Simulation of the mechanical behavior of railway ballast by intelligent computing

Abstract

Ballast is one of the main components of railway track foundations, thus, an accurate prediction of its mechanical behavior is crucial for stability of railway tracks. In this paper, one of the most commonly used intelligent computing techniques, i.e. Artificial Neural Networks (ANNs), is utilized to model the mechanical behavior of ballast under static loading conditions. Experimental results from a series of large-scale consolidated drained triaxial compression tests collected from the literature are used for ANN model calibration and validation. The results indicate that predictions from the ANN model compare well with those obtained from the large-scale experiments. In particular, ANN predictions demonstrate a high degree of accuracy in simulating the stress-strain and volume change characteristics of ballast. The plastic dilation and contraction of ballast at various confining pressures, and the strain-hardening and post-peak strain-softening are also well simulated.