Probabilistic Comparative Assessment of Structural Cement Composite Compressive Strength-Porosity Models

Compressive strength models for ordinary portland cement composites are typically deterministic and rely on parameters such as water-to-cement ratio, porosity, and chemical composition. While these models can estimate compressive strength, they often lack the ability to quantify uncertainty in their predictions. Various factors contribute to this uncertainty, including diverse testing protocols, limited data, overly complex mathematical formulations, and constraints in capturing the variability of the phenomenon through model parameters. This study introduces a comparative evaluation of representative compressive strength-porosity models for cement paste, adopting a probabilistic perspective and using data from various testing protocols found in the literature. Experimental data encompassing compressive strength and porosity, spanning porosity values from 1% to 30%, were gathered from hot-pressed and conventionally cast specimens. Results were obtained following the Bayesian Analysis Reporting Guidelines (BARG) with combined decision criteria using a 95% High Posterior Density (HPD) interval and the Range of Practical Equivalence (ROPE). A probabilistic version of Ryshkewitch's model emerges as the most plausible based on porosity data. These results hold promise for integration into risk analysis and decision-making tailored for structural analysis and design. Ultimately, this research contributes insights into enhancing the reliability of compressive strength predictions for cement composites by accounting for the inherent uncertainty of the underlying models.