ANN-based bubble tracking algorithm for clay slurries containing large gas bubbles using X-ray CT

Abstract

Gassy clay is a widely-distributed natural composite material consisting of a saturated clay matrix incorporating large gas bubbles. This study aims to develop a novel method to non-destructively monitor the strain field evolution of drying kaolinite slurry samples, with entrained gas bubbles, using X-ray computed tomography (CT). During the drying process, the kaolinite sample is scanned several times at decreasing water content (reducing soil volume. Individual bubbles are identified within the kaolinite sample and their morphology and location are measured via image processing. A bubble-tracking algorithm links associated bubbles between different scans using bubble location and morphology information. The morphology of individual bubbles changes as the soil volume reduces under increased suction during the drying process. Therefore, an artificial neural network (ANN) is adopted to enhance the bubble-tracking algorithm. This method shows great accuracy based on verification with manually linked bubbles. Hence, the three-dimensional (3D) displacement field of bubbles was determined within the soil sample. Finally, the evolution of the 3D displacement field is discussed, both before and after desiccation crack formation.