Perspective on the Application of Machine Learning Algorithms for Flow Parameter Estimation in Recycled Concrete...

Area of Science:

• Civil Engineering
• Materials Science
• Data Science

Background:

• Expanding construction necessitates sustainable practices, including waste management and the reuse of recycled concrete aggregate (RCA).
• The non-homogeneous nature of RCA presents challenges in accurately determining its engineering parameters, such as permeability.
• Machine learning offers potential solutions for efficient and reliable material characterization.

Purpose of the Study:

• To compare the effectiveness of k-Nearest Neighbors (k-NN) and Artificial Neural Network (ANN) algorithms for predicting the permeability coefficient of RCA.
• To evaluate the performance of these machine learning models against a linear regression baseline.
• To interpret the machine learning models using SHAP to understand feature importance in predicting RCA permeability.

Main Methods:

• Two distinct types of RCA were tested.
• Filtration tests were conducted on samples prepared with varying compaction energies (0.17 and 0.59 J/cm³).
• k-NN and ANN algorithms were trained and tested to predict the permeability coefficient, with results analyzed using the coefficient of determination (R²) and SHAP.

Main Results:

• The k-NN model achieved higher prediction accuracy, with R² values of 0.947 for training and 0.980 for testing data.
• The ANN model yielded R² values ranging from 0.877 to 0.936.
• SHAP analysis provided insights into the parameters influencing the prediction of RCA permeability, highlighting model interpretability.

Conclusions:

• The k-NN algorithm is highly effective for predicting the permeability coefficient of recycled concrete aggregate.
• Machine learning, particularly k-NN, offers a robust and accurate method for characterizing heterogeneous recycled materials in civil engineering.
• Interpretable AI methods like SHAP are crucial for understanding material behavior and validating predictive models in sustainable construction.