Forecasting the Dielectric Properties of Obsidian Containing Cement Mortars Using Distinct Approaches in Machine Learning
Iranian Journal of Science and Technology - Transactions of Civil Engineering, 2026 (Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1007/s40996-026-02195-z
- Dergi Adı: Iranian Journal of Science and Technology - Transactions of Civil Engineering
- Derginin Tarandığı İndeksler: Scopus, ABI/INFORM, INSPEC, Middle East & Africa Database (ProQuest), Natural Science Collection (ProQuest), Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
- Anahtar Kelimeler: Dielectric properties, Machine learning, Obsidian, Supplementary cementitious materials
- Recep Tayyip Erdoğan Üniversitesi Adresli: Evet
Özet
Concrete is one of the well-nigh universally utilized construction materials thanks to its strong mechanical strength, long life, and resistance to chemical effects. It is also inexpensive and easy to get. However, concrete is also used for functional purposes in addition to structural ones. In this context, dielectric properties critically impact electromagnetic properties, especially electromagnetic shielding. Additionally, dielectric properties provide information about the material’s physical properties, such as moisture content, void structure, and microstructure. Therefore, quickly and reliably forecasting the dielectric features of building materials is important for their performance. This research is of critical importance because conventional laboratory measurements for determining these functional properties are time consuming and costly, making rapid, nondestructive evaluations essential for modern infrastructure. This examination utilized different machine learning frameworks to foretell the dielectric features of cementitious mortars with pozzolanic properties, such as obsidian (OB), fly ash (FA), and Ground Granulated Blast Furnace Slag (GGBS). The examination employed algorithms with distinct machine learning perspectives RF, DT, GB, SVR, and MLP. The study predicted the 28 day dielectric properties of mortar samples. The study identified complex relationships between important variables, such as mixing parameters, and dielectric properties. The results from the study suggest that Gradient Boosting had the best overall predictive capability for this application as seen by the R2, R2training and RMSE values of 0.93, 0.99, and 0.00345, respectively. Second place was Random Forest with a 0.86 R2 value, 0,00503 RMSE value, and Decision Tree model showed relatively good performance with a values of 0.89 for R2 and 0,00457 for RMSE. All ensemble models were found to have made most of their predictions within a close margin of error ± 10% – ±25%. SVR provided the least accurate predictions with the greatest amount of error, RMSE = 0.00912 and an R2 = 0.55. This investigation exhibits an innovative, data driven approach for forecasting the dielectric aspects of cementitious mortars. The primary advantage of this methodology is its capability to deliver highly accurate predictions that dramatically reduce the need for physical testing, thereby saving significant time and resources. Ultimately, the examination’s findings reveal significant potential for environmentally friendly building materials with electromagnetic shielding properties. The study’s novelty lies in its pioneering use of machine learning to map the high frequency (2.45 GHz) dielectric behavior of sustainable obsidian blended mortars, providing a rapid alternative to labor intensive physical testing.