Optimization and predictive measurement of compressive strength of iron ore slag modified concrete using data-driven...

Area of Science:

• Materials Science
• Civil Engineering
• Data Science

Background:

• Concrete production contributes significantly to global CO2 emissions.
• Utilizing industrial wastes like glass powder, marble powder, and iron ore slag in concrete offers a sustainable alternative.
• Predicting the compressive strength (CS) of concrete with these supplementary cementitious materials is crucial for mix design.

Purpose of the Study:

• To develop an integrated machine learning-experimental framework for predicting the CS of concrete containing ternary industrial wastes.
• To evaluate the performance of various ML algorithms and hybrid models in predicting concrete CS.
• To identify key factors influencing concrete strength and develop a practical GUI for industrial application.

Main Methods:

• Compiled a dataset of 366 concrete mix ratios and CS values.
• Employed advanced ML algorithms: extreme gradient boosting (XGB), gradient boosting (GBR), and random forest (RF).
• Utilized hybrid models (XGB-GBR, XGB-RF) and validated predictions through experimental CS evaluation and scanning electron microscopy.

Main Results:

• The hybrid XGB-GBR model exhibited superior performance with R² values of 0.911 for training and 0.869 for testing.
• Feature importance analysis identified curing age and coarse aggregate as the most influential factors affecting CS prediction.
• An interactive graphical user interface (GUI) was developed for practical application.

Conclusions:

• The integrated ML-experimental framework effectively predicts the CS of concrete incorporating ternary industrial wastes.
• The developed GUI enhances the industrial applicability of ML-based concrete optimization, promoting cost reduction and sustainable practices.
• This research supports the efficient use of industrial wastes in concrete mix design, contributing to a circular economy.