Sensitivity analysis of nonlinear railway vehicle models using linearized proxy analyses

International audienceVirtual prototyping can facilitate the design and certification of railway vehicles with the aim of ensuring passenger comfort and safety. Sensitivity analysis (SA) has proven to be an essential tool for ranking the most influential model parameters in order to effectively reduce the design space before implementing costly model selection procedures, for example design optimization, model calibration, or enhancement of critical component models. Meanwhile, the computational burden of performing global sensitivity analysis on the nonlinear transient simulations used in the railway industry can prove challenging. The present work investigates the potential of replacing these nonlinear simulations by a series of proxy modal analyses performed at different linearization points in the design space. The methodology is illustrated using the Morris design sensitivity method on an academic system: a two degree of freedom (DOF) system with a nonlinear stiffness. Results of a complete nonlinear sensitivity analysis are compared with the results based on the proxy linearized analyses