Evaluation of Stiffness and Tensile Strength in Hardened Cement Paste Using Bayesian Updating

Abstract The objective of this study is to propose a framework to evaluate the mechanical properties of hardened cement paste, such as stiffness and tensile strength, using Bayesian updating. To supplement time-consuming experiments conducted for the evaluation of material properties, virtual experiments conducted using simulations have been found to be effective and promising. However, even with the help of simulations, accurate and reliable evaluation of the mechanical properties of cement paste is challenging. Owing to the heterogeneous and complex nature of the microstructure of cement paste, considerable time and effort are required to reduce the uncertainties in its responses. The need for a large number of samples from the frequentist approach induces uncertainties and precludes accurate probabilistic assessments. Therefore, this study proposes a framework for a Bayesian approach to assess the mechanical properties of hardened cement pastes using a reduced number of data sets. Mechanical properties evaluated from actual microstructures obtained using micro-computed tomography were adopted as priors, and properties from reconstructed microstructures using generative artificial intelligence were selected as the observed data set. The conjugate prior distribution was then updated using the observed data set, resulting in a posterior distribution. From the updated statistical measures (mean and standard deviation) of the stiffness and tensile strength, it was found that the mechanical properties of cement paste could be estimated with reduced uncertainty using the Bayesian framework.