Physics-Informed Deep Learning for Predicting Tensile Strength and Microhardness in Al/GO/MWCNT Composites: Addressing Experimental and Mechanistic Defects

This study investigates the tensile strength and microhardness of aluminum matrix composites supplemented with graphene oxide (GO) and multi-walled carbon nanotubes (MWCNTs), utilizing experimental characterisation, physics-based modeling, and deep learning prediction. To made hybrid Al/GO/MWCNT composites using powder metallurgy and extrusion methods. Then, to used microstructural analysis to find important characteristics including grain size, porosity, and dispersion index. Mechanical testing showed a big strengthening effect. The yield strength went from 181 MPa (unreinforced Al) to 270 MPa at 0.3 wt% GO and 0.5 wt% CNT, while the ultimate tensile strength (UTS) went from 200 MPa to 310 MPa. The microhardness also went up from 68 HV to 108 HV, showing that hybrid reinforcement works. Physics-based models, such as Halpin–Tsai, Orowan strengthening, and load-transfer mechanisms, shown strong concordance with experimental data, with prediction errors around 2%. A physics-informed deep learning framework was created to make things even more accurate. It combined mechanistic features with microstructural and process parameters. The suggested hybrid model did better than XGBoost, CatBoost, Random Forest, and TabTransformer baselines when it came to making predictions. It had MAE values of 5.2 MPa (UTS), 1.8 HV, and R² values of 0.98 and 0.97. The results show that combining experimental data, physics-based descriptors, and deep learning is a strong way to forecast properties that speeds up the design of new lightweight structural composites.