The need to optimize inspection intervals of railway wheelsets involves, among other aspects, a reliable estimation of fatigue crack growth in the axles. With this in mind, the present paper focuses on the description of the considered axle made of A1N steel, and on some typical load spectra. Application of an opportune algorithm, derived from the dedicated soft-ware NASGRO®, made it possible to obtain life predictions in terms of cycles, and to discuss the influence of service loads. Life estimates were then analysed together with probability of detection curves of different non-destructive testing techniques in order to determine the total probability of observing, during periodical inspections, cracks or defects.