Preliminaries for probabilistic railway sleeper design

Licentiate thesis, 2006

Two investigations are presented, which can be seen as contributions to the aim of introducing a probabilistic design strategy of railway sleepers in the future. When using probabilistic design the statistics of all relevant parameters considered as stochastic must be known. One such is the loading environment of the railway sleeper that can best be seen as stochastic. For example there is a variation of the sleeper embedding both from site to site and in its distribution under each sleeper. We plan to perform tests in order to collect statistics of the latter. A pretest investigation including a numerical method for obtaining the maximum allowed sensor spacing for accurate sampling of this distribution is one of the investigations presented in this work. The numerical method is based on a hypothesis of equality between this sensor distance and a critical correlation length of the embedding stiffness. Before applying reliability analysis on the sleeper dynamics several aspects must be considered. One such is efficiency of the analysis method. The complexity, and thus efficiency, of the probabilistic design method that is required is basically determined by the properties of the performance function used for probability of failure estimation. An investigation, including both factorial design and one-factor-at-a time analyses, of the appearances of performance functions related to sleeper bending moments as functions of varying loading parameters is the second investigation presented in this work.