MACHINE LEARNING-AIDED OPTIMIZATION OF ULTRA-HIGH PERFORMANCE CONCRETE COMPRESSIVE STRENGTH. Masters thesis, King Fahd University of Petroleum and Minerals.

PDF (MSc Thesis) Muhammed Balogun_Msc_Thesis_Revise.pdf - Accepted Version Restricted to Repository staff only until 30 January 2026. Download (2MB)

Arabic Abstract

‬‬‫الخرسانة‬‫فائقة‬ ‫األداء‬ ‫(‬ ‫‪UHPC‬‬‫)‬‫معروفة‬ ‫بخصائصها‬ ‫الميكانيكية‬ ‫الفائقة‬ ‫والمتانة‬ ‫في‬ ‫تطبيقات‬ ‫البنية‬ ‫التحتية‪.‬‬ ‫يُعتبر‬ ‫التنبؤ‬ ‫أمر‬‫ا‬‫بالغ‬ ‫األهمية‬ ‫كمرحلة‬ ‫سابقة‬ ‫إلجراءات‬ ‫تصميم‬ ‫الهياكل‪.‬‬ ‫تُعد‬ ‫الصيغ‬ ‫التجريبية‬ ‫التقليدية‬ ‫بمقاومة‬‫االنضغاط‬ ‫للـ‬ ‫‪UHPC‬‬ ‫ً‬ ‫طويًل‬‫‪،‬‬‫مما‬ ‫يستدعي‬ ‫الحاجة‬ ‫إلى‬ ‫نهج‬ ‫أكثر‬ ‫كفاءة‪.‬‬ ‫تستكشف‬ ‫هذه‬ ‫الدراسة‬ ‫جدوى‬ ‫للـ‬‫‪UHPC‬‬ ‫غالبًا‬ ‫مجهدة‬ ‫وتستغرق‬ ‫وقتًا‬ ‫استخدام‬‫نماذج‬ ‫التعلم‬ ‫اآللي‬ ‫(‬ ‫‪ML‬‬‫)‬‫لتحسين‬ ‫وتوقع‬ ‫مقاومة‬ ‫االنضغاط‬ ‫للـ‬ ‫‪UHPC‬‬ ‫‪.‬‬‫تم‬ ‫استخدام‬ ‫عدة‬ ‫نماذج‬ ‫للتعلم‬ ‫اآللي‬ ‫في‬‫عملية‬ ‫التنبؤ‪،‬‬ ‫بما‬ ‫في‬ ‫ذلك‬ ‫نموذج‬ ‫الجيران‬ ‫األقرب‬ ‫(‬ ‫‪KNN‬‬‫)‪،‬‬‫شجرة‬ ‫القرار‬ ‫(‬ ‫‪DT‬‬‫)‪،‬‬‫الغابة‬ ‫العشوائية‬ ‫(‬ ‫‪RF‬‬‫)‪،‬‬‫تعز‬ ‫يز‬ ‫التدرج‬‫المتطرف‬ ‫(‬ ‫‪XGB‬‬‫)‪،‬‬‫آلة‬ ‫المتجهات‬ ‫الداعمة‬ ‫(‬ ‫‪SVM‬‬‫)‪،‬‬‫باإلضافة‬ ‫إلى‬ ‫االنحدار‬ ‫الخطي‪.‬‬ ‫ألغراض‬ ‫النمذجة‪،‬‬ ‫تم‬ ‫جمع‬‫بيانات‬ ‫الـ‬ ‫‪UHPC‬‬ ‫بنا‬ ‫ً‬ ‫ء‬‫على‬ ‫معايير‬ ‫تصميم‬ ‫الخليط‬ ‫من‬ ‫األدبيات‬ ‫مفتوحة‬ ‫المصدر‬ ‫والتي‬ ‫شملت‬ ‫‪810‬‬ ‫مًلحظة‬ ‫و‬ ‫‪15‬‬ ‫معلمة‬‫إدخال‬ ‫"ميزات"‪.‬‬ ‫في‬ ‫البداية‪،‬‬ ‫تم‬ ‫اعتماد‬ ‫تقنية‬ ‫اختي‬ ‫ار‬‫الميزات‬ ‫بنا‬ ‫ً‬ ‫ء‬‫على‬ ‫تصنيف‬ ‫االرتباط‬ ‫لتقليل‬ ‫معلمات‬ ‫اإلدخال‬ ‫من‬‫‪15‬‬ ‫إلى‬ ‫‪8‬‬ ‫‪،‬‬‫مما‬ ‫أدى‬ ‫إلى‬ ‫تحسين‬ ‫التنبؤ‬ ‫بمقاومة‬ ‫االنضغاط‬ ‫للـ‬ ‫‪UHPC‬‬ ‫‪.‬‬‫تم‬ ‫استخدام‬ ‫مقاييس‬ ‫تقييم‬ ‫النماذج‪،‬‬ ‫بما‬ ‫في‬ ‫ذلك‬ ‫متوسط‬‫الخطأ‬ ‫المطلق‬ ‫(‬ ‫‪MAE‬‬‫)‪،‬‬‫الجذر‬ ‫التربيعي‬ ‫لمتوسط‬ ‫الخطأ‬ ‫المربع‬ ‫(‬ ‫‪RMSE‬‬‫)‪،‬‬‫ومعامل‬ ‫التحديد‬ ‫(‬ ‫‪²‬‬‫‪R‬‬‫)‪،‬‬‫لتقييم‬ ‫دقة‬ ‫توقع‬‫النموذج‪.‬‬ ‫أشارت‬ ‫النتائج‬ ‫إلى‬ ‫أن‬ ‫نماذج‬ ‫الـ‬ ‫‪RF‬‬ ‫والـ‬ ‫‪XGB‬‬ ‫تفوقت‬ ‫على‬ ‫النماذج‬ ‫األخرى‬ ‫في‬ ‫دقة‬ ‫التنبؤ‪.‬‬ ‫وتؤكد‬ ‫هذه‬ ‫النتائج‬‫على‬ ‫إمكانيات‬ ‫نماذج‬ ‫التعلم‬ ‫اآللي‬ ‫في‬ ‫تحسين‬ ‫كفاءة‬ ‫وفعالية‬ ‫تكلفة‬ ‫تصميم‬ ‫الخليط‬ ‫للـ‬ ‫‪UHPC‬‬ ‫بشكل‬ ‫كبير‪.‬‬ ‫تقدم‬ ‫هذه‬ ‫إطار‬‫ا‬‫قويًا‬ ‫يعتمد‬ ‫على‬ ‫الحاسوب‬ ‫يمكن‬ ‫استخدامه‬ ‫لتحسين‬ ‫تصميم‬ ‫الـ‬ ‫‪UHPC‬‬ ‫بسرعة‬ ‫ودقة‪،‬‬ ‫مما‬ ‫يمهد‬ ‫الطريق‬ ‫الدراسة‬ ‫لتطبيقه‬‫بشكل‬ ‫أوسع‬ ‫في‬ ‫صناعة‬ ‫البناء‬ ‫‪.‬‬

