Statistical analysis of defects within concrete under elevated temperatures based on SEM image

Abstract

Concrete is a multiphase composite material consisting of aggregates, mortar, and initial defects. This work aims to quantitatively study the distribution and evolution of initial defects under elevated temperatures. To minimize the effect of hydration process and thermal and endogenous shrinkage, concrete samples cured in water for 12 months is tested. In order to study the effect of external load on the initial distribution of defects within concrete, concrete samples without pre-loading and pre-applied with 30% of the failure load are selected as research objects. The evolution of defects within the concrete samples are characterized using scanning electron microscopy (SEM). By employing an image processing software, the two-dimensional configurations of defects are extracted from the backscattered images obtained by SEM. Each defect is identified and approximated using an optimal ellipse. Thereafter, the defects are automatically recognized and classified into microcracks and pores by utilizing the image processing software and MATLAB programming. Consequently, the distribution and propagation of microcracks and pores in concrete could be quantitatively and separately analyzed. Statistical analysis of area, perimeter, and fitting ellipse parameters (e.g., major axis, minor axis, coordinates of the center, axis orientation, aspect ratio, and solidity) can provide the distribution of initial defects within concrete. It is demonstrated that the area of microcracks consistently increases with temperature. Log-normal and normal distributions may be suitable for the distribution of aspect ratio and solidity of microcracks, respectively, for both concrete samples with and without pre-applied loading under all elevated temperatures (40, 105, 150, 200, and 250 ℃). However, the distributions of area, coordinates of the center, and axis orientation of microcracks may vary with the initial concrete loading condition and temperature. For the pre-loaded concrete sample, the distribution of parameters does not conform to any type of distributions. Moreover, for the sample without pre-loading, the distributions of area, coordinates of the center, and axis orientation of microcracks exhibit good fitting precisions with log-normal, normal, and normal distributions, respectively at lower temperatures (40–150 ℃); however, such relationships do not apply at higher temperatures (200 and 250 ℃). As reference for identifying the variations in microstructures, 150 ℃ is selected as the threshold temperature. The statistical approach employed in this study can lay the theoretical and experimental bases for determining the microstructure evolution in concrete. It can also be used for identifying the meso-defect and macro-defect and establishing their relationship with concrete properties.