<![CDATA[Smart manufacturing’s (Industry 4.0) complexity demands automated quality control (AQC), as manual inspection is a major bottleneck. A critical gap exists in integrating “passive” Computer Vision (CV) detection with “active” mechatronic intervention, creating a “siloed” research problem. This research aims to design, develop, and validate a closed-loop AQC framework, integrating deep learning CV and mechatronics to autonomously perform the full QC cycle from detection to real-time physical intervention. An experimental systems integration design was employed. A Convolutional Neural Network (CNN) was trained on a 17,000-image dataset. A Robotic Operating System (ROS) framework was utilized as the integration layer for “hand-eye” calibration, synchronizing the CV node with a 6-axis robotic arm on a test rig. The CV model achieved 99.7% mAP (42ms latency) and calibration yielded ±0.35mm precision. The fully integrated system validation achieved a 99.15% Defect Detection Rate (DDR), a 0.11% False Positive Rate (FPR), and a 97.4% Successful Rejection Rate (SRR). The research empirically validates a holistic, closed-loop AQC framework, successfully solving the “siloed” gap. The system provides a proven, scalable blueprint for moving beyond passive detection to fully autonomous quality control in smart manufacturing.]]>
Copyrights © 2025
