Journal article 590 views 102 downloads
Anomaly Detection of DC Nut Runner Processes in Engine Assembly
,
Cinzia Giannetti ,
Hessel Van Dijk
AI, Volume: 4, Issue: 1, Pages: 234 - 254
Swansea University Author:
Cinzia Giannetti
-
PDF | Version of Record
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/)
Download (15.08MB)
DOI (Published version): 10.3390/ai4010010
Abstract
In many manufacturing systems, anomaly detection is critical to identifying process errors and ensuring product quality. This paper proposes three semi-supervised solutions to detect anomalies in Direct Current (DC) Nut Runner engine assembly processes. The nut runner process is a challenging anomal...
| Published in: | AI |
|---|---|
| ISSN: | 2673-2688 |
| Published: |
MDPI AG
2023
|
| Online Access: |
Check full text
|
| URI: | https://cronfa.swan.ac.uk/Record/cronfa62914 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Abstract: |
In many manufacturing systems, anomaly detection is critical to identifying process errors and ensuring product quality. This paper proposes three semi-supervised solutions to detect anomalies in Direct Current (DC) Nut Runner engine assembly processes. The nut runner process is a challenging anomaly detection problem due to the manual nature of the process inducing high variability and ambiguity of the anomalous class. These characteristics lead to a scenario where anomalies are not outliers, and the normal operating conditions are difficult to define. To address these challenges, a Gaussian Mixture Model (GMM) was trained using a semi-supervised approach. Three dimensionality reduction methods were compared in pre-processing: PCA, t-SNE, and UMAP. These approaches are demonstrated to outperform the current approaches used by a major automotive company on two real-world datasets. Furthermore, a novel approach to labelling real-world data is proposed, including the concept of an ‘Anomaly No Concern’ class, in addition to the traditional labels of ‘Anomaly’ and ‘Normal’. Introducing this new term helped address knowledge gaps between data scientists and domain experts, as well as providing new insights during model development and testing. This represents a major advancement in identifying anomalies in manual production processes that use handheld tools. |
|---|---|
| Keywords: |
GMM, UMAP, PCA, t-SNE, quality assurance, anomaly detection, nut runner |
| College: |
Faculty of Science and Engineering |
| Funders: |
Funded by M2A COATED2 CDT |
| Issue: |
1 |
| Start Page: |
234 |
| End Page: |
254 |