MACHINE LEARNING-ASSISTED HOMOGENIZATION APPROACH FOR PREDICTING THE ELASTOPLASTIC BEHAVIOR OF SHORT FIBER-REINFORCED COMPOSITES
INTERNATIONAL CONGRESS ON NEW TRENDS IN MECHANICS (ICNTM25), Baku, Azerbaycan, 2 - 05 Eylül 2025, cilt.1, sa.1, ss.829-842, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 1
- Basıldığı Şehir: Baku
- Basıldığı Ülke: Azerbaycan
- Sayfa Sayıları: ss.829-842
- Bursa Uludağ Üniversitesi Adresli: Evet
Özet
Short fiber-reinforced composites (SFRCs) are widely used in industries such as automotive, aerospace, and industrial manufacturing due to their lightweight structure and high mechanical strength. However, the anisotropic and nonlinear mechanical behavior resulting from fiber-matrix interactions presents significant challenges in predicting their response under various loading conditions. Traditional computational homogenization techniques are commonly used to estimate the macroscopic mechanical properties of SFRCs. However, in recent years, machine learning (ML) methods have gained increasing attention for enhancing the efficiency and accuracy of these processes. In this study, the effectiveness of advanced ML models in predicting the elastoplastic behavior of SFRCs was evaluated. The applied ML models included Random Forest (RF), Extreme Gradient Boosting (XGBoost), Artificial Neural Networks (ANN), and Support Vector Regression (SVR). The training dataset was generated using Sobol sampling, systematically varying nine key micromechanical parameters that influence the homogenized material response. The predictive performances of the ML models were assessed using statistical metrics such as R², MAE, MAPE, and RMSE. The findings demonstrate that ML-based approaches significantly reduce computational time while maintaining high accuracy, making them a reliable alternative for engineering applications