Percussion Based Detection Method for Localization of Pipe Inspection Gauge using Advanced Machine Learning Classification and Clustering Techniques.

Pipelines carry the blood upon which our 21st century is built. They are used all over the world, transporting oil, natural gas, water, and other valuable liquids allowing for basic necessities and travel. However, pipeline infrastructure on the decline from old age and constant use, maintenance and mitigation has emerged as a necessity to upkeep the piping. In recent decades, pipe inspection gauges (PIGs) have emerged as key tools which can flow through a pipe to conduct cleaning or inspection tasks. These PIGs can span up to multiple feet in length, and due to varying conditions in the pipe from construction shape to contaminant blockage, PIGs can become stuck themselves in pipelines leading to major downtime and contingency operations to remove the PIG. This can be extremely costly and time consuming. Various PIG localization techniques have been invented, using a whole suite of advanced instruments, from global positioning to magnetic based tracking to percussion based. Percussion based detection in particular stands out due to its ability to be done in all pipeline environments and its simplicity and cost effective nature. With no current methods in the industry currently to track stuck PIGs in pipelines, this thesis focused on using percussion based detection and machine learning methods to localize missing PIGs. From a simple strike on a pipe system, this thesis compared advanced supervised and unsupervised machine learning techniques (Support Vector Machine, Convolutional Neural Network + Long-Short Term Memory Network, Feedforward Neural Network, gradient boosting, Gaussian Mixed Model and K-means clustering) to find cost effective localization techniques. Results showed that MFCC feature extraction and Convolutional Neural Network + Long-Short Term Memory Network and Gaussian Mixed Model clustering techniques were able to best classify at high accuracy missing pipe inspection gauges in an experimental pipeline system.

Name thesis:degree_name

Master of Science

Discipline thesis:degree_discipline

Aerospace Engineering

Grantor

University of Houston

Year dc:date.issued

2025

Author dc:creator

• Cohen, Carter Patrick

Advisor dc:contributor.advisor

• Song, Gangbing

Committee members dc:contributor.committeemember

• Grigoriadis, Karolos
• Ryou, Jae-Hyun

• Aerospace engineering

Language dc:language.iso

English

Handle dc:identifier.uri

https://hdl.handle.net/10657/20634

OAI identifier oai:identifier

oai:uh-ir.tdl.org:10657/20634