Machine learning-based predictive modeling of machining forces and temperatures in alumina-reinforced jute/epoxy functional composites
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY, vol.144, no.7-8, pp.5217-5237, 2026 (SCI-Expanded, Scopus)
- Publication Type: Article / Article
- Volume: 144 Issue: 7-8
- Publication Date: 2026
- Doi Number: 10.1007/s00170-026-18181-8
- Journal Name: INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY
- Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, IBZ Online, Compendex, INSPEC, DIALNET
- Page Numbers: pp.5217-5237
- Keywords: Alumina reinforcement, Cutting forces, Data-driven modeling, Drilling process, Machining temperature, Natural fiber composites, Process-material interaction, Surface integrity
- Recep Tayyip Erdoğan University Affiliated: No
Abstract
This study develops a robust, multi-algorithm, performance-weighted ensemble modeling framework to predict cutting forces and temperatures during the machining of baseline jute/epoxy and 10% alumina-reinforced jute/epoxy functional composites. The framework integrates an enhanced empirical modeling, optimized Support Vector Regression (SVR), and customized Neural Networks (NNs). Feature engineering incorporated physics-informed terms to capture non-linear relationships between cutting parameters (feed rate and spindle speed) and responses. The ensemble model dynamically weights each algorithm's prediction based on its coefficient of determination (R & sup2;). Experimental results revealed a dramatic 1510% increase in cutting force and a significant 19% increase in mean temperature for the hybrid composite, highlighting the profound impact of abrasive alumina particles on machinability. In addition, three-dimensional surface roughness measurements were employed as a complementary qualitative assessment, revealing distinct surface integrity responses between the baseline and hybrid composites and underscoring the limitations of average roughness parameters in capturing localized drilling-induced damage. The proposed ensemble framework meets the predefined accuracy threshold, achieving R & sup2; values exceeding 0.97 for both cutting forces and temperatures across both materials. Overall, this work provides a reliable data-driven tool for industrial process optimization, enabling improved management of the elevated mechanical and thermal stresses associated with machining alumina-reinforced natural fiber composites.