Optimization of the Ballast Layer in High-Speed Railway Tracks with Genetic Algorithms

The present paper deals with multi-objective optimization of the ballast layer in high-speed railway tracks. The track model is a simplified one, with only two dimensions, following the work of Zhai et al. [1]: the rails are modeled as beams, and discrete springs, dampers and mass simulate the stiffness, damping and inertia of the underlying structure (namely the rail-pads, sleepers, ballast and subgrade). The model is implemented in the commercial explicit dynamic software LS-DYNA. The design space is formed by the ballast height and by parameters representing its mechanical properties, which in turn are a function of the ballast’s material granulometry, derived from contact mechanics. Since published work on railway track simplified models uses mostly parameters obtained from field experiments, various expressions are proposed to obtain all necessary parameters from the mechanical and geometrical properties of the ballast. They aim to provide values close to the experimental ones, while being based on the theoretical mechanical behavior of the involved media. The objective functions cover minimization of the maximum displacement, velocity and acceleration in the main structural elements, namely the rail (with implications in vehicle stability and passenger comfort) and the ballast (important for track degradation and maintenance), due to the passage of moving loads representative of various railway vehicles. The optimization process is accomplished by the implementation of genetic algorithms for multi-objective optimization (Fonseca and Fleming [2]). Each objective is optimized individually, and then in pairs and triplets to obtain Pareto frontiers and decision maps, respectively.