Date of Award


Document Type

Open Access Dissertation


Civil and Environmental Engineering

First Advisor

Juan M. Caicedo


Constitutive models available in the literature typically describe the compressive strength of ordinary Portland cement paste (OPC) in a deterministic fashion, in some cases as a function of variables such as W/C, porosity, and the clinker chemical composition. The significant effect of the proportion of W/C on the final strength, porosity and density have been reported in several works in material science. Some studies have shown that the addition of MWCNTs can also modify the microstructure of cement significantly, and improve compressive strength and other mechanical and physical properties. In principle, W/C, MWCNTs addition, porosity, density, and compressive strength are related in different mechanical and physical degrees, which indicates that a compound multilevel model could include all these variables to predict multiple properties simultaneously. The development of mathematical models for properties such as the compressive strength has been intensively discussed in the literature. The construction of simple compressive strength models allowed for practical engineering applications. However, the deterministic models constructed to predict the compressive strength of the cement paste; as a function of the physical properties, mixing process and chemical composition of the cement paste, are not able to express uncertainty in the predicted values. In these cases, the measurement and modeling errors are ignored in the predictions. In this work, we proposed a Probabilistic Multilevel Constitutive Model (PMCM) using the variables mentioned. In multilevel modeling, nested sources of variability are considered in each level of the model, improving the precision of the different level iv predictors. Bayesian Inference was used to inform the model using experimental data from the literature and porosity characterization. Part of the clustered data was used to check the model goodness of fit in each group-level. Additionally, we proposed a Bayesian framework to analyze Mercury Intrusion Porosimetry data. The purpose of the approach was to incorporate all the uncertainty derived from sample preparation, drying methods, sample size, pore shape, contact angle, and surface tension assumptions