Prediction of wheelset axle loading from operating conditions using statistical modelling

Fracture mechanics is a technical discipline that assumes the presence of defects in the material and studies their propagation under loading. It allows to mathematically estimate the relevance of a defect and its period of stable propagation before catastrophic failure. Crack propagation calculations are therefore an established engineering tool to ensure operational safety of components. This approach is often called “damage-tolerant design”. In the railway industry, particular attention is paid to the wheelset axle: a component that in the past has been object of catastrophic failures due to unexpected critical crack propagation. The goal of crack propagation calculations is ensuring sufficiently short non-destructive testing (NDT) intervals to detect the presence of potentially dangerous cracks before they become critical. Crack propagation is heavily dependent on the load spectra the component is subjected to during service. Higher stress amplitudes cause faster crack propagation than lower stress amplitudes. The proportion of high and low stress amplitudes and their order are a deciding factor of the stable crack propagation. Knowledge of the load history is therefore essential to obtain a realistic crack propagation prediction. In practice, recording vehicle-, service- and track- specific loads spectra using measurements is a time-consuming and expensive endeavour. Additionally, the information about the loads is not available during the development phase of a vehicle and can only be determined in retrospect. With growing demands by operators and National Safety Boards on the specificity of crack propagation calculations and the length of NDT intervals to reduce maintenance costs, methods to determine realistic axle loads spectra during the development phase are needed. This research aimed to develop a statistical model to estimate wheelset axle bending loads as a function of track and service parameters, with particular focus on cost competitiveness, ease of use, accuracy and consistency. The model was built based on measurement data gathered during dedicated on-track tests using the method of linear regression.