Timely detection of failures in critical components of internal combustion engines is one of the key factors in extending engine lifetime and reducing repair and maintenance costs. In this study, an intelligent machine vision-based system was designed and implemented to detect failures (wear) in Hydraulic Lash Adjusters (HLA). The main objective of this system is to replace traditional and costly inspection methods, such as coordinate measuring machines, with a faster, more accurate, and cost-effective approach. The dataset consisted of color images of normal and defective parts. After a preprocessing stage including resizing, normalization, and data augmentation, Resnet18 was employed for Normal/Damaged classification, while a U-Net architecture was used for defect segmentation. The proposed system was evaluated using accuracy, recall, precision, and overlap-based metrics. Results demonstrated that the proposed model, compared with traditional methods, not only achieved higher accuracy (over 97.6% in classification and final Dice Score of 0.750) but also reduced inspection time by more than 70%. Furthermore, a comparison between the system’s output and expert human evaluation showed high consistency, confirming the reliability and practical applicability of the proposed method.