Updating the prior parameters of concrete compressive strength through Bayesian statistics for structural reliability assessment

Abstract

The stochastic model of concrete compressive strength is a critical aspect for the reliability-based design of new concrete structures. The distribution scatter of this variable can compromise the quality of structural reliability assessments. The existing stochastic models in Probabilistic Model Code – a reference document for structural engineering practitioners published by the Joint Committee on Structural Safety – are based on previous investigations conducted during the 1980’s, which might be outdated. This paper investigates the validity of these models by using an extensive database of measured strength values of ready-mixed concretes by means of Bayesian updating methods with the use of prior information. Furthermore, a mathematical procedure is proposed to update the stochastic models when new strength values are available. This investigation confirms that the stochastic model for the most common strength classes can be ameliorated. For most classes, the revision resulted in larger 5% quantile values and lower mean values than the ones from those early studies, which could be an indication of a more precise concrete production. Additionally, the results suggest that the updating might be sensitive to the sample configuration. Finally, recommendations are given to improve the newly-derived models by means of additional strength values.