A Novel Hierarchical Extreme Machine-Learning-Based Approach for Linear Attenuation Coefficient Forecasting

Area of Science:

• Materials Science and Engineering
• Nuclear Engineering and Radiation Shielding
• Computational Science and Machine Learning

Background:

• High-energy photon shielding (X-rays, gamma-rays) is crucial in industrial and healthcare settings.
• Polymer composites and mineral admixtures offer potential for enhancing concrete's shielding capabilities.
• Traditional shielding calculations are often time-consuming and resource-intensive.

Purpose of the Study:

• To develop a dataset for assessing gamma-ray shielding behavior of concrete composites using magnetite and mineral powders.
• To investigate the efficacy of data-driven machine learning (ML) approaches as an alternative to theoretical calculations.
• To compare the performance of various ML models in predicting the linear attenuation coefficient (LAC) of concrete composites.

Main Methods:

• A dataset was created using magnetite and seventeen mineral powder combinations with concrete, varying densities and water/cement ratios.
• Photon cross-sections were computed using the National Institute of Standards and Technology (NIST) XCOM database to determine LAC.
• Multiple ML regressors, including Support Vector Machine (SVM), Convolutional Neural Network (CNN), and Hierarchical Extreme Learning Machine (HELM), were trained and evaluated.

Main Results:

• The Hierarchical Extreme Learning Machine (HELM) architecture significantly outperformed other ML models (SVM, CNN, Random Forest, etc.) in predicting LAC.
• HELM demonstrated strong consistency with XCOM-simulated LAC values, validated by stepwise regression and correlation analysis.
• The HELM model achieved the highest R2score and the lowest Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), indicating superior accuracy.

Conclusions:

• Data-driven ML techniques, particularly HELM, can effectively replicate and predict the gamma-ray shielding properties of concrete composites.
• The developed dataset and ML models provide a valuable, efficient alternative for assessing radiation shielding materials.
• HELM offers a promising approach for optimizing composite material design for enhanced radiation protection in various applications.