A pre-trained model selection for transfer learning of remaining useful life prediction of grinding wheel
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Park, Seung-Ho | - |
dc.contributor.author | Park, Kyoung-Su | - |
dc.date.accessioned | 2024-06-23T12:00:25Z | - |
dc.date.available | 2024-06-23T12:00:25Z | - |
dc.date.issued | 2024-06 | - |
dc.identifier.issn | 0956-5515 | - |
dc.identifier.issn | 1572-8145 | - |
dc.identifier.uri | https://scholarworks.bwise.kr/gachon/handle/2020.sw.gachon/91635 | - |
dc.description.abstract | Grinding tools can be used for over 90% of their lifetime, but most practical production tools are replaced between 50 and 80%. Accurate tool condition monitoring, which can monitor the quality of grinding tools, can improve product quality and reduce costs. The transfer learning technique can solve the data distribution and data deficiency issues of deep learning methods, enabling precise tool condition prediction in a grinding system. However, selecting the optimal pre-trained model is crucial for enhancing transfer learning performance due to the relationship between the source and target domains and the negative transfer problem. To address this issue, an encoding metric-based model selection technique was developed that accurately reflects the time series similarity between the remaining useful life (RUL) of the grinding tool and the encoding value and features between the pre-trained model and target data. Encoding metrics that reflect the critical characteristics of transfer learning, such as the target label, data, and pre-trained model, were computed to assess the pre-trained models directly. The target RUL and the extracted encoding values from the target input data and pre-trained models were compared using measurements and correlation to evaluate how closely they matched. The encoding selection method demonstrated superiority over the current source selection criteria, with encoding-based root mean square error (RMSE), mean absolute error, and dynamic time warping showing a high association with the approach (over a Pearson correlation coefficient of 0.66). Furthermore, the method was independent of the quantity and quality of the sources, making it robust for various datasets. The experimental dataset was used to validate the pre-trained model's encoding RMSE selection method, and the method was verified with a strong correlation (Pearson correlation coefficient of 0.79). In conclusion, the proposed encoding metric-based model selection technique can improve both the efficiency and intelligence of the machining process in terms of both quantity (exchange periods) and quality of grinding tools. Accurate tool condition monitoring enabled by this technique can lead to enhanced product quality and reduced costs. | - |
dc.format.extent | 18 | - |
dc.language | 영어 | - |
dc.language.iso | ENG | - |
dc.publisher | SPRINGER | - |
dc.title | A pre-trained model selection for transfer learning of remaining useful life prediction of grinding wheel | - |
dc.type | Article | - |
dc.identifier.wosid | 001008415800001 | - |
dc.identifier.doi | 10.1007/s10845-023-02154-9 | - |
dc.identifier.bibliographicCitation | JOURNAL OF INTELLIGENT MANUFACTURING, v.35, no.5, pp 2295 - 2312 | - |
dc.description.isOpenAccess | N | - |
dc.identifier.scopusid | 2-s2.0-85161953295 | - |
dc.citation.endPage | 2312 | - |
dc.citation.startPage | 2295 | - |
dc.citation.title | JOURNAL OF INTELLIGENT MANUFACTURING | - |
dc.citation.volume | 35 | - |
dc.citation.number | 5 | - |
dc.type.docType | Article | - |
dc.publisher.location | 네델란드 | - |
dc.subject.keywordAuthor | Transfer learning | - |
dc.subject.keywordAuthor | Pre-trained model selection | - |
dc.subject.keywordAuthor | Remaining useful life | - |
dc.subject.keywordAuthor | Encoding metrics | - |
dc.subject.keywordAuthor | Pearson correlation coefficient | - |
dc.subject.keywordAuthor | Tool condition monitoring | - |
dc.subject.keywordPlus | TIME | - |
dc.subject.keywordPlus | WEAR | - |
dc.relation.journalResearchArea | Computer Science | - |
dc.relation.journalResearchArea | Engineering | - |
dc.relation.journalWebOfScienceCategory | Computer Science, Artificial Intelligence | - |
dc.relation.journalWebOfScienceCategory | Engineering, Manufacturing | - |
dc.description.journalRegisteredClass | scie | - |
dc.description.journalRegisteredClass | scopus | - |
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.
1342, Seongnam-daero, Sujeong-gu, Seongnam-si, Gyeonggi-do, Republic of Korea(13120)031-750-5114
COPYRIGHT 2020 Gachon University All Rights Reserved.
Certain data included herein are derived from the © Web of Science of Clarivate Analytics. All rights reserved.
You may not copy or re-distribute this material in whole or in part without the prior written consent of Clarivate Analytics.