Prediction of Discrete Element Breakage Parameter for Ballast Particles Based on Genetic Algorithm–Back Propagation Neural Network Model

To optimize the deterioration assessment and maintenance of ballasted tracks, it is of great value to study the breakage process and mechanism of ballast particles. Through a uniaxial breakage test on the single ballast particle, the equivalent stress required for its failure was determined. The deformation behavior under load was analyzed based on the ballast particle breakage process and loading force. Laser grating scanning of the ballast particle geometry was performed, and a minimum bounding rectangle method was used for specification. Rigid blocks were used for ballast particle packing, and a comparison was made with the traditional spherical particle packing method. The breakage process of ballast particles constructed with rigid blocks and the initiation of microcracks within the ballast particles were analyzed. In addition, the discrete element contact parameters for ballast particles with different geometries were studied, and a neural network model optimized by a genetic algorithm, namely GA-BP was used to predict the bond strength for ballast particles with different equivalent particle sizes. The results show that in the discrete element model, the bond strength of the ballast particles increases with the increase in its equivalent particle sizes. Specifically, for equivalent particle sizes in the ranges of [25, 39), [39, 48), [48, 56), [56, 64), and [64, 80) mm, the corresponding average bond strengths are 151.85, 159.45, 166.71, 175.29, and 185.29 MPa, respectively.