Machine Learning-Driven Corrosion Detection and Classification in Pipelines

Pipeline corrosion is a critical issue in the petrochemical industry, with significant implications for both operational safety and economic efficiency. This research focuses on developing a robust system for detecting and classifying pipeline corrosion using advanced signal processing techniques and machine learning algorithms. The study is grounded in a positivist research philosophy, employing a deductive approach to apply established theories in a real-world context. Data for this research was collected from a project conducted by a petrochemical company in China, where Fiber Bragg Grating (FBG) sensors were deployed along an oil pipeline to monitor vibrations caused by liquid flow. These vibrations were analyzed to detect potential corrosion. The raw sensor data, often noisy due to environmental factors, was processed using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Bhattacharyya Variance Distance (BVD) algorithms to enhance signal quality. Key features indicative of pipeline condition were extracted and subjected to clustering analysis using the K-means algorithm, categorizing the data into distinct groups representing different levels of corrosion severity. Subsequently, classification models, including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and XGBoost, were applied to predict corrosion severity. The XGBoost model demonstrated superior performance, achieving perfect accuracy, precision, recall, and F1 scores. The study also addresses the ethical considerations of data privacy and the responsible application of machine learning in industrial settings. The findings highlight the effectiveness of the proposed methodologies in accurately detecting and classifying pipeline corrosion, offering significant potential for improving pipeline maintenance and safety. Recommendations for further research include increasing the dataset size and exploring additional or hybrid models to further enhance system accuracy.

Name dc:type.qualificationname

msc

Level dc:type.qualificationlevel

masters

Grantor dc:publisher.institution

University of Wales Trinity Saint David

Year dc:date.issued

2024

Author dc:creator

• Liu, Liwei

• TK Peirianneg drydanol. Peirianneg Niwclear Electronig

Dc Identifier Grantnumber

UWTSD

OAI identifier oai:identifier

oai:repository.uwtsd.ac.uk:3305