Scholarlink

<p>The wire arc additive manufacturing (WAAM) process is a subset of direct energy deposition (DED) within, in the context of metal additive manufacturing (MAM), and has been attracting significant attention in both industrial and scientific productions fields, mainly due to its high productivity, demonstrating great potential for manufacturing large-sized parts within an attractive timeframe for industrial implementation. However, given its limitations surrounding the certification of manufactured products, critical for high-responsibility industrial sectors, monitoring and controlling defects in the WAAM process represent an alternative approach to increase the reliability of the process and its products. Thus, this study presents a systematic literature review (SLR) focusing on investigating the application of machine learning (ML) approach techniques in defect detection in WAAM. This systematic review investigates the application of ML techniques in defect detection in WAAM, analyzing different methodologies and key challenges associated with their implementation. For this purpose, the review was guided by following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol guidelines, 150 studies were retrieved initially, and the publication search was performed on retrieved from Web of Science, Scopus, ACM Digital Library, and IEEE Xplore databases. After applying inclusion and exclusion criteria, 28 studies were selected for in-depth analysis. The findings indicate that the ML-based strategies for defect detection in WAAM effectively validate data processing frameworks, with supervised learning algorithms such as convolutional neural networks (CNNs), support vector machines (SVMs), and ensemble methods showing promising results in predictive modeling. In general, the results demonstrated that the machine learning approach for defect control in WAAM proves to be effective in validating data processing structures, showing good results in terms of evaluation metrics. Additionally, hybrid approaches, including boosting models and the integration of multiple algorithms, were particularly effective in enhancing data processing capabilities. The study also identifies key challenges in WAAM defect monitoring, such as data acquisition limitations, real-time processing constraints, and the multifactorial nature of defect formation of boosting models or the application of multiple models favors the optimization of data processing.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning approach in defect detection in wire arc additive manufacturing: a literature review

  • 期刊论文

  • 发表日期：

    2025年9月2日

查看全文
  • Joyce Ingrid Venceslau de Souto,
  • Keila Lucas dos Santos,
  • Walman Benício de Castro,
  • Jefferson Segundo de Lima

摘要

The wire arc additive manufacturing (WAAM) process is a subset of direct energy deposition (DED) within, in the context of metal additive manufacturing (MAM), and has been attracting significant attention in both industrial and scientific productions fields, mainly due to its high productivity, demonstrating great potential for manufacturing large-sized parts within an attractive timeframe for industrial implementation. However, given its limitations surrounding the certification of manufactured products, critical for high-responsibility industrial sectors, monitoring and controlling defects in the WAAM process represent an alternative approach to increase the reliability of the process and its products. Thus, this study presents a systematic literature review (SLR) focusing on investigating the application of machine learning (ML) approach techniques in defect detection in WAAM. This systematic review investigates the application of ML techniques in defect detection in WAAM, analyzing different methodologies and key challenges associated with their implementation. For this purpose, the review was guided by following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol guidelines, 150 studies were retrieved initially, and the publication search was performed on retrieved from Web of Science, Scopus, ACM Digital Library, and IEEE Xplore databases. After applying inclusion and exclusion criteria, 28 studies were selected for in-depth analysis. The findings indicate that the ML-based strategies for defect detection in WAAM effectively validate data processing frameworks, with supervised learning algorithms such as convolutional neural networks (CNNs), support vector machines (SVMs), and ensemble methods showing promising results in predictive modeling. In general, the results demonstrated that the machine learning approach for defect control in WAAM proves to be effective in validating data processing structures, showing good results in terms of evaluation metrics. Additionally, hybrid approaches, including boosting models and the integration of multiple algorithms, were particularly effective in enhancing data processing capabilities. The study also identifies key challenges in WAAM defect monitoring, such as data acquisition limitations, real-time processing constraints, and the multifactorial nature of defect formation of boosting models or the application of multiple models favors the optimization of data processing.