<p>Forging, as a fundamental manufacturing process in mechanical engineering, finds extensive applications across aerospace, marine equipment, automotive manufacturing, and other sectors. However, defects that frequently occur during the forging process significantly impact product quality and performance. This paper systematically reviews forging defect prediction methods and optimization strategies, with a particular focus on the application of finite element methods (FEM) in defect prediction and process optimization. Through detailed analysis of three typical defects—cracks, underfilling, and folding—the following prediction methods are summarized: (1) crack defects are primarily predicted based on ductile fracture criteria, with the Cockcroft–Latham damage theory being the most widely applied; (2) underfilling defect prediction methods include surface observation, billet-die contact detection, and minimum distance calculation; (3) folding defect prediction methods are the most diverse, encompassing surface wrinkle observation through finite element simulation, surface discontinuity analysis, the Euclidean distance method, flow field convergence analysis, folding index analysis, and characteristic value analysis. Furthermore, this paper discusses strategies for eliminating defects through optimization of billet shape, die structure, and process parameters. Finally, the limitations of current prediction methods are identified, and future research directions are proposed. These research findings provide important theoretical guidance for improving forging process quality and product performance.</p>