This paper presents an efficient numerical tool for the prediction of railway dynamic response. A behavior calibration of the infinite Euler-Bernoulli beam resting on continuous viscoelastic foundation is proposed. Constitutive laws of the discrete elements are determined for a rectilinear ballasted track. A three-dimensional model coupled with an adaptive meshing scheme is employed to calibrate the beam model impedances by finding the similarity between the output signals using the genetic algorithm. The model shows an important performance with significant reduction in computational effort. This study emphasizes the major impact of the excitation characteristics on the parameters of the discrete models.