Internal defects such as voids/cavities significantly hamper the weld quality, resulting in lower weld efficiency and mechanical properties. Therefore, there is a scope for developing a relationship between quantum of defects and weld efficiency. This research uses a thermomechanical model based on coupled Eulerian-Lagrangian (CEL) to predict defect formation during friction stir welding (FSW) of a 3 mm thick AA2024. A square pin tool is used in current research. A novel concept of defect volume is developed using the Eulerian volume fraction within the weld domain, and a mathematical equation is developed to calculate weld efficiency as a function of defect volume, stir zone temperature, and axial force. The developed model predicted weld efficiency with less than 5 % error with R-squared coefficients of 0.9. The model established effective material flow behavior having an intercalated zig-zag path, resulting in defect-free weld with 86 % weld efficiency at 1500 rpm and 150 mm/min. It has a defect volume of 0.45 mm3. EBSD analysis shows an average grain size of 4.9 and 6.1 µm for defect-free and defective welds, respectively. Material flow, mechanical, and metallurgical characterization of lowest and highest weld efficiency is carried out to understand the underlying weld failure mechanism.