Compressive Strength-Based Classification of Eco-Friendly Concretes Using Machine Learning Models.

Area of Science:

• Materials Science
• Sustainable Construction
• Artificial Intelligence

Background:

• Accurate compressive strength prediction is crucial for sustainable construction materials.
• Replacing cement with recycled glass powder presents challenges in material performance assessment.
• Eco-friendly concretes require reliable methods for quality control and optimization.

Purpose of the Study:

• To evaluate and compare five machine learning models for classifying compressive strength in concrete with recycled glass powder.
• To assess the viability of recycled glass powder as a sustainable cement substitute.
• To identify the most effective AI algorithms for optimizing low-carbon concrete performance.

Main Methods:

• Utilized a dataset of 846 experimental concrete samples with varying mix designs and curing ages.
• Compared the performance of Naïve Bayes, Random Forest, Decision Tree, Support Vector Machine (SVM), and k-Nearest Neighbors (k-NN) models.
• Analyzed model accuracy, generalizability, and interpretability for compressive strength classification.

Main Results:

• Naïve Bayes and Random Forest models demonstrated the highest accuracy and generalizability.
• Recycled glass powder incorporation did not lead to significant data instability.
• Decision Tree offered high interpretability regarding mixture parameter influence; SVM and k-NN excelled in extreme strength categories.

Conclusions:

• Probabilistic and ensemble learning methods are superior to deterministic and proximity-based algorithms for classifying variable concrete compositions.
• Artificial intelligence offers a reliable, scalable, and non-destructive tool for optimizing sustainable concrete performance.
• Recycled glass powder is a viable and sustainable alternative to traditional cement.