Numerical Modeling for Prediction of Surface Morphology and Volumetric Defect Using Coupled Eulerian–Lagrangian Approach during Friction Stir Welding of Marine Grade Aluminum Alloy

Defective welds substantially hinder the quality of a welded joint, and choosing the best process parameter to ensure tunnel defect-free welds requires extensive experimentation. A coupled Eulerian–Lagrangian (CEL) finite element (FE) model was proposed to predict the defect formation, thermal history, and material flow behavior of the friction stir welding (FSW) process. The formation of defects was examined by varying tool rotation speed (TRS) between 600 and 1800 rpm and welding speed (WS) between 30 and 90 mm/min, with a constant tilt angle of 0°. It was evident that selecting appropriate process parameters established the condition for a tunnel defect-free joint. Less temperature of 300.2 °C can’t initiate proper material plasticization, leading to reduced material consolidation and more flash generation, thereby generating a long-stretched tunnel defect sample (S1). In contrast, a maximum temperature rise of 595.6 °C in the tunnel defect-free specimen (S2) ensures sufficient frictional contact between the tool-workpiece, thereby increasing material plasticization and enhancing material flow. Variation of less than 10% was found between the observed and simulated transient thermal histories. A minor percentage error of 5.44% in length and 1.28% in width was measured between the experimental and numerical tunnel defect morphology of S1. Contrarily, percentage errors of 4.16% in length and 3.75% in width were recorded for S2. Complete dynamic recrystallization in S2 hinders the formation of lamellar grain structures and flow segregation, leading to fine granular morphology. Maximum von Mises residual stress of 370 MPa was evident to spread all over the welded path in S1, thus hindering its service life. In comparison, 470 MPa were recorded at the crown of the chips of S2, which can be machined. A percentage error of 11.9% for sample S1 and 4.07% for S2 was marked when validating for average axial force data deviation. The proposed methodology was fairly well validated with the experimental trials.