Influence of mesoscopic pore characteristics on the splitting-tensile strength of cellular concrete through deep-learning based image segmentation

Properties of cellular concrete are significantly influenced by the mesoscopic pore structures, but the porosity-property relationships are still unclear. This paper tried to understand the pore effect on the splitting-tensile strength of cellular concrete by the specially prepared specimens with controlled additions of porosity. A series of cellular concrete specimens with different mix design were carefully prepared and subjected to splitting-tensile tests at a specified loading rate of 1 mm/min. Then, an integrated framework was developed for recognizing and measuring meso-structure based on the deep-learning model. The meso-characteristics of pores on splitting surfaces directly related to the splitting-tensile strength were captured and quantified automatically, and the relationship between meso-characteristics of pores and splitting-tensile strength was discussed in details. It was found that the splitting-tensile strength decreases as the porosity or pore size increases, accompanied by the variability of other pore characteristics such as uniformity coefficient and fractal dimension. At last, pore size and surface porosity on the splitting surface were identified as the main factors for splitting-tensile strength, and an empirical model was suggested to predict it with the above main factors. The findings in this paper further improve the understanding of the influencing mechanism for pores on the splitting-tensile strength, which can be used to optimal design of cellular concrete.