Stochastic modeling and statistical analysis of fatigue tests on prestressed concrete beams under cyclic loadings

Abstract

To evaluate the fatigue behavior of prestressed concrete, bridges for instance, it is necessary to determine the built in tendons’ fatigue strength. Therefore, prestressing steel samples (strands), obtained from an existing bridge built in 1957, were examined and tested by TU Dortmund University, see [1, 2]. Additionally, similar prestressing steels were tested in comparable experiments. As large experiments on prestressed concrete beams under cyclic load with small stress range are very time-consuming and expensive, an early prediction of failure trend in the experiment is desirable. Here, it is shown that a crack width function can be evolved dependent on the process of single wire failures. This process will differ for each experiment because of the randomness of single wire failure. Description of this uncertainty is the first step and is achieved by a predictive distribution for the counting process of wire failure. The second step is to include these results into the model for the crack width process for which a nonlinear regression model based on a physically evolved function depending on the counting process is suitable. For both modeling steps, we present a Bayesian estimation and prediction procedure.