Failure Detection in Oil and Gas Industry Pipelines Using Computer Vision and Convolutional Neural Networks

• Grant number: 24/16778-5
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: March 01, 2025
• End date: February 28, 2026
• Field of knowledge: Engineering - Mechanical Engineering
• Principal Investigator: Marcelo Becker
• Grantee: Giovanna Herculano Tormena
• Host Institution: Escola de Engenharia de São Carlos (EESC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

This project aims to develop a computational model for detecting failures in pipelines in the oil and gas industry, using computer vision and convolutional neural networks (CNNs). The oil industry constantly faces challenges in maintaining its infrastructure, and pipeline failures, such as cracks and corrosion, pose serious environmental and economic risks. Traditional inspection techniques, such as manual inspection and non-destructive testing (NDT), have significant limitations, including reliance on human interpretation, difficult access, and lengthy surface preparation.The project proposes replacing these traditional methods with a robotic inspection system equipped with cameras and employing convolutional networks capable of identifying complex patterns in large datasets. The dataset used to train, validate, and test the model will be acquired by the BIKE robot, which specializes in pipeline inspection, and the image database is owned by Petrobras. The system to be developed aims to increase the efficiency and safety of inspections, minimizing the risks of unexpected failures and associated costs.