Prediction of concrete compressive strength using a Deepforest-based model.

Area of Science:

• Civil Engineering
• Materials Science
• Data Science

Background:

• Traditional concrete compressive strength testing is time-consuming and labor-intensive.
• Machine learning (ML) offers a promising alternative for predicting concrete strength.
• A comprehensive comparison of ML models is needed to identify the optimal predictor.

Purpose of the Study:

• To develop and compare 12 distinct ML-based regressors for concrete compressive strength prediction.
• To identify the optimal ML model for accurate strength prediction.
• To analyze the correlation between strength and key constituent factors.

Main Methods:

• Developed 12 ML regressors, including Deepforest.
• Selected key factors: blast furnace slag, superplasticizer, age, cement, and water.
• Utilized grid search and fivefold cross-validation for hyperparameter tuning.

Main Results:

• The Deepforest-based model outperformed the other 11 ML models.
• Achieved a high performance metric (R² of 0.91) on an independent testing dataset.
• Demonstrated superior accuracy compared to state-of-the-art models.

Conclusions:

• The Deepforest model is the optimal predictor for concrete compressive strength.
• This ML approach significantly enhances the efficiency and accuracy of construction quality control.
• The findings provide a robust foundation for implementing advanced predictive analytics in civil engineering.