Fatigue Life Prediction in Cementitious Materials Through Bayesian Probabilistic Models

Abstract

First, we present a probabilistic framework for analyzing fatigue in cementitious materials, particularly concrete, within a Bayesian context. This is realized by incorporating the Sparks–Menzies relation, which correlates fatigue life to the secondary strain rate, into a probabilistic model. This study emphasizes the transition from deterministic to probabilistic methodologies, enhancing the prediction and understanding of fatigue behavior under varying load conditions. Next, we reformulate a strain-based failure criterion, previously deterministic, into a probabilistic model that better captures the inherent variability of material properties. The Bayesian model estimates parameters through probability density functions rather than fixed coefficients, leveraging extensive experimental data. Finally, we extract a generic Sparks–Menzies relation for all the cementitious materials studied, thus a strain-based criterion for their fatigue life prediction, with obtained probabilistic density functions. Such a customized application of the Bayesian analysis represents a rather novel approach for fatigue prediction and facilitates the early identification of potential failure points during the secondary phase of fatigue testing.