robustness of the optimization algorithm. The average error of the natural frequencies decreased from 8.5% before calibration to 3.2% after calibration, and the average MAC values improved from 0.911 to 0.950, revealing a significant improvement of the initial numerical model.