Estimation of friction stir welding parameters of 3D printed sheets using ANN

Nergizhan Anaç; Oğuz Koçar; Erhan Baysal

Article Details

Estimation of friction stir welding parameters of 3D printed sheets using ANN

Authors:

Nergizhan Anaç | Oğuz Koçar | Erhan Baysal

Publisher:

Nigde Omer Halisdemir University

Main Topics:

Makine Mühendisliği

Journal:

Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.13 Issue 1, pp.176-187

Document Type:

Article

Publication Date:

1/2024

Document Language:

Turkish

Document ID (Org No.):

2001309716

pISSN:

2564-6605

Keywords:

Abstract:

 In this study, sheets were printed from PLA Wood and PLA-CF filaments using a 3D printer. The prepared sheets were welded using the FSW method. The process parameters were selected as different mixed tip geometries, tool feed rate, and rotational tool speed. The data set obtained from the experiments was used to train and test artificial neural networks. To optimize the ANN model, 30 models were created by changing the training algorithm, transfer function, and the number of neurons in the hidden layer. The training and test results of the models were statistically evaluated. As a result, in the best prediction model trainlm, number of neurons 14 and tansig, the training result was 96 accurate and the test result was 99 accurate. The training and testing results of the ANN model show that ANN can be used to estimate the value of process parameters in the FSW method.

 In this study, sheets were printed from PLA Wood and PLA-CF filaments using a 3D printer. The prepared sheets were welded using the FSW method. The process parameters were selected as different mixed tip geometries, tool feed rate, and rotational tool speed. The data set obtained from the experiments was used to train and test artificial neural networks. To optimize the ANN model, 30 models were created by changing the training algorithm, transfer function, and the number of neurons in the hidden layer. The training and test results of the models were statistically evaluated. As a result, in the best prediction model trainlm, number of neurons 14 and tansig, the training result was 96 accurate and the test result was 99 accurate. The training and testing results of the ANN model show that ANN can be used to estimate the value of process parameters in the FSW method.

 In this study, sheets were printed from PLA Wood and PLA-CF filaments using a 3D printer. The prepared sheets were welded using the FSW method. The process parameters were selected as different mixed tip geometries, tool feed rate, and rotational tool speed. The data set obtained from the experiments was used to train and test artificial neural networks. To optimize the ANN model, 30 models were created by changing the training algorithm, transfer function, and the number of neurons in the hidden layer. The training and test results of the models were statistically evaluated. As a result, in the best prediction model trainlm, number of neurons 14 and tansig, the training result was 96 accurate and the test result was 99 accurate. The training and testing results of the ANN model show that ANN can be used to estimate the value of process parameters in the FSW method.

 Bu çalışmada 3B yazıcı kullanılarak PLA Wood Ahşap katkılı polilaktik Asit ve PLA-CF karbon fiber katkılı polilaktik Asit filamentlerden plakalar basılmıştır. Hazırlanan plakalar SKK metodu kullanılarak birleştirilmiştir. SKK metodunda işlem parametreleri farklı karıştırıcı uç geometrisi kare, üçgen ve vida, takım ilerleme hızı 20, 40 ve 60 mmdk ve takım dönme hızı 1250, 1750 ve 2250 devdk olarak seçilmiştir. Yapılan deneyler sonucunda elde edilen veri seti 54 deney yapay sinir ağlarının eğitim ve testi için kullanılmıştır. YSA modelinin optimizasyonu için eğitim algoritması, transfer fonksiyonu ve gizli katmandaki nöron sayısı değiştirilerek 30 model oluşturulmuştur. Modellerin eğitim ve test sonuçları istatistiksel olarak değerlendirilmiştir. Sonuç olarak en iyi tahmin modelinde trainlm, nöron sayısı 14 ve tansig eğitim sonucu 99, test sonucu 96 doğrulukla yapılmıştır. YSA modelinin eğitim ve test sonuçları, YSAnın SKK metodunda işlem parametrelerinin değerini tahmin etmek için kullanılabileceğini göstermektedir.

Article Details

Estimation of friction stir welding parameters of 3D printed sheets using ANN

Authors:

Nergizhan Anaç | Oğuz Koçar | Erhan Baysal

Publisher:

Nigde Omer Halisdemir University

Main Topics:

Makine Mühendisliği

Journal:

Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.13 Issue 1, pp.176-187

Document Type:

Article

Publication Date:

1/2024

Document Language:

Turkish

Document ID (Org No.):

2001309716

pISSN:

2564-6605

Additive manufacturing

3D printing

Thermoplastic material

Filler reinforced filament

Friction stir welding

Artificial neural networks (ANN)Eklemeli imalat

3B yazdırma

Termoplastik malzeme

Dolgu takviyeli filament

Sürtünme karıştırma kaynağı

Yapay sinir ağları (YSA)

Issues

Article Details

Authors: Nergizhan Anaç | Oğuz Koçar | Erhan Baysal

Publisher: Nigde Omer Halisdemir University

Main Topics: Makine Mühendisliği

Journal: Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.13 Issue 1, pp.176-187

Document Type: Article

Publication Date: 1/2024

Document Language: Turkish

Document ID (Org No.): 2001309716

pISSN: 2564-6605

Issues

Authors:

Nergizhan Anaç | Oğuz Koçar | Erhan Baysal

Publisher:

Nigde Omer Halisdemir University

Main Topics:

Makine Mühendisliği

Journal:

Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.13 Issue 1, pp.176-187

Document Type:

Article

Publication Date:

1/2024

Document Language:

Turkish

Document ID (Org No.):

2001309716

pISSN:

2564-6605