The present thesis aims at advancing an innovative computational methodology that simulates steel and composite material fracture under cyclic loading following a phenomenological approach, with calibration from both small scale and large scale testing. This work addresses fatigue processes ranging from high cycle to ultra-low-cycle fatigue. An assessment of the current state of the art is done for all the different fatigue types. Following, for ultra-low cycle fatigue a new constitutive law is proposed and validated with experimental results obtained on small scale samples. Industrial applications are shown for a large diameter straight pipe under monotonic loading conditions and for a bent pipe under cyclic loading. Emphasis is made on the capacity of the model to represent different failure modes depending on the loading conditions. The research regarding this part has been used in the frame of the European Project: ¿Ultra low cycle fatigue of steel under cyclic high-strain loading conditions¿ (ULCF). Regarding high cycle fatigue, a classic damage model is presented in combination with an automatic load advancing strategy that saves computational time when dealing with load histories of millions of cycles. Numerical examples are shown in order to demonstrate the capabilities of the advancing strategy and a validation of the model is done on small scale samples. A new constitutive model is presented for Low Cycle Fatigue that uses the classic plasticity and damage theories and simultaneously integrates both processes in the softening regime. The capabilities of the model are shown in numerical examples. Finally, the high cycle fatigue damage model is applied to the constituents of a composite material and the structural behaviour is obtained by means of the serial/parallel rule of mixtures. Validation of the constitutive formulation is done on pultruded glass fiber reinforced polymer profiles.