Improving the utility of MIP analysis for cementitious systems through Gaussian process regression modeling to predict electrical resistivity

Durability of concrete is a key design parameter and is strongly related to the penetrability of the cementitious binder matrix. This research evaluated methods to estimate the durability of concrete in service, using different procedures to characterize the penetrability of concrete, mortar, and paste. Mercury intrusion porosimetry (MIP) was compared to several standardized testing methods (surface resistivity, bulk resistivity, and bulk diffusion) on specimens made from cementitious paste, mixed mortar, sieved mortar, and concrete. MIP measurements were compared to results from standardized methods, with the goal of developing useful correlations. Gaussian process regression (GPR) modeling was applied to predict the electrical resistivity of concrete using MIP measurements obtained from mortar specimens, in combination with chemical oxide compositions of the mortar components. The resulting GPR model was compared to the Katz-Thompson method for correlating MIP data to electrical resistivity measurements. In this study, GPR modeling offered an improvement over the Katz Thompson method and was able to predict concrete surface resistivity using porosity data obtained from mortar, providing a potentially useful tool for rapid durability screening of supplementary cementitious materials.

Automated summary

This study evaluated methods to estimate concrete durability by analyzing penetrability, comparing MIP with standardized testing methods, and developing GPR models. The results showed that GPR modeling was more accurate than the Katz-Thompson method in predicting concrete surface resistivity, providing a potentially useful tool for rapid screening of supplementary cementitious materials.