In the first stages of the design of vehicle structures for crashworthiness, multicriteria optimization tools can be very useful in order to evaluate how conflicting requirements compete. However, crashworthiness problems are by nature time consuming, if the standard non-linear finite elements simulation tools are used. Simplified models based on multibody dynamics can be developed and used for crashworthiness simulations as proposed in this work. With these simplified models, the computational time can be drastically reduced and evolutionary or genetic algorithms, that require large numbers of simulations, can be used, as proposed. The vehicle or structures are described as a set of rigid or flexible bodies connected by non-linear springs. Different models are presented hereunder dealing with different needs or stages of the projects of railway vehicles, namely, I D models that are mass-spring models, without kinematic constraints developed for the simulation of train sets and 2D rigid-flexible bodies developed for the study of structures and energy absorption devices with planar motion. These models include a Hertz contact-impact model (2D) for the simulation of the contact between the bodies and barriers and Coulomb friction for the wheel-rail contact. These two simulation tools are linked with evolutionary optimization algorithms to the multicriteria design of train structures. The design methodology developed is general and allows evaluating different conflicting design functions together, such as accelerations or measures of the acceleration, deformations, energies, masses or velocities. Also and in particular for the simulation of collisions of train sets different scenarios can be considered, which corresponds to the evaluation of different models in a single iteration. Three examples are presented in order to illustrate the design methodology developed herein. A single criteria optimization problem of the setup of an experimental train crash test; a multicriteria optimization problem of the energy absorption devices of a train in a multi-scenario collision with I D models; and the multicriteria optimization of an energy absorption structure with a 2D flexible model. The methodology has been implemented two general-purpose multibody codes and any response parameter can be used as the design function or a design constraint. The design methodologies presented have been applied for crashworthiness problems; however, they are quite general and can be applied for other field related with multibody dynamics such as, mechanisms and vehicle dynamics. The main limitation of the methodologies presented is that is the complexity of the models increases in the case of 2D flexible models, it will not be possible to obtain the Pareto front in a reasonable time. However, the simulations required in each generation are independent and the efficiency can be improved with parallel processing.