On the Derivation of Design S-N Curves Based on Limited Fatigue Test Data

Qualification of new characteristic S-N curves for fatigue life assessment of structures is considered to be a significant engineering challenge. First, representative fatigue test data for the actual structural connections have to be derived. Then these test data have to be transferred into characteristic S-N curves that represent a predefined probability of survival. Characteristic S-N curves are also often denoted design S-N curves as these curves are often used directly for fatigue life assessment of structures without application of a material factor. A few large scale tests can add significant confidence to a design S-N curve dependent on the type of structural detail to be designed. The reason for this is that a prototype test specimen can be fabricated in a similar way as the actual connection and it is similar in geometry, material characteristics, residual stress, and fabrication tolerances. In addition it can likely be subjected to a more relevant loading and boundary conditions as compared with that of small scale test specimens. When a limited number of test data are available, it is questioned how a characteristic S-N curve can be derived with a well defined probability of survival. The mentioned issues are further considered in this paper together with some recommendations on how to derive design S-N curves based on limited data.