Design of welded structures working under random loading

This paper presents a design method for predicting the fatigue life of T-joint assemblies loaded by random loads, based on a statistical analysis of tests. This study was based on the correlation between the types of loading observed in practice and test results obtained for fatigue life determination. The work follows three steps: analysis of the statistical distributions of random loads that illustrate extreme values from Markov matrix representations; statistical analysis of lifetimes obtained when the specimens are submitted to random loads defined earlier; design of a set of endurance curves [stress-number of cycles to failure (S-N) curves], called 'random' S-N curves. These S-N curves are shifted compared with that obtained under sinusoidal loading. Random S-N curve positions in the S-N plane are optimized depending on the lifetime for the random tests. A Basquin S-N model is introduced in order to represent S N curves. This model is able to take into account the damage due to the small cycles that are often present in actual loading. The use of random S-N curves for fatigue life calculations gives results matching with the real fatigue lifetime obtained with a T-joint assembly of a bogie chassis used for railway applications. Different analyses show the robustness of the proposed approach.