DescriptionFatigue, recognized as one of the main causes of failures of mechanical and electrical components, is a class of structural damage that occurs when material is exposed to cyclic application of stress with varying or constant amplitudes. It is necessary to develop suitable lifetime distributions to predict the reliability and useful life of components or systems which experience fatigue failure due to random or constant amplitude loading during their usage. Birnbaum and Saunders (1968) propose the standard Birnbaum-Saunders (SB-S) distribution which belongs to the normal distribution family to fit the fatigue failure data to a life distribution. This model has a single dominant crack which appears and grows as the structure experiences repeated load cycles up to the point that the crack is sufficiently long to cause failure. The underlying assumptions, derivations, probability density function (pdf), cumulative density function (cdf), hazard function, reliability function, characteristics and properties of SB-S distribution have been investigated. The hazard function of SB-S distribution is shown to be always upside-down, which is limited because it fails to cover other types of failure conditions. In this thesis, we overcome the limitations of the SB-S distribution in modeling different types of failure rates and generalize the SB-S to be generalized Birnbaum-Saunders (GB-S) distribution. We present the characteristics of the distribution and the sufficient and necessary conditions that enable it to model multiple failure conditions. We also verify that the GB-S distribution provides better fit to failure data and propose new methodologies for the estimation of the parameters of the GB-S distribution for different sample sizes and shape parameters. An accelerated life testing (ALT) model can utilize the accelerated failure data to predict the reliability and lifetime of components or systems under normal environments. In this research, we develop the GB-S based ALT models, and deal with the inference procedure. We compare the performance of GB-S accelerated model with other ALT models. We also develop the GB-S based accelerated life testing plans for reliability performance prediction since a properly designed ALT plan makes the reliability estimation and prediction more efficient.