Industrial defect detection has become an indispensable task in advanced manufacturing systems, as it directly influences product reliability, customer trust, and overall production efficiency. Although machine vision techniques have evolved considerably, achieving accurate detection of subtle and heterogeneous defects in X-ray images remains a significant challenge due to noise interference, complex background patterns, and variability in defect characteristics. In this study, we propose a hybrid Deep Learning (DL) framework that integrates Convolutional Neural Networks (CNNs) with tailored feature enhancement techniques to overcome these limitations. Unlike conventional CNN-based approaches, the proposed method emphasizes the refinement of discriminative features prior to classification, thereby improving detection robustness across diverse defect types and orientations. The experimental evaluation was conducted on a dataset of 3,500 X-ray images encompassing five major defect categories with varying scales and orientations, closely resembling real-world industrial conditions. The results demonstrate that the proposed framework achieves state-of-the-art performance, with an accuracy of 96.2%, precision of 95.5%, recall of 94.8%, and F1-score of 95.1%. Comparative analyses against existing methods further confirm the superiority of our approach in terms of both accuracy and generalization capability. Overall, this study contributes a reliable and scalable solution for automated quality inspection in industrial environments. The findings underline the practical applicability of the framework, offering manufacturers a cost-effective and efficient alternative to manual inspection while ensuring consistent quality assurance in large-scale production.