Effect of Ironing Parameters on Surface Quality and Energy Demand of FDM-Fabricated Proximal Humerus Plates: A Box–Behnken and Machine Learning Study
Journal of Applied Polymer Science, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1002/app.71624
- Dergi Adı: Journal of Applied Polymer Science
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Applied Science & Technology Source, Chemical Abstracts Core, Chimica, Compendex, INSPEC, Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
- Anahtar Kelimeler: Bayesian optimization, Box–Behnken design, fused deposition modeling, Gaussian process regression, ironing parameters, machine learning prediction, proximal humerus bone plate
- Kocaeli Üniversitesi Adresli: Evet
Özet
This study investigates the effects of ironing parameters on surface quality and energy demand in FDM fabrication of proximal-humerus-inspired PLA plate-like demonstrator, aiming to quantify how ironing settings influence surface roughness (Ra) and electrical energy consumption (EC), thereby revealing the quality–efficiency trade-off in the finishing stage. Four ironing related process variables including ironing flow (IF), ironing line spacing (ILS), ironing speed (IS), and ironing inset (II) were investigated at three levels using a BBD. Experimental results showed that Ra varied between 4.52 and 8.52 μm, whereas EC ranged from 261.15 to 545.39 Wh, indicating that improved surface finish generally requires higher energy input. Analysis of variance confirmed that IF, ILS, and IS significantly affect Ra (p < 0.001), while EC is predominantly governed by IS and IF (p < 0.001). To model and predict Ra and EC from ironing parameters, LR, ANN, and GPR-Bayes models were developed and evaluated using a repeated group-based fivefold cross-validation strategy. For Ra prediction, LR and GPR-Bayes showed similar performance, while ANN showed lower performance. For EC prediction, GPR-Bayes achieved the best overall performance, with R2 = 0.8624 ± 0.1741.