Facilitating detection and sizing of crack defects in pipes by 3D K-means clustering
Date
2024
Authors
Mazinani, Fatemeh
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis presents a novel approach for detection and sizing of surface-breaking crack defects in pipes using 3D K-Means clustering of ultrasound imaging data. The proposed method processes volumetric ultrasound data (obtained from a moving transducer array inside a pipe) to identify distinct clusters, effectively reducing noise and isolating critical crack-related features. Experimental validation has been performed on three pipe samples with different crack sizes and locations. The results show that 3D K-Means clustering improves crack detection and sizing, outperforming 2D K-means clustering in most cases. This research contributes to the field of ultrasonic nondestructive testing by providing an efficient solution for assessing the structural integrity of critical infrastructure components, such as pipelines.
Description
Keywords
k-means clustering