Statistical modeling of thermal conductivity for cement-based foam

Abstract

This paper presents the formulation of two multivariate regression models for predicting the thermal conductivity of cement-based foam through statistical analysis. The cement-based foam was produced by replacing Portland cement with fly ash, silica fume, and metakaolin, up to 20% by mass, for densities that ranged from 400 to 800 kg/m3. In the first model, the porosity, moisture content, and thermal conductivity of the solid cement paste were considered as independent variables. Results reveal the presence of multicollinearity among these independent variables. In the second formulation, the porosity, mass ratio of the pozzolanic admixture in the binder, and maturity (in days) were taken as independent variables. This model was found statistically significant and was validated against a range of independently determined experimental values for thermal conductivity of such materials.