Steel Surface Parameterisation for On-line Quantification with Machine Learning Models

MILNE, ALEXANDER

doi:10.23889/SUThesis.71779

E-Thesis 35 views 12 downloads

Steel Surface Parameterisation for On-line Quantification with Machine Learning Models / ALEXANDER MILNE

Swansea University Author: ALEXANDER MILNE

PDF | E-Thesis – open access

Copyright: the author, Alex Milne, 2026, Distributed under the terms of a Creative Commons Attribution NonCommercial 4.0 License (CC BY-NC 4.0)
Download (19.97MB)

DOI (Published version): 10.23889/SUThesis.71779

Abstract

Accurately estimating steel surface roughness parameters in real-time during industrial manufacturing is a critical challenge. Traditional post-production stylus-based measurements of surface roughness, speciﬁcally the arithmetic mean roughness (Ra), are slow and sample only small sections of a stee...

Full description

Published:	Swansea 2026
Institution:	Swansea University
Degree level:	Doctoral
Degree name:	Ph.D
Supervisor:	Xie, X., and Tam, G.
URI:	https://cronfa.swan.ac.uk/Record/cronfa71779

first_indexed	2026-04-22T12:12:39Z
last_indexed	2026-04-24T04:17:01Z
id	cronfa71779
recordtype	RisThesis
fullrecord	<?xml version="1.0"?><rfc1807><datestamp>2026-04-22T13:20:41.4951365</datestamp><bib-version>v2</bib-version><id>71779</id><entry>2026-04-22</entry><title>Steel Surface Parameterisation for On-line Quantification with Machine Learning Models</title><swanseaauthors><author><sid>c6da9da5c99754b6850e882895b86ca5</sid><firstname>ALEXANDER</firstname><surname>MILNE</surname><name>ALEXANDER MILNE</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2026-04-22</date><abstract>Accurately estimating steel surface roughness parameters in real-time during industrial manufacturing is a critical challenge. Traditional post-production stylus-based measurements of surface roughness, speciﬁcally the arithmetic mean roughness (Ra), are slow and sample only small sections of a steel coil, leading to production inefﬁciencies, costly rework or downgrading of coils, and potential downstream delays in demanding just-in-time manufacturing environments due to re-production when quality is not met. While the on-line, non-contact laser reﬂection measurement device used by our partners offers a continuous monitoring solution, its accuracy is often insufﬁcient compared to stylus-based ground truth measurements, negating its utility for process control.This research explores the application of machine learning (ML) to enhance the accuracy of surface roughness estimation from raw laser reﬂection data. By leveraging both limited la-belled and abundant unlabelled industrial datasets, this work investigates various ML-driven methods. These include: data imputation techniques aimed at improving existing calculation methods by addressing data quality issues; end-to-end models designed for direct mapping from sensor input to the stylus-equivalent Ra value; efﬁciency enhancements for promising model architectures; and the incorporation of representation learning via self-supervised learning (SSL) to effectively utilise unlabelled data and overcome labelled data scarcity.The ﬁndings demonstrate the substantial advantages of ML approaches over conventional methods. End-to-end deep learning models consistently achieve signiﬁcantly higher accuracy in predicting Ra compared to the baseline sensor calculations. Our comprehensive comparative studies reveal that architectures such as Temporal Convolutional Networks (TCNs) and certain 2D Convolutional Neural Networks provide a robust combination of high predictive ac-curacy and computational feasibility for this speciﬁc industrial data type. Furthermore, lever-aging large unlabelled datasets via pretraining proved highly effective; representation learning strategies, particularly a novel task-speciﬁc coil classiﬁcation pretext task and a TCN-based autoencoder, yielded models with enhanced generalisation and accuracy upon ﬁne-tuning with limited labelled data. Methodological insights also include the necessity of a statistics-aware loss function for imputation tasks requiring the preservation of surface texture characteristics, and the successful GPU acceleration of the ROCKET model for enhanced