Since immanent defects are considered as the primary cause of component fatigue failure, in this research work the effects of defects on the fatigue behaviour of cast aluminium are investigated. Therefore, crankcases of an Al-Si-Cu alloy, grade EN-AC 46200, are evaluated in terms of metallographic, quasi-static, fracture mechanical and fatigue strength. The specimens are extracted from a priori defined sampling positions in order to diminish technological size effects due to varying microstructural properties. To force defect-based fatigue failure, the specimen geometries are numerically shape-optimized. Thus, an uniform stress distribution within the testing region is ensured, such that the most critical heterogeneity eventually leads to fatigue crack initiation, cyclic growth and subsequently to burst failure. Moreover, the length of the constant testing diameter is varied. Therefore, the distinctive difference in the highly-stressed volume between the specimen types enables the evaluation of a statistical size effect. Common defect-based fatigue strength models are enhanced, featuring their applicability in both the finite-life region and long-life region. Additionally, the effect of defects taking into account in-service temperatures is investigated. By non-destructive evaluation of immanent heterogeneities, the critical flaw size distribution of both the investigated component, as well as the mass-production correlated extreme value distribution, may be derived, subsequently acting as statistically distributed equivalent intrinsic crack size. Furthermore, extensive investigations of the load ratio dependent crack-closure effects are conducted in order to statistically evaluate the cyclic crack resistance curve. In addition, the stress-based mean stress sensitivity is expanded such that the raised fatigue assessment approach is valid for both imperfect and near-defect free cast material. By setting up a bivariate load distribution of critical defect sizes and local load stresses, in line with the statistically characterised crack threshold and fatigue strength, a statistical, fracture-mechanical-based fatigue assessment approach is introduced. Finally, a novel probabilistic fatigue assessment map is developed, in order to enable fatigue design towards utmost probabilities of survival, additionally taking the highly-stressed volume of a critical component into account. The developed approach is validated by means of experimental data, along with state-of-the-art models. The verification procedure reveals that the improved fatigue map matches the experimental data well, whereat commonly applied engineering approaches may even lead to non-conservative design of cast components.