Todays components are faced with an increase of load cycles without the decrease of performance capabilities. Material fatigue can be described as continuous process originating from external an internal imperfections. The focus of many investigations is about application-oriented figures including the determination of different influences on the material fatigue process. The first part of this work (Publikation A, Publikation B) presents the analytical and numerical modelling of a new high frequency testing technique. The proposed setup showed a linear correlation between the mass acceleration and the induced stresses within the specimen cross section. Furthermore the setup is capable of fatigue testing up to 10^9 cycles to failure at frequencies of 930 ± 5Hz respectively 928 ± 8Hz. Long term tests showed a specimen temperature increase to T = 33 ± 0.5◦C enabling fatigue tests to reach 10^9 cycles within 12.8 days. Validation (Publikation B) of the testing technique was carried out at 928 ± 8Hz with a maraging steel X5CrNiCuNb16-4 (referred to as M-B). High frequency fatigue results were compared with conventional servo-hydraulic tests at 30Hz. Fatigue tests at 928 ± 8Hz showed a disproportionate increase of the fatigue strength. In addition to that a clear dependence could be determined regarding internal crack initiation. So called fish-eye fractures could be found in a wide scatter band from 10^6 - 10^9 cycles to failure. Therefore defect based analyses (Publikation C) were carried out concerning M-B material. Internal crack initiation respectively fish-eye fractures mainly emanated from non-metallic inclusions, Al2O3 and its chemical complexes. Specimen geometries with test diameters of D 4 = 4mm and D 7.5 = 7.5mm showed defect sizes of ( √ area) 28.98µm and 10.98µm. Additional fatigue tests at T = 350◦C (D 7.5 = 7.5mm) by contrast introduced slip band (subsurface) induced crack initiation. In this case a multistage Wöhler curve is suggested (Publikation D) due to the high scatter between 10 6 - 10 9 cycles to failure. The fatigue life of smaller volumes is underestimated. Various simulations on the specimen size and the stress gradient approach (Publikation E and Publikation F) showed a good applicability regarding rotating bending fatigue tests. The modification of the initial Wöhler curve was confirmed by fatigue experiments. On the other hand it can be said that lifetime calculations of higher stress gradients were less conservative.