A non-central version of the Birnbaum-Saunders distribution for reliability analysis

The Birnbaum-Saunders distribution has largely been applied to material fatigue and reliability studies to relate the time until failure to some type of cumulative damage. This damage is produced by the growth of a dominant crack in material specimens, the propagation of which is due to cyclic patterns of stress. However, the Birnbaum-Saunders model was constructed under restrictive conditions that may not be valid for certain applications. In particular, this model was developed assuming that the crack extension in each cycle has a constant s-mean over time. In this article, we assume the crack extensions in each cycle have a non-constant s-mean. This leads us to introduce a non-central version of the Birnbaum-Saunders distribution. Specifically, a comprehensive description of the properties and characteristics of this new model is presented. The suitability of this non-central distribution is shown by using real, and simulated data.