Open Collections

Abstract

The detection of diversions in residential gas pipelines is critical to maintaining public safety, preventing unauthorized usage, and ensuring the integrity of distribution networks. To minimize service disruptions, non-intrusive and non-destructive sensing methods are preferred. This thesis proposes a technique for the localization of diversions in high-density polyethylene (HDPE) pipes using ultrasonic sensors, signal processing, and a vision transformer (ViT) architecture. The approach specifically addresses the challenge of determining the distance from the measurement point to the diversion site. Zadoff-Chu (ZC) modulated ultrasonic signals are transmitted through the pipe using a custom-designed array of piezoelectric transmitters. Reflected signals, captured by a corresponding array of piezoelectric receivers, are processed and analyzed by a vision transformer to estimate diversion locations. Experimental validation carried out in buried pipeline setups demonstrates a localization accuracy of 11 cm under conditions that approximate real-world environments.