Additive manufacturing helps designers to make better use of materials. It is essential to check the quality and consistency of 3D-printed parts because even small errors can affect how well they work and how safe they are. For finding errors in complicated shapes, traditional ways are difficult to implement. The aim of this research is to generate an approach which helps to find defects in 3D-printed parts. This method improves quality control in additive manufacturing by using both advanced deep learning methods for extraction of features and machine learning algorithms for classification. The study helps in finding defects in Fused Deposition Modeling (FDM) 3D-printed parts and sorts them into six groups: blobs, cracks, spaghetti, stringing, warping, and no defect. Combining these methods makes the system more accurate and better at generalization. The hybrid model combining EfficientNetB7 with the ensemble of Random Forest and XGBoost outperformed other models with a test accuracy of 79.83% and validation accuracy of 78.33%. This study helps in the research of AI-driven quality control in additive manufacturing. This will aim to make 3D-printed parts more reliable, which could be very useful in important fields like aerospace, automobile, and biomedical engineering.