Automated Flange Inspection using AI-Powered Vision System

This paper presents an AI-integrated vision-based inspection system for automated quality assessment of industrial flange components. The proposed framework utilizes a YOLOv5s object detection model trained on a custom dataset to identify common manufacturing defects such as hole misalignment and missing apertures. Real-time image acquisition is performed using a fixed-position USB webcam aligned above a motor-driven conveyor system. Preprocessing of the captured frames is executed using Canny Edge Detection to enhance contour clarity before classification. Upon detection, the system leverages OpenCV to synchronize the conveyor movement, halting for precise frame analysis. The control logic, executed on a Raspberry Pi 3B, interprets model output and actuates the sorting mechanism via GPIO-controlled relays and a servo motor. Components classified as defective are automatically diverted to a rejection path, ensuring uninterrupted flow of compliant units. This simple system is compact, cost-effective, and reduces human interaction, allowing faster inspection, as well as consistent evaluation of quality.