Delamination and thrust force analysis in GLARE: Influence of tool geometry and prediction with machine learning models


EKİCİ E., Pazarkaya İ., UZUN G.

Journal of Composite Materials, 2024 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Basım Tarihi: 2024
  • Doi Numarası: 10.1177/00219983241305706
  • Dergi Adı: Journal of Composite Materials
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Applied Science & Technology Source, Chimica, Communication Abstracts, Compendex, Computer & Applied Sciences, INSPEC, Metadex, Civil Engineering Abstracts
  • Anahtar Kelimeler: analysis of variance, delamination, Glass fibre aluminium reinforced epoxy, long-short-term memory, machine learning, uncut fiber
  • Çanakkale Onsekiz Mart Üniversitesi Adresli: Evet

Özet

The multi-layered (fiber/metal) structure of glass fibre aluminium reinforced epoxy (GLARE) makes it difficult to obtain acceptable damage-free holes that meet aerospace standards. This paper investigated the effects of tool geometry and drilling parameters on reducing delamination damage and uncut fibers at the hole exit surface in drilling GLARE. The hole surfaces were examined by scanning electron microscope (SEM) at various magnifications. In addition, deep neural network (DNN) and long-short-term memory (LSTM) machine learning models were used to predict delamination (Fda), uncut fiber (UCF), and thrust forces using experimental data. No positive contribution of the special geometry tool was observed, while the standard geometry tool was found to be ideal for drilling conditions. Analysis of variance (ANOVA) revealed that feed rate contributed 57.83% to delamination damage, while tool geometry contributed 74.31% and 92.33% for uncut fiber and thrust force, respectively. SEM analysis revealed high deformation zones in the aluminum layers and fiber fracture and separation in the glass fibre reinforced polymer (GFRP) layers. DNN and LSTM models were found to provide accurate predictions with R2 values greater than 95% and 98%, respectively.