Parametric optimization of FSAM-Fabricated Al7075/Graphene/B4C hybrid composites using a Taguchi-ensemble machine learning framework

Friction Stir Additive Manufacturing (FSAM) avoids melting-related defects and is useful for repairing and building aluminium structures, but challenges remain with interlayer bonding, reinforcement dispersion, and surface wear. To address these, this study reinforced Al7075 with graphene and boron carbide (B4C). Graphene promotes load transfer, improves thermal conductivity and material flow (reducing tool/workpiece friction), and helps interlayer bonding B4C provides high hardness, wear resistance, and grain refinement. Using a groove-filling route and layer-by-layer stirring, two-layer Al7075/graphene/ B4C hybrid composites were fabricated. A Taguchi L16 design studied five process parameters: tool rotation (600–1200 rpm), traverse speed (20–80 mm/min), axial force (3–9 kN), tilt angle (0–3°), and shoulder-to-pin ratio (D/d = 3.0-4.5). Ultimate tensile strength (UTS) and Vickers hardness were the responses. The best condition (1200 rpm, 20 mm/min, 9 kN, 1° tilt, D/d = 4.0) gave UTS of 420 MPa and hardness of 160 HV. ANOVA showed tool rotation and shoulder-to-pin ratio as the most significant factors for both responses, with tilt angle important for defect suppression and layer bonding. To enhance prediction and optimization, ensemble machine-learning models (RF, GB, ET) were trained; all performed well (R2 > 0.98), with Gradient Boosting giving the lowest test errors (RMSE = 1.1 MPa for UTS and 0.64 HV for hardness). These results show that combining graphene and B4C reinforcements with FSAM, guided by Taguchi design and ML, offers a practical route to stronger and harder Al7075 components for aerospace, marine, and repair applications.