MINISTRY OF EDUCATION AND TRAINING HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY AND EDUCATION GRADUATION THESIS MAJOR: MACHINE MANUFACTURING TECHNOLOGY DESIGN AND FABRICATION OF A VISUAL INSPECTION WORKSTATION FOR MANUAL ARC WELDING APPLIED IN WELDING TRAINING INSTRUCTOR: TRAN NGOC THIEN,ME. STUDENT: TONG HUYNH TANH TRAN VAN THOAI LUONG HOANG HIEN Ho Chi Minh city, July 2024 MINISTRY OF EDUCATION AND TRAINING HCMC UNIVERSITY OF TECHNOLOGY AND EDUCATION FACULTY OF MECHANICAL ENGINEERING GRADUATION THESIS DESIGN AND FABRICATION OF A VISUAL INSPECTION WORKSTATION FOR MANUAL ARC WELDING APPLIED IN WELDING TRAINING Supervisor: Tran Ngoc Thien, M.E Students: ID: Class: Tong Huynh Tanh 20134024 20134A Tran Van Thoai 20134025 20134A Luong Hoang Hien 20143069 20143CL5B Year of Admission: 2020-2024 Ho Chi Minh City, July 2024 HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY AND EDUCATION FACULTY OF MECHANICAL ENGINEERING ---------------------------------- DEPARTMENT OF MECHATRONICS GRADUATION THESIS DESIGN AND FABRICATION OF A VISUAL INSPECTION WORKSTATION FOR MANUAL ARC WELDING APPLIED IN WELDING TRAINING Supervisor: Tran Ngoc Thien, M.E Student: TONG HUYNH TANH Student ID: 20134024 Student: TRAN VAN THOAI Student ID: 20134025 Student: LUONG HOANG HIEN Student ID: 20143069 Class: 20134A, 20143CL5B Year of Admission: 2020 - 2024 Ho Chi Minh City, July 2024 TRƯỜNG ĐẠI HỌC SƯ PHẠM KỸ THUẬT TP. HCM CỘNG HOÀ XÃ HỘI CHỦ NGHĨA VIỆT NAM KHOA CƠ KHÍ CHẾ TẠO MÁY Độc lập - Tự do - Hạnh phúc NHIỆM VỤ ĐỒ ÁN TỐT NGHIỆP Học kỳ II / năm học 2023-2024 Giảng viên hướng dẫn: ThS. Trần Ngọc Thiện Sinh viên thực hiện: Tống Huỳnh Tánh MSSV: 20134024 Điện thoại: 0916716095 Sinh viên thực hiện: Trần Văn Thoại MSSV: 20134025 Điện thoại: 0354408059 Sinh viên thực hiện: Lương Hoàng Hiện MSSV: 20143069 Điện thoại: 0962269011 1.
Đề tài tốt nghiệp: - Mã số đề tài: CKM-155 - Tên đề tài: Thiết Kế, Chế Tạo Bàn Thao Tác Đánh Giá Ngoại Quan Mối Hàn Hồ Quang Tay Ứng Dụng Trong Đào Tạo Hàn 2. Các số liệu, tài liệu ban đầu: - Hình ảnh mối hàn hồ quang tay của sinh viên. - Tài liệu về các lỗi trong hàn hồ quang tay. Nội dung chính của đồ án: - Thiết kế và chế tạo bàn quan sát mối hàn.
- Xây dựng mô hình AI để phân loại và phát hiện lỗi của mối hàn. - Xây dựng giao diện thao tác với mối hàn. Các sản phẩm dự kiến - Báo cáo thuyết trình và sản phẩm 5. Ngôn ngữ trình bày: Bản báo cáo: Tiếng Anh Tiếng Việt Trình bày bảo vệ: Tiếng Anh Tiếng Việt TRƯỞNG BỘ MÔN GIẢNG VIÊN HƯỚNG DẪN (Ký, ghi rõ họ tên) (Ký, ghi rõ họ tên) Được phép bảo vệ …………………………………………… (GVHD ký, ghi rõ họ tên) i COMMITMENT - Project: DESIGN AND FABRICATION OF A VISUAL INSPECTION WORKSTATION FOR MANUAL ARC WELDING APPLIED IN WELDING TRAINING - Supervisor: Tran Ngoc Thien, M.E Students: ID: Phone Number: Tong Huynh Tanh 20134024 0916716095 Tran Van Thoai 20134025 0354408059 Luong Hoang Hien 20143069 0962269011 - Graduation thesis submission date: 07/2024 - Commitment: “We hereby certify that this graduation thesis is our own research and work.
We have not copied from any published articles without proper citation. If there is any violation, we will take full responsibility.” ii ACKNOWLEDGMENTS We wish to express our sