Machine Learning Techniques for Surface Defect and Anomaly Detection in Steel Sheets: A Hybrid Approach using Xception and Random Forest

Dour, Ramit (2024) Machine Learning Techniques for Surface Defect and Anomaly Detection in Steel Sheets: A Hybrid Approach using Xception and Random Forest. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (2MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

This study investigates the application of multiple machine learning and image segmentation models for automated surface defect detection on steel sheets. Datasets such as the NEU Surface Defect Database, the Severstal Steel Defect Dataset, and KSDD2 are used for training. Traditional manual inspection methods for detecting surface defects like scratches, dents, and marks are laborious and can lead to errors, highlighting the need for automated solutions in industries like automotive, electrical appliances, and electronics. The research evaluates various deep learning models, including custom Convolutional Neural Networks (CNN), ResNet50, InceptionV3, EfficientNetB0, VGG19, Xception, and U-Net, to identify the most effective approach for defect detection. Among these, the hybrid model combining Xception for feature extraction and Random Forest for classification achieved the highest test accuracy of 82.22%, with a precision of 0.8967, making it the most accurate model in this study. Additionally, the Segment Anything Model (SAM) was evaluated for its segmentation capabilities, achieving a Dice coefficient of 0.72 on the validation set. These findings contribute to the development of scalable and reliable deep learning-based defect detection systems that can significantly enhance product output quality in production by reducing dependency on manual inspection.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Simiscuka, Anderson UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning H Social Sciences > HD Industries. Land use. Labor > Specific Industries > Manufacturing Industry
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	15 Aug 2025 17:46
Last Modified:	15 Aug 2025 17:46
URI:	https://norma.ncirl.ie/id/eprint/8555

Actions (login required)

View Item