Railway sleepers have important roles in the complex railway system. Due to different loading condition, poor maintenance of sleeper or bad quality of ballast, a random load distribution along the sleeper-ballast interface may occur. A sleeper design, and also the track system design, which do not consider the random load distribution, could influence the performance of the sleeper and even damage the whole railway system. Thus, a numerical static and dynamic analysis for a pre-stressed concrete mono-block railway sleeper is carried out using finite element method. The structural behaviour of a single sleeper subjected to a random sleeper-ballast interaction is studied in three steps. First, four typical scenarios of support condition for sleeper are discussed in numerical analysis. Second, large enough numerical results under different random support conditions are conducted. Finally, Neural Network methodology is used to study the performance of sleeper under a stochastic support condition. Results of vertical displacement of rail seat, tensile stress at midpoint and underneath rail seat are presented. Moreover, the worst support condition is also identified.