Performance Evaluation of ANN and Ensemble Learning Methods in Predicting Wear Properties of Porcelain Ceramic Composites

Articles in Press

Document Type : Article

Author

SCU Artificial Intelligence and Data Science Application and Research Center, University of Sivas Cumhuriyet, Sivas, 58140, Turkiye; SCU Nanophotonic Application Research Center, Sivas, Turkiye

10.24200/sci.2025.64976.9228

Abstract

This study extensively investigated the fabrication and wear properties of aluminum titanate (Al₂TiO₅) and mullite (3Al₂O₃-2SiO₂) doped porcelain ceramic-composites produced by the powder metallurgy method. The porcelain ceramics were prepared by powder metallurgy and the wear resistance and other mechanical properties were evaluated based on the data obtained. The experimental wear results were modeled and analyzed using ensemble learning(EL) methods and Artificial Neural Networks (ANN). Boosting and random forest(RF) algorithms were employed among the ensemble methods. Basic statistical measures such as R², RMSE, MAE, and MAPE were utilized to evaluate model performance. Boosting and RF methods also produced successful results, but ANN was found to be superior in terms of accuracy and overall performance. In the study, pure porcelain (P), mullite doped porcelain (PM), aluminum titanate doped porcelain (PAT) and aluminum titanate-mullite doped (PMAT) porcelain models were investigated and compared separately. The findings provide valuable contributions to the development of high-performance ceramic-composites in materials engineering and optimization of the wear behavior of these materials. In this context, the applicability of advanced machine learning methods in materials science and the advantages of these methods are discussed in detail.

Keywords

Main Subjects

Scientia Iranica

Articles in Press, Accepted Manuscript
Available Online from 20 January 2025

Files

History

  • Receive Date: 22 July 2024
  • Revise Date: 24 November 2024
  • Accept Date: 20 January 2025

Share

How to cite

Statistics