throughput.The key contributions of this thesis therefore encompass not only the successful application and comparison of various ML techniques (imputation; end-to-end regression; and pretraining strategies like contrastive learning, autoencoding, and coil classiﬁcation) to this speciﬁc industrial problem, but also methodological advancements such as the statistics-aware imputation loss and efﬁciency improvements to the ROCKET model.These ﬁndings show that machine learning models, particularly those enhanced with representation learning techniques, can signiﬁcantly improve the accuracy of on-line surface rough-ness estimation. This paves the way for reliable real-time quality control, enabling faster process feedback, reducing waste, facilitating process optimisation, and ultimately opening possibilities for automated closed-loop control systems in steel manufacturing.</abstract><type>E-Thesis</type><journal/><volume/><journalNumber/><paginationStart/><paginationEnd/><publisher/><placeOfPublication>Swansea</placeOfPublication><isbnPrint/><isbnElectronic/><issnPrint/><issnElectronic/><keywords>Machine learning, Surface roughness, Ra, On-line measurement, Steel manufacturing, Temper rolling, Time Series Extrinsic Regression (TSER), Deep Learning, Self-Supervised Learning, Data Imputation</keywords><publishedDay>3</publishedDay><publishedMonth>3</publishedMonth><publishedYear>2026</publishedYear><publishedDate>2026-03-03</publishedDate><doi>10.23889/SUThesis.71779</doi><url/><notes/><college>COLLEGE NANME</college><CollegeCode>COLLEGE CODE</CollegeCode><institution>Swansea University</institution><supervisor>Xie, X., and Tam, G.</supervisor><degreelevel>Doctoral</degreelevel><degreename>Ph.D</degreename><degreesponsorsfunders>EPSRC Industrial Case award</degreesponsorsfunders><apcterm/><funders>EPSRC Industrial Case award</funders><projectreference/><lastEdited>2026-04-22T13:20:41.4951365</lastEdited><Created>2026-04-22T12:47:10.2591281</Created><path><level id="1">Faculty of Science and Engineering</level><level id="2">School of Mathematics and Computer Science - Computer Science</level></path><authors><author><firstname>ALEXANDER</firstname><surname>MILNE</surname><order>1</order></author></authors><documents><document><filename>71779__36549__916a91c0a2974fcc9b63f5d8dd029d86.pdf</filename><originalFilename>2026_Milne_A.final.71779.pdf</originalFilename><uploaded>2026-04-22T13:10:40.8136516</uploaded><type>Output</type><contentLength>20937092</contentLength><contentType>application/pdf</contentType><version>E-Thesis – open access</version><cronfaStatus>true</cronfaStatus><documentNotes>Copyright: the author, Alex Milne, 2026, Distributed under the terms of a Creative Commons Attribution NonCommercial 4.0 License (CC BY-NC 4.0)</documentNotes><copyrightCorrect>true</copyrightCorrect><language>eng</language><licence>https://creativecommons.org/licenses/by-nc/4.0/</licence></document></documents><OutputDurs/></rfc1807>
spelling	2026-04-22T13:20:41.4951365 v2 71779 2026-04-22 Steel Surface Parameterisation for On-line Quantification with Machine Learning Models c6da9da5c99754b6850e882895b86ca5 ALEXANDER MILNE ALEXANDER MILNE true false 2026-04-22 Accurately estimating steel surface roughness parameters in real-time during industrial manufacturing is a critical challenge. Traditional post-production stylus-based measurements of surface roughness, speciﬁcally the arithmetic mean roughness (Ra), are slow and sample only small sections of a steel coil, leading to production inefﬁciencies, costly rework or downgrading of coils, and potential downstream delays in demanding just-in-time manufacturing environments due to re-production when quality is not met. While the on-line, non-contact laser reﬂection measurement device used by our partners offers a continuous monitoring solution, its accuracy is often insufﬁcient compared to stylus-based ground truth measurements, negating its utility for process control.This research explores the application of machine learning (ML) to enhance the accuracy of surface roughness estimation from raw laser reﬂection data. By leveraging both limited la-belled and abundant unlabelled industrial datasets, this work investigates various ML-driven methods. These