Prediction of Compressive Strengths of Silica Fume Substituted Cements Using ANN and ANFIS Models

Yasemin Erdem; Eyyüp Gülbandılar; Yılmaz Koçak

Article Details

Prediction of Compressive Strengths of Silica Fume Substituted Cements Using ANN and ANFIS Models

Authors:

Yasemin Erdem | Eyyüp Gülbandılar | Yılmaz Koçak

Publisher:

Bursa Uludağ University

Main Topics:

Yazılım Testi | Doğrulama ve Validasyon

Journal:

Uludağ Üniversitesi Mühendislik Fakültesi Dergisi Vol.30 Issue 2, pp.355-374

Document Type:

Article

Publication Date:

8/2025

Document Language:

Turkish

Document ID (Org No.):

2001355787

pISSN:

2148-4155

Keywords:

Abstract:

 Environmental awareness has led to intensive efforts to reduce the amount of clinker and improve various properties of cementitious composites by using both artificial and natural mineral additives in cement production stages. In this context, studies have focused on silica fume, which is harmful to the environment but beneficial in terms of improving many properties of cementitious composites. In this study, the experimental results of silica fume substituted cement mortars were analyzed and three models were designed for the prediction of compressive strength using artificial neural networks ANN and adaptive network based fuzzy inference system ANFIS. In the architecture of the models, 120 specimens obtained from the experiments with compressive strength results at 2, 7, 28 and 90 days were used for the training stages. In the test phase, 20 results were used and hydration age days, Portland cement PC, silica fume were used as input parameters and compressive strength of concrete was used as output parameter in the models. When the test results were compared with the experimental results, it was determined gave the best results that ANN-2 model with 0.9995, 0.5490 and 0.3572, and ANFIS-3 model with 0.9987, 1.0200 and 0.5664 for R2, MAPE and RMSE, respectively. According to these data, it is concluded that the agreement between the experimental results and the predicted results is good and that high accuracy can be predicted with the ANN and ANFIS models.

 Environmental awareness has led to intensive efforts to reduce the amount of clinker and improve various properties of cementitious composites by using both artificial and natural mineral additives in cement production stages. In this context, studies have focused on silica fume, which is harmful to the environment but beneficial in terms of improving many properties of cementitious composites. In this study, the experimental results of silica fume substituted cement mortars were analyzed and three models were designed for the prediction of compressive strength using artificial neural networks ANN and adaptive network based fuzzy inference system ANFIS. In the architecture of the models, 120 specimens obtained from the experiments with compressive strength results at 2, 7, 28 and 90 days were used for the training stages. In the test phase, 20 results were used and hydration age days, Portland cement PC, silica fume were used as input parameters and compressive strength of concrete was used as output parameter in the models. When the test results were compared with the experimental results, it was determined gave the best results that ANN-2 model with 0.9995, 0.5490 and 0.3572, and ANFIS-3 model with 0.9987, 1.0200 and 0.5664 for R2, MAPE and RMSE, respectively. According to these data, it is concluded that the agreement between the experimental results and the predicted results is good and that high accuracy can be predicted with the ANN and ANFIS models.

 Environmental awareness has led to intensive efforts to reduce the amount of clinker and improve various properties of cementitious composites by using both artificial and natural mineral additives in cement production stages. In this context, studies have focused on silica fume, which is harmful to the environment but beneficial in terms of improving many properties of cementitious composites. In this study, the experimental results of silica fume substituted cement mortars were analyzed and three models were designed for the prediction of compressive strength using artificial neural networks ANN and adaptive network based fuzzy inference system ANFIS. In the architecture of the models, 120 specimens obtained from the experiments with compressive strength results at 2, 7, 28 and 90 days were used for the training stages. In the test phase, 20 results were used and hydration age days, Portland cement PC, silica fume were used as input parameters and compressive strength of concrete was used as output parameter in the models. When the test results were compared with the experimental results, it was determined gave the best results that ANN-2 model with 0.9995, 0.5490 and 0.3572, and ANFIS-3 model with 0.9987, 1.0200 and 0.5664 for R2, MAPE and RMSE, respectively. According to these data, it is concluded that the agreement between the experimental results and the predicted results is good and that high accuracy can be predicted with the ANN and ANFIS models.

