| Fatigue life estimation, reliability and durability are important in acquisition, maintenance and operation of vehicle systems. Fatigue is considered as one of the most important failure modes of a mechanical system. Fatigue life is random because of the stochastic load, the inherent variability of material properties and the uncertainty in the definition of the S-N curve. Degradation of the material properties of a system throughout time may cause unexpected fatigue failures that eventually increase the lifecycle costs due to warranty costs, repairs and loss of market share.;This dissertation has two main parts. In the first part, fatigue life prediction methods are investigated for linear and non-linear systems excited by Gaussian and non-Gaussian loading. For the latter, a general methodology to calculate the statistics of the output process considering the effects of skewness and kurtosis is used. Real operational conditions of ground vehicles involve non-Gaussian loading whose characterization is challenging. The excitation is first characterized using the first four moments (mean, variance, skewness and kurtosis) and a correlation structure. Then, the first four moments and the correlation structure of the response process are calculated using Polynomial Chaos Expansion (PCE) and Karhunen-Loeve (KL) expansion. Simulated trajectories from the response stochastic metamodel are rainflow counted to obtain realizations of the fatigue life random variable based on Miner's damage model. Finally, the Saddlepoint Approximation (SPA) method provides the PDF and percentiles of the fatigue life.;In the second part of this dissertation, we develop a new Accelerated Life Testing (ALT) methodology using Gaussian or non-Gaussian excitations without assuming the type of life distribution or the relationship between life and stress level. The accuracy of fatigue life prediction at nominal loading conditions is affected by model uncertainty (system model and fatigue model error) and material uncertainty such as the coefficients of the S-N curve. The uncertainty of fatigue life prediction is reduced by performing tests at higher loading levels. This reduces the test duration. We will develop an ALT methodology to minimize the cost of testing while improving the accuracy of fatigue life prediction. All dissertation developments will be demonstrated with representative examples. |
