Reliability analysis of mechanical components

During the design stage, realistic life time predictions for components help to reduce test cycles and allow for an optimization with respect to both production and repair costs (see e.g. Schuëller et al. 1996). Accurate predictions of the mean life time and its standard deviation can be obtained from a fracture mechanics approach that takes the uncertainties in the material parameters into account. In this paper, the fatigue life time has been estimated and compared with experimental results. Failure was due to crack growth at critical sections of welding seams. As experiments indicated, the number of cycles to unstable crack growth showed a large scatter. Also, the direction of unstable crack growth seemed to be random. The theoretical model aims at representing the distribution of the cycle number until unstable crack growth. First, stress time series were obtained from given load histories for typical load cases. These time series were further reduced to equivalent constant amplitude loading, which served as input for crack growth calculations. A random variable crack growth model with failure criteria based on the R6 curve was employed. By means of importance sampling, failure probabilities for different cycles numbers were computed, from which life time predictions were estimated. Comparison between experimental and simulated first and second order moments of the life time showed a very good agreement for all load cases.