Application of Artificial Intelligence Methods for Predicting the Compressive Strength of Self-Compacting Concrete...

Area of Science:

• Civil Engineering
• Materials Science
• Computer Science

Background:

• Class F fly ash can replace cement in concrete, promoting sustainability and reducing greenhouse gas emissions.
• Accurate prediction of compressive strength is crucial for utilizing fly ash in self-compacting concrete (SCC).

Purpose of the Study:

• To develop and compare machine learning models for predicting the compressive strength of SCC containing Class F fly ash.
• To identify the optimal predictive model for SCC with fly ash replacement.

Main Methods:

• Utilized a dataset of 327 experimentally tested SCC samples.
• Developed and evaluated models including regression trees (RTs), Gaussian process regression (GPR), support vector regression (SVR), and artificial neural networks (ANNs).
• Analyzed both individual and ensemble model accuracies using Mean Absolute Error (MAE) and correlation coefficient (R).

Main Results:

• An ensemble of ANNs demonstrated the highest prediction accuracy with MAE of 4.37 MPa and R of 0.96.
• Individual multi-gene genetic programming (MGGP) and regression tree (RT) models showed comparable or superior accuracy to individual ANN models.
• MGGP model achieved MAE of 5.70 MPa and R of 0.93; RT model achieved MAE of 6.64 MPa and R of 0.89.

Conclusions:

• Machine learning, particularly ensemble ANNs, can accurately predict the compressive strength of SCC with fly ash.
• Transparent models like MGGP and RT offer viable alternatives with competitive predictive performance.
• Accurate prediction facilitates the increased use of fly ash in SCC, contributing to sustainable construction practices.