English Abstract

Ultra-high-performance concrete (UHPC) is known for its superior mechanical properties and durability for infrastructure applications. The prediction of UHPC compressive strength is crucial in the preliminary stages of structural design. Traditional empirical mixture formulations for non-proprietary UHPC are often labor-intensive and time- consuming, highlighting the need for more efficient approaches. This study explores the feasibility of machine learning (ML) models to optimize and predict the UHPC compressive strength. Several ML models were employed for the prediction process, including K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB), Support Vector Machine (SVM), and Linear Regression. For the modeling purpose, a dataset based on UHPC mix design parameters was collected from the open-source literature, comprising 810 observations with 15 input parameters (“features”). At first, the correlation rank feature selection technique was adopted to reduce the input parameters from 15 to 8, thereby optimizing the prediction of UHPC compressive strength. Model evaluation metrics, including Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and determination coefficient (R²), were used to assess the model’s predictive accuracy. Results indicated that RF and XGB models outperformed other models in predictive accuracy. The findings underscore the potential of ML models to enhance the efficiency and cost-effectiveness of UHPC mix design significantly. This study provides a robust computer-aided framework for rapid and precise optimization of UHPC, paving the way for its broader application in the construction industry.

Item Type:	Thesis (Masters)
Subjects:	Civil Engineering Civil Engineering > Structural Engineering
Department:	College of Design and Built Environment > Civil and Environmental Engineering
Committee Advisor:	Adekunle, Saheed K.
Committee Members:	Olatunji, Sunday O. and Yaseen, Zaher M.
Depositing User:	MUHAMMED BALOGUN (g202110150)
Date Deposited:	09 Feb 2025 05:31
Last Modified:	09 Feb 2025 05:31
URI:	http://eprints.kfupm.edu.sa/id/eprint/143274