Improving Manufacturing Processes via Automated Inspection Advancements

Abstract

This project proposes a transformative solution to revolutionize manufacturing processes at Mansha Plastic Factory’s sack production unit through the implementation of an automated quality assurance system. The existing manual inspection method suffers from inefficiencies, delays, and increased labor costs, necessitating a more reliable and efficient approach. The primary goal is to design and deploy a real-time embedded re-trainable web system, leveraging advanced computer vision and deep learning techniques. The focal point is the creation of a system adept at accurately distinguishing between defective and non-defective sacks, thus elevating overall operational efficiency. Mansha Plastic Factory currently faces inefficiencies, limited capacity, and cost challenges in its manual inspection process. This project proposes a robust, fault-tolerant solution to address these issues, ensuring sustained quality assurance, cost-effectiveness, and long-term viability. The methodology encompasses crucial stages, including data collection, preprocessing, labeling, and the utilization of YOLO architecture for model selection. Further stages involve fine-tuning the model, implementing real-time inference, seamlessly integrating into the production process, and rigorous testing and validation. Key third-party tools, including YOLO, Python, OpenCV, labeling tools, documentation tools, and React for web application development, constitute the technological foundation of the project. The envisioned re-trainable automated inspection system promises multifaceted benefits, including enhanced quality assurance, operational efficiency, cost savings, consistency, reliability, data-driven decision-making, competitive advantage, and long-term sustainability within the manufacturing industry. The success of the project relies on a judicious blend of hardware and software tools, featuring automated inspection equipment, networking components, deep learning frameworks like YOLO, computer vision libraries such as OpenCV, labeling tools, Python as the primary programming language, database management systems, and documentation tools. This comprehensive approach positions the project as a transformative force in modernizing manufacturing quality control processes.