A deep learning-based approach for defect detection in powder bed fusion additive manufacturing using transfer learning

Duman, Burhan; Özsoy, Koray

doi:10.17341/gazimmfd.870436

Cited by 8 publications

(1 citation statement)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, one of the significant challenges associated with AM is the need for better surface quality of printed parts. This issue has prompted the development of pretreatment techniques to minimize errors and determine optimal printing parameters [1][2][3][4][5]. Unfortunately, these preprocessing methodologies often have limited effectiveness.…”

Section: Introductionmentioning

confidence: 99%

Experimental investigation and optimization of the effect garnet vibratory tumbling as a post-process on the surface quality of 3D printed PLA parts

KARTAL,

KAPTAN

2024

European Mechanical Science

View full text Add to dashboard Cite

The method known as additive manufacturing causes high surface roughness between layers depending on the technique used at the end of the product development process. This can be an important problem in three-dimensional (3D) manufacturing depending on the usage area. To solve this problem, in this experimental study, the effect of vibratory tumbling (VT) on surface roughness in 3D printing was investigated using garnet abrasive particles. Optimization with the best parameters was also performed and the results were analyzed. This experimental study investigated the effect of vibratory tumbling on surface roughness in 3D printing produced from Polylactic acid (PLA) material using garnet abrasive particles. The surface roughness (Ra) values were measured at different vibration durations for each mesh size. The results provide insights into the impact of vibratory tumbling on surface roughness in 3D-printed parts. The study involved subjecting the printed parts to vibratory tumbling using garnet abrasive particles of various mesh sizes (80, 90, 100, 120, 150, 180, and 220 mesh). Surface roughness measurements were taken at different vibration durations (2, 4, 6, 8, 10, and 12 hours) for each mesh size. A surface roughness measuring device was used to obtain the roughness values. The findings reveal that vibratory tumbling with garnet abrasive particles effectively reduces surface roughness in 3D printed parts. As the vibration duration increased, smoother surfaces were achieved. The data collected for each mesh size and vibration duration offer valuable insights into the relationship between vibratory tumbling and surface roughness in 3D printing. The surface roughness of the printed samples was reduced by 60% on average by using the optimum values after post-process. This research highlights the potential of vibratory tumbling as a viable method for improving surface roughness in 3D printing applications. Emphasis is placed on optimizing the vibration duration and selecting the appropriate mesh size to achieve the desired surface quality. Overall, this study contributes to our understanding of the effect of vibratory tumbling on surface roughness in 3D printing and provides considerable insights for enhancing surface quality in additive manufacturing processes.

show abstract

Section: Introductionmentioning

confidence: 99%

Experimental investigation and optimization of the effect garnet vibratory tumbling as a post-process on the surface quality of 3D printed PLA parts

KARTAL,

KAPTAN

2024

European Mechanical Science

View full text Add to dashboard Cite

show abstract

Analysis of Defect Associated with Powder Bed Fusion with Deep Learning and Explainable AI

Pratap

Sardana

Utomo

et al. 2023

2023 15th International Conference on Knowledge and Smart Technology (KST)

View full text Add to dashboard Cite

Domates yapraklarında hastalık tespiti için önerilen hafif evrişimli sinir ağı ile önceden eğitilmiş ağların performans karşılaştırması

Ecemis

İlhan

2022

Gazi Üniversitesi Mühendislik Mimarlık Fakültesi Dergisi

View full text Add to dashboard Cite

Domateslerde oluşan hastalıkların çoğu yaprakların gözlemlenmesi ile belirlenebilir. Bu nedenle domates yapraklarının gözlemlenmesi, sebze yetiştiriciliğinde kritik bir öneme sahiptir. Yaprakların üzerinde bulunan izler, renk değişimleri, kıvrıklık vb. hastalığı ya da hastalık riski oluşturabilecek bir virüsü işaret edebilir. Yaprakların sağlığının korunması domateslerin de verimliliğini artıracağından, yaprak gelişimlerinin gözlemlenmesi, olası hastalık risklerinin önceden doğru ve hızlı tespitinin yapılması bir gerekliliktir. Önerilen çalışmada, domates bitkisinin yapraklarında oluşan 9 farklı hastalığı ve sağlıklı yaprakları tespit edebilen "T-LeafNet" Evrişimsel Sinir Ağı (ESA-CNN) modeli önerilmiştir. Önerilen ağ sıfırdan eğitilerek sınıflandırma başarısı ölçülmüştür. Ayrıca, öğrenme aktarımı (transfer learning) kapsamında hazır ağ modellerinden MobileNetV2, AlexNet ve VGG16 mimarileri de veri kümesi üzerinde test edilmiş ve sonuçlar karşılaştırılmıştır. Testler orjinal ve artırılmış veriler üzerinde farklı öğrenme katsayıları ile uygulanarak, veri artırma ve öğrenme katsayı parametresinin başarıma etkisi de incelenmiştir. T-LeafNet, AlexNet, MobileNetV2 ve VGG16 ağlarının en yüksek sınıflandırma performansları %97,32, %98,32, %99,1 ve %99,21 olarak ölçülmüştür. Eğitim zamanı ve sınıflandırma performansları göz önünde bulundurulduğunda, önerilen T-LeafNet ağının, test edilen diğer ağlara kıyasen daha kısa süre ve az kaynakla, yüksek performans verdiği gözlemlenmiştir.

show abstract

A deep learning-based approach for defect detection in powder bed fusion additive manufacturing using transfer learning

Cited by 8 publications

References 49 publications

Experimental investigation and optimization of the effect garnet vibratory tumbling as a post-process on the surface quality of 3D printed PLA parts

Experimental investigation and optimization of the effect garnet vibratory tumbling as a post-process on the surface quality of 3D printed PLA parts

Analysis of Defect Associated with Powder Bed Fusion with Deep Learning and Explainable AI

Domates yapraklarında hastalık tespiti için önerilen hafif evrişimli sinir ağı ile önceden eğitilmiş ağların performans karşılaştırması

Contact Info

Product

Resources

About