include: data imputation techniques aimed at improving existing calculation methods by addressing data quality issues; end-to-end models designed for direct mapping from sensor input to the stylus-equivalent Ra value; efﬁciency enhancements for promising model architectures; and the incorporation of representation learning via self-supervised learning (SSL) to effectively utilise unlabelled data and overcome labelled data scarcity.The ﬁndings demonstrate the substantial advantages of ML approaches over conventional methods. End-to-end deep learning models consistently achieve signiﬁcantly higher accuracy in predicting Ra compared to the baseline sensor calculations. Our comprehensive comparative studies reveal that architectures such as Temporal Convolutional Networks (TCNs) and certain 2D Convolutional Neural Networks provide a robust combination of high predictive ac-curacy and computational feasibility for this speciﬁc industrial data type. Furthermore, lever-aging large unlabelled datasets via pretraining proved highly effective; representation learning strategies, particularly a novel task-speciﬁc coil classiﬁcation pretext task and a TCN-based autoencoder, yielded models with enhanced generalisation and accuracy upon ﬁne-tuning with limited labelled data. Methodological insights also include the necessity of a statistics-aware loss function for imputation tasks requiring the preservation of surface texture characteristics, and the successful GPU acceleration of the ROCKET model for enhanced throughput.The key contributions of this thesis therefore encompass not only the successful application and comparison of various ML techniques (imputation; end-to-end regression; and pretraining strategies like contrastive learning, autoencoding, and coil classiﬁcation) to this speciﬁc industrial problem, but also methodological advancements such as the statistics-aware imputation loss and efﬁciency improvements to the ROCKET model.These ﬁndings show that machine learning models, particularly those enhanced with representation learning techniques, can signiﬁcantly improve the accuracy of on-line surface rough-ness estimation. This paves the way for reliable real-time quality control, enabling faster process feedback, reducing waste, facilitating process optimisation, and ultimately opening possibilities for automated closed-loop control systems in steel manufacturing. E-Thesis Swansea Machine learning, Surface roughness, Ra, On-line measurement, Steel manufacturing, Temper rolling, Time Series Extrinsic Regression (TSER), Deep Learning, Self-Supervised Learning, Data Imputation 3 3 2026 2026-03-03 10.23889/SUThesis.71779 COLLEGE NANME COLLEGE CODE Swansea University Xie, X., and Tam, G. Doctoral Ph.D EPSRC Industrial Case award EPSRC Industrial Case award 2026-04-22T13:20:41.4951365 2026-04-22T12:47:10.2591281 Faculty of Science and Engineering School of Mathematics and Computer Science - Computer Science ALEXANDER MILNE 1 71779__36549__916a91c0a2974fcc9b63f5d8dd029d86.pdf 2026_Milne_A.final.71779.pdf 2026-04-22T13:10:40.8136516 Output 20937092 application/pdf E-Thesis – open access true Copyright: the author, Alex Milne, 2026, Distributed under the terms of a Creative Commons Attribution NonCommercial 4.0 License (CC BY-NC 4.0) true eng https://creativecommons.org/licenses/by-nc/4.0/
title	Steel Surface Parameterisation for On-line Quantification with Machine Learning Models
spellingShingle	Steel Surface Parameterisation for On-line Quantification with Machine Learning Models ALEXANDER MILNE
title_short	Steel Surface Parameterisation for On-line Quantification with Machine Learning Models
title_full	Steel Surface Parameterisation for On-line Quantification with Machine Learning Models
title_fullStr	Steel Surface Parameterisation for On-line Quantification with Machine Learning Models
title_full_unstemmed	Steel Surface Parameterisation for On-line Quantification with Machine Learning Models
title_sort	Steel Surface Parameterisation for On-line Quantification with Machine Learning Models
author_id_str_mv	c6da9da5c99754b6850e882895b86ca5
author_id_fullname_str_mv	c6da9da5c99754b6850e882895b86ca5_***_ALEXANDER MILNE
author	ALEXANDER MILNE
author2	ALEXANDER MILNE
format	E-Thesis
