Predictive modelling of tensile strength and hardness of microwave assisted AA7075/Sic composite through supervised machine learning algorithms

• ML models accurately predict tensile strength and hardness of AA7075/SiC. • KNN best predicts tensile strength; SVM excels for hardness. • Optimal range: 7% SiC, 800 MPa, 550°C, 120 min sintering. • Compaction pressure drives strength; SiC governs hardness. • Residuals and SHAP confirm model reliability and accuracy. Integrating machine learning (ML) into materials science is crucial for predicting and optimizing material properties. This study applies ML methods to analyze the properties of microwave-assisted AA7075/SiC composites. Composite powders were fabricated under consistent parameters, ensuring uniform reinforcement dispersion, confirmed by scanning electron microscopy (SEM). The dataset included 172 points with independent variables such as SiC composition, compaction pressure, sintering temperature, and time, alongside tensile strength and hardness as dependent variables. Five regression models were used to monitor mechanical properties, identifying optimal ranges: SiC composition around 6–7 wt.%, compaction pressure in the range of 750–800 MPa, sintering temperature near 540–560°C, and sintering time between 115–125 minutes. Hyperparameters were tuned using GridSearchCV, and Ridge regularization mitigated overfitting, resulting in high predictive accuracy across models. The K-Nearest Neighbors (KNN) model provided the best tensile strength predictions (R² = 0.9688), while the Support Vector Machine (SVM) model excelled in hardness predictions (R² = 0.9870). Residual diagnostics indicated minimal autocorrelation and well-behaved residuals, confirming model reliability. Feature importance analysis revealed compaction pressure as the most influential factor on tensile strength, while SiC composition significantly affected hardness. This study demonstrates the potential of ML to enhance material property prediction and optimization.