Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Clustering-based Adaptive Data Augmentation for Class-imbalance in Machine Learning (CADA): Additive Manufacturing Use-case
Blekinge Institute of Technology, Faculty of Computing, Department of Computer Science.ORCID iD: 0000-0002-3311-2530
Blekinge Institute of Technology, Faculty of Computing, Department of Computer Science.ORCID iD: 0000-0002-4390-411x
GKN Aerospace Engine Systems.
Blekinge Institute of Technology, Faculty of Computing, Department of Computer Science.
2022 (English)In: Neural Computing & Applications, ISSN 0941-0643, E-ISSN 1433-3058Article in journal (Refereed) Epub ahead of print
Abstract [en]

Large amount of data are generated from in-situ monitoring of additive manufacturing (AM) processes which is later used in prediction modelling for defect classification to speed up quality inspection of products. A high volume of this process data is defect-free (majority class) and a lower volume of this data has defects (minority class) which result in the class-imbalance issue. Using imbalanced datasets, classifiers often provide sub-optimal classification results i.e. better performance on the majority class than the minority class. However, it is important for process engineers that models classify defects more accurately than the class with no defects since this is crucial for quality inspection. Hence, we address the class-imbalance issue in manufacturing process data to support in-situ quality control of additive manufactured components.  For this, we propose cluster-based adaptive data augmentation (CADA) for oversampling to address the class-imbalance problem. Quantitative experiments are conducted to evaluate the performance of the proposed method and to compare with other selected oversampling methods using AM datasets from an aerospace industry and a publicly available casting manufacturing dataset. The results show that CADA outperformed random oversampling and the SMOTE method and is similar to random data augmentation and cluster-based oversampling. Furthermore, the results of the statistical significance test show that there is a significant difference between the studied methods.  As such, the CADA method can be considered as an alternative method for oversampling to improve the performance of models on the minority class. 

Place, publisher, year, edition, pages
Springer London, 2022.
Keywords [en]
Class-imbalance, Melt-pool defects classification, Aerospace application, Additive Manufacturing, Polar Transformation, Random Forests
National Category
Computer Sciences
Identifiers
URN: urn:nbn:se:bth-22028DOI: 10.1007/s00521-022-07347-6ISI: 000800995800001OAI: oai:DiVA.org:bth-22028DiVA, id: diva2:1586517
Note

open access

Available from: 2021-08-20 Created: 2021-08-20 Last updated: 2022-06-10Bibliographically approved
In thesis
1. Predictive Modelling to Support Design and Manufacturing in Aerospace Engineering
Open this publication in new window or tab >>Predictive Modelling to Support Design and Manufacturing in Aerospace Engineering
2021 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

A crucial issue in the design of aircraft components is the evaluation of a large number of potential design alternatives.  This evaluation involves too expensive procedures, consequently, it slows down the search for optimal design samples.  As a result, a scarce or small number of design samples with high dimensional parameter space will pose issues in the learning of surrogate models. These issues bring the need to investigate methods for surrogate modelling for the most effective use of available data. Furthermore, during the manufacturing of components, it is crucial to monitor (in-situ process monitoring) the welding process for quality assurance. A large amount of process data is generated from these in-situ monitoring methods, which can be used to build prediction models for defects classification. However, the process data are unstructured, and defects are unknown, which brings the need for investigations to address these issues to build defect classification models. 

 The thesis goal is to support engineers in the early design and manufacturing phases of aircraft engine components via (1) surrogate modelling for the purpose of exploration of larger search spaces and for speeding up the evaluation of design configurations, and (2) defects classification to support in-situ process monitoring to speed up defects' analysis. 

The first part of the thesis focuses on addressing challenges in design data when building surrogate models. For this, the thesis explores, evaluates, and improves tree models for design space exploration. The second part of the thesis focuses on addressing challenges in process data when building defect classification models. For this, the thesis (1) investigates the performance of selected handcrafted feature extraction techniques, (2) proposes an oversampling technique to balance process datasets, and (3) proposes an active learning approach for labelling data. 

Place, publisher, year, edition, pages
Karlskrona: Blekinge Tekniska Högskola, 2021
Series
Blekinge Institute of Technology Doctoral Dissertation Series, ISSN 1653-2090 ; 7
Keywords
surrogate modelling, defects classification, machine learning, aerospace, additive manufacturing
National Category
Computer Sciences
Research subject
Computer Science
Identifiers
urn:nbn:se:bth-22033 (URN)978-91-7295-428-1 (ISBN)
Public defence
2021-10-18, Distance/J1630, Blekinge Institute of Technology, Karlskrona, 13:30 (English)
Opponent
Supervisors
Available from: 2021-08-23 Created: 2021-08-20 Last updated: 2021-11-15Bibliographically approved

Open Access in DiVA

fulltext(1218 kB)149 downloads
File information
File name FULLTEXT01.pdfFile size 1218 kBChecksum SHA-512
ed824471a829fa5df690798232edd47337a503cadbcf18e11e26ea26e315c0297aecaa88998124b38c15ee76cc719f4e81c8c70caae3e00732115732cf94b2d8
Type fulltextMimetype application/pdf

Other links

Publisher's full text

Authority records

Dasari, Siva KrishnaCheddad, AbbasLundberg, Lars

Search in DiVA

By author/editor
Dasari, Siva KrishnaCheddad, AbbasLundberg, Lars
By organisation
Department of Computer Science
In the same journal
Neural Computing & Applications
Computer Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 149 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
urn-nbn

Altmetric score

doi
urn-nbn
Total: 351 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf