Document Type
Extended Abstract
Abstract
Assessing water absorption, or 'sorptivity,' in cement-based materials is a crucial method for evaluating durability. The ASTM C1585 standard provides a straightforward approach for measuring sorptivity, yet it is often labor-intensive and time-consuming. This study presents two innovative strategies to accelerate and automate sorptivity measurements through computer vision. The first method involves a droplet test, demonstrating that the initial dynamics of a water droplet's contact angle, characterized by CNNs, strongly predict the initial sorptivity of paste samples. By analyzing data from 63 different paste systems, this method estimates the 6-hour initial sorptivity in mere seconds. The second method utilizes a vision-based algorithm to monitor the advancing waterfront in a sample, leveraging the wetted area ratio to predict sorptivity and streamline the measurement process. By training the algorithm on a dataset of over 6,000 images and 1,400 data points, the system achieves real-time predictions of initial and secondary sorptivities with an R² exceeding 0.9, closely aligning with ASTM standards. These cost-effective solutions, using cameras priced at around $30, provide laboratories worldwide with an accessible and automated method to enhance sorptivity testing, especially for assessing the durability of new low-carbon cementitious materials.
Keywords
Sorptivity Measurement, Durability, Computer Vision.
DOI
10.5703/1288284318006
Computer Vision-Driven Sorptivity Tests for Cementitious Materials
Assessing water absorption, or 'sorptivity,' in cement-based materials is a crucial method for evaluating durability. The ASTM C1585 standard provides a straightforward approach for measuring sorptivity, yet it is often labor-intensive and time-consuming. This study presents two innovative strategies to accelerate and automate sorptivity measurements through computer vision. The first method involves a droplet test, demonstrating that the initial dynamics of a water droplet's contact angle, characterized by CNNs, strongly predict the initial sorptivity of paste samples. By analyzing data from 63 different paste systems, this method estimates the 6-hour initial sorptivity in mere seconds. The second method utilizes a vision-based algorithm to monitor the advancing waterfront in a sample, leveraging the wetted area ratio to predict sorptivity and streamline the measurement process. By training the algorithm on a dataset of over 6,000 images and 1,400 data points, the system achieves real-time predictions of initial and secondary sorptivities with an R² exceeding 0.9, closely aligning with ASTM standards. These cost-effective solutions, using cameras priced at around $30, provide laboratories worldwide with an accessible and automated method to enhance sorptivity testing, especially for assessing the durability of new low-carbon cementitious materials.