 Çevresel farkındalık, çimento üretim aşamalarında gerek yapay gerekse doğal mineral katkılar kullanılarak klinker miktarının azaltılması ve çimentolu kompozitlerin çeşitli özelliklerini iyileştirilmesi amacıyla yoğun çalışmaların yapılmasını sağlamaktadır. Bu bağlamda yapılan çalışmalarda, çevre için zararlı, ancak çimentolu kompozitlerin birçok özelliğini iyileştirmesi açısından faydalı olan silis dumanı üzerinde yoğunlaşıldığı görülmektedir. Bu çalışmada, silis dumanı ikameli çimento harçlarının deneysel sonuçları incelenerek yapay sinir ağları YSA ve uyarlamalı ağ tabanlı bulanık çıkarım sistemi ANFIS ile basınç dayanımlarının tahmini için üçer model tasarımı yapılmıştır. Modellerin mimarisinde deneylerden elde edilen ve 2, 7, 28 ve 90. gün basınç dayanım sonuçları olan 120 numune eğitim aşamaları için kullanılmıştır. Test aşamasında ise 20 sonuç kullanılmış ve modellerde hidratasyon yaşı gün, Portland çimento PÇ, silis dumanı giriş verisi, betonun basınç dayanımı ise çıkış verisi olarak kullanılmıştır. Test sonuçları ile deneysel sonuçlar karşılaştırıldığında, en iyi sonuçları veren modelin R2, MAPE ve RMSE için sırasıyla 0,9995, 0,5490 ve 0,3572 ile YSA-2, 0,9987, 1,0200 ve 0,5664 ile ANFIS-3 modeli olduğu belirlenmiştir. Bu verilere göre, deneysel sonuçlarla tahmin edilen sonuçlar arasındaki uyumun iyi olduğu ve hem YSA hem de ANFIS ile oluşturulan modellerle yüksek doğruluk derecesinde tahmin yapılabileceği kanaatine ulaşılmıştır.

Article Details

Prediction of Compressive Strengths of Silica Fume Substituted Cements Using ANN and ANFIS Models

Authors:

Yasemin Erdem | Eyyüp Gülbandılar | Yılmaz Koçak

Publisher:

Bursa Uludağ University

Main Topics:

Yazılım Testi | Doğrulama ve Validasyon

Journal:

Uludağ Üniversitesi Mühendislik Fakültesi Dergisi Vol.30 Issue 2, pp.355-374

Document Type:

Article

Publication Date:

8/2025

Document Language:

Turkish

Document ID (Org No.):

2001355787

pISSN:

2148-4155

Portland cement

Silica fume

Compressive strength

ANN

ANFIS.Portland çimento

Silis dumanı

Basınç dayanımı

YSA

ANFIS.

Issues

Article Details

Authors: Yasemin Erdem | Eyyüp Gülbandılar | Yılmaz Koçak

Publisher: Bursa Uludağ University

Main Topics: Yazılım Testi | Doğrulama ve Validasyon

Journal: Uludağ Üniversitesi Mühendislik Fakültesi Dergisi Vol.30 Issue 2, pp.355-374

Document Type: Article

Publication Date: 8/2025

Document Language: Turkish

Document ID (Org No.): 2001355787

pISSN: 2148-4155

Issues

Authors:

Yasemin Erdem | Eyyüp Gülbandılar | Yılmaz Koçak

Publisher:

Bursa Uludağ University

Main Topics:

Yazılım Testi | Doğrulama ve Validasyon

Journal:

Uludağ Üniversitesi Mühendislik Fakültesi Dergisi Vol.30 Issue 2, pp.355-374

Document Type:

Article

Publication Date:

8/2025

Document Language:

Turkish

Document ID (Org No.):

2001355787

pISSN:

2148-4155