Evaluation of mechanical component fatigue behavior under random loads: Indirect frequency domain method

Abstract The fatigue behavior assessment for mechanical components subjected to random loads is traditionally conducted in time domain. An alternative frequency domain procedure is applicable under the hypothesis of stationary Gaussian stress state. The main objectives of this work consist in the description of the limits of applicability for the frequency domain methods and in the proposal of an original procedure (indirect method), that combines the advantages of the dynamic analysis conducted in the frequency domain, with the fatigue direct assessment criteria in time domain. Due to the stochastic nature of the fatigue damage under random stresses, a deterministic approach cannot be adopted, however, according to the common practice, a mean value of the damage can be assessed, paying attention that the time history used is sufficiently long. This makes sure the random error is negligible as it is of the same order of magnitude of the other approximations contained in the simulation process. In this paper the authors investigate under which hypothesis this method is valid and they propose a tool for the significance test of time data with regards to their capability in reproducing the right fatigue cycle amplitude distribution of the specific random process.