Quantification of cracks in concrete thin sections considering current methods of image analysis

Prior research has shown that manual crack quantification in concrete thin sections is time-consuming and subjective. It was already known that epoxy impregnation improves contrast in thin sections for imaging. No prior work had resolved how to automate crack measurement with machine learning. This gap motivated the need for a reproducible workflow. Existing methods rely on manual measurements, which lack consistency. The uncertainty in crack width distribution analysis remains unresolved. Researchers have not yet integrated Python-based automation for crack quantification. This paper introduces a novel approach to address these limitations.

The aim of this work is to develop an automated workflow for quantifying cracks in concrete thin sections. The specific problem involves the lack of objective and efficient methods for crack analysis. The motivation comes from the need to assess freeze-thaw damage in concrete structures. The study seeks to improve accuracy and reduce subjectivity in crack measurements. It also aims to provide a width distribution plot for cracks. The workflow is designed to handle areas up to 40 cm². The study compares manual and automated methods for validation. It addresses the challenge of quantifying crack density in thin sections.

The study uses yellow epoxy resin to impregnate concrete thin sections. Image preprocessing steps enhance contrast between cracks and other phases. Machine learning algorithms are applied to detect and quantify cracks. Python scripts automate the workflow for crack analysis. Crack area, length, and width are estimated using image processing tools. Validation involves manual measurements for comparison. Two manual methods are used to assess accuracy. Crack density is calculated to evaluate inner degradation.

The workflow successfully quantifies cracks in up to 40 cm² of thin sections. Crack area, length, and width are estimated automatically. Pixelwise analysis provides a width distribution plot for cracks. The crack density is calculated to assess freeze-thaw damage. Manual measurements differ from automated results in crack length. The yellow epoxy resin improves contrast for image analysis. The workflow reduces subjectivity in crack quantification. Results show that automated methods can complement manual analysis.

The authors propose that automated workflows can improve crack quantification in concrete. They suggest that yellow epoxy resin enhances image contrast for analysis. The workflow provides a width distribution plot for cracks. Crack density is a useful metric for assessing degradation. Automated methods differ from manual measurements in crack length. The study supports the use of machine learning for crack analysis. The workflow handles large areas efficiently. The findings suggest that automation complements traditional methods.