Comparative Study of Supervised Machine Learning Algorithms for Predicting the Compressive Strength of Concrete at...

Area of Science:

• Materials Science
• Civil Engineering
• Computer Science

Background:

• High temperatures negatively impact concrete's ingredient properties, reducing its strength.
• Predicting concrete compressive strength is challenging and time-consuming.
• Supervised machine learning offers a viable solution for accurate strength prediction.

Purpose of the Study:

• To forecast the compressive strength of concrete exposed to high temperatures.
• To evaluate the performance of various machine learning algorithms for this prediction task.
• To identify the most effective machine learning approach for concrete strength assessment.

Main Methods:

• Utilized supervised machine learning models: Decision Tree (DT), Artificial Neural Network (ANN), Bagging, and Gradient Boosting (GB).
• Employed Python coding in Anaconda Navigator with 207 data points.
• Input parameters included water, cement, aggregates, admixtures, and temperature; output was compressive strength.

Main Results:

• Ensemble algorithms (Bagging and Gradient Boosting) achieved high R² values (0.90 and 0.88, respectively).
• Individual DT and ANN models showed R² of 0.83 and 0.82.
• Statistical indicators (R², MAE, MSE, RMSE) and k-fold cross-validation confirmed the superior performance of ensemble methods.

Conclusions:

• Ensemble machine learning algorithms significantly enhance the prediction accuracy of concrete compressive strength at high temperatures.
• Sensitivity analysis indicated the contribution of each input variable to the model's performance.
• Machine learning provides an efficient tool for predicting concrete performance under thermal stress.