DETECTION OF SURFACE-BASED ANOMALIES FOR SELF-TAPPING SCREWS IN PLASTIC HOUSINGS USING SUPERVISED MACHINE LEARNING

West, N; Trianni, A; Deuse, J

DETECTION OF SURFACE-BASED ANOMALIES FOR SELF-TAPPING SCREWS IN PLASTIC HOUSINGS USING SUPERVISED MACHINE LEARNING

West, N Trianni, A

Deuse, J

Permalink

Publication Type:: Conference Proceeding
Citation:: Proceedings of International Conference on Computers and Industrial Engineering CIE, 2024, 2024-December, pp. 1395-1404
Issue Date:: 2024-01-01

Closed Access

	Filename	Description	Size
	2024_West-et-al_CIE51-Submission_final.pdf	Published version	1.02 MB	Adobe PDF	View/Open
	CIE51 notification for paper 225.pdf	Supporting information	101.27 kB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	West, N
dc.contributor.author	Trianni, A https://orcid.org/0000-0002-1699-7787
dc.contributor.author	Deuse, J https://orcid.org/0000-0003-4066-4357
dc.date.accessioned	2025-06-10T08:24:54Z
dc.date.available	2025-06-10T08:24:54Z
dc.date.issued	2024-01-01
dc.identifier.citation	Proceedings of International Conference on Computers and Industrial Engineering CIE, 2024, 2024-December, pp. 1395-1404
dc.identifier.issn	2164-8689
dc.identifier.uri	http://hdl.handle.net/10453/187676
dc.description.abstract	This paper presents a comprehensive study on anomaly detection in screw connections using supervised machine learning techniques. We introduce a novel, open-source dataset comprising 12,500 time series observations from screw tightening processes, including both normal operations and seven distinct error types related to surface properties. The dataset encompasses multiple tightening cycles, allowing for the analysis of progressive wear effects. To establish a benchmark, four supervised classifiers (Naive Bayes, Multilayer Perceptron, k-Nearest Neighbours, and Random Forest) are evaluated across six different scenarios. These scenarios consider various screw cycles and labelling strategy granularities. Results demonstrate that classification performance improves as the number of classes decreases, with binary classification yielding the highest accuracies. The Random Forest model consistently outperforms other algorithms. However, all models exhibit a marked decrease in performance when analysing data from multiple tightening cycles, highlighting challenges posed by wear effects and subtle temporal changes. Our findings provide insights into the varying degree of detectability of different error types and the challenges of automated quality assurance in manufacturing processes. This study contributes to the field of anomaly detection in manufacturing, offering a benchmark for further research and demonstrating the potential of machine learning in identifying complex error patterns in screw connections.
dc.language	en
dc.relation.ispartof	Proceedings of International Conference on Computers and Industrial Engineering CIE
dc.rights	info:eu-repo/semantics/closedAccess
dc.title	DETECTION OF SURFACE-BASED ANOMALIES FOR SELF-TAPPING SCREWS IN PLASTIC HOUSINGS USING SUPERVISED MACHINE LEARNING
dc.type	Conference Proceeding
utslib.citation.volume	2024-December
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology/School of Mechanical and Mechatronic Engineering
pubs.organisational-group	University of Technology Sydney/UTS Groups
pubs.organisational-group	University of Technology Sydney/UTS Groups/Centre for Advanced Manufacturing (CAM)
utslib.copyright.status	closed_access	*
dc.date.updated	2025-06-10T08:24:50Z
pubs.publication-status	Published
pubs.volume	2024-December

Abstract:

This paper presents a comprehensive study on anomaly detection in screw connections using supervised machine learning techniques. We introduce a novel, open-source dataset comprising 12,500 time series observations from screw tightening processes, including both normal operations and seven distinct error types related to surface properties. The dataset encompasses multiple tightening cycles, allowing for the analysis of progressive wear effects. To establish a benchmark, four supervised classifiers (Naive Bayes, Multilayer Perceptron, k-Nearest Neighbours, and Random Forest) are evaluated across six different scenarios. These scenarios consider various screw cycles and labelling strategy granularities. Results demonstrate that classification performance improves as the number of classes decreases, with binary classification yielding the highest accuracies. The Random Forest model consistently outperforms other algorithms. However, all models exhibit a marked decrease in performance when analysing data from multiple tightening cycles, highlighting challenges posed by wear effects and subtle temporal changes. Our findings provide insights into the varying degree of detectability of different error types and the challenges of automated quality assurance in manufacturing processes. This study contributes to the field of anomaly detection in manufacturing, offering a benchmark for further research and demonstrating the potential of machine learning in identifying complex error patterns in screw connections.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/187676