Laboratorio de Pronóstico Inteligente y Sistemas Estructurales Ciber-físicos

Anticipating and characterizing the damage induced by fatigue loadings is a challenging problem in the composites science and technology. In contrast to metals, fatigue degradation of composites is presented since the initial stages of the process leading to a decrease in the mechanical performance. The literature covers a large number of fatigue models for composites, the majority of them are only valid in their experimental conditions. In this thesis a novel bayesian inverse strategy to reconstruct fatigue damage over lifetime is proposed. This model has been developed to be extensible to different material configurations and loading conditions, in a coherent statistical sense. Finally this result has led to a bayesian model class selection, by which it is possible to select the most plausible model parameterization. The proposed methodology has been validated against fatigue damage data. This bayesian framework has shown versatility to take into account all possible information about data, models and the relation between them. The updated information inserted into the reliability problem is shown to confer a way to considered the long term reliability without the need to make hypothesis for time to failure.