publishDate	2026
institution	Swansea University
doi_str_mv	10.23889/SUThesis.71779
college_str	Faculty of Science and Engineering
hierarchytype
hierarchy_top_id	facultyofscienceandengineering
hierarchy_top_title	Faculty of Science and Engineering
hierarchy_parent_id	facultyofscienceandengineering
hierarchy_parent_title	Faculty of Science and Engineering
department_str	School of Mathematics and Computer Science - Computer Science{{{_:::_}}}Faculty of Science and Engineering{{{_:::_}}}School of Mathematics and Computer Science - Computer Science
document_store_str	1
active_str	0
description	Accurately estimating steel surface roughness parameters in real-time during industrial manufacturing is a critical challenge. Traditional post-production stylus-based measurements of surface roughness, speciﬁcally the arithmetic mean roughness (Ra), are slow and sample only small sections of a steel coil, leading to production inefﬁciencies, costly rework or downgrading of coils, and potential downstream delays in demanding just-in-time manufacturing environments due to re-production when quality is not met. While the on-line, non-contact laser reﬂection measurement device used by our partners offers a continuous monitoring solution, its accuracy is often insufﬁcient compared to stylus-based ground truth measurements, negating its utility for process control.This research explores the application of machine learning (ML) to enhance the accuracy of surface roughness estimation from raw laser reﬂection data. By leveraging both limited la-belled and abundant unlabelled industrial datasets, this work investigates various ML-driven methods. These include: data imputation techniques aimed at improving existing calculation methods by addressing data quality issues; end-to-end models designed for direct mapping from sensor input to the stylus-equivalent Ra value; efﬁciency enhancements for promising model architectures; and the incorporation of representation learning via self-supervised learning (SSL) to effectively utilise unlabelled data and overcome labelled data scarcity.The ﬁndings demonstrate the substantial advantages of ML approaches over conventional methods. End-to-end deep learning models consistently achieve signiﬁcantly higher accuracy in predicting Ra compared to the baseline sensor calculations. Our comprehensive comparative studies reveal that architectures such as Temporal Convolutional Networks (TCNs) and certain 2D Convolutional Neural Networks provide a robust combination of high predictive ac-curacy and computational feasibility for this speciﬁc industrial data type. Furthermore, lever-aging large unlabelled datasets via pretraining proved highly effective; representation learning strategies, particularly a novel task-speciﬁc coil classiﬁcation pretext task and a TCN-based autoencoder, yielded models with enhanced generalisation and accuracy upon ﬁne-tuning with limited labelled data. Methodological insights also include the necessity of a statistics-aware loss function for imputation tasks requiring the preservation of surface texture characteristics, and the successful GPU acceleration of the ROCKET model for enhanced throughput.The key contributions of this thesis therefore encompass not only the successful application and comparison of various ML techniques (imputation; end-to-end regression; and pretraining strategies like contrastive learning, autoencoding, and coil classiﬁcation) to this speciﬁc industrial problem, but also methodological advancements such as the statistics-aware imputation loss and efﬁciency improvements to the ROCKET model.These ﬁndings show that machine learning models, particularly those enhanced with representation learning techniques, can signiﬁcantly improve the accuracy of on-line surface rough-ness estimation. This paves the way for reliable real-time quality control, enabling faster process feedback, reducing waste, facilitating process optimisation, and ultimately opening possibilities for automated closed-loop control systems in steel manufacturing.
published_date	2026-03-03T07:50:43Z
_version_	1863424245501001728
score	11.103076

Steel Surface Parameterisation for On-line Quantification with Machine Learning Models / ALEXANDER MILNE

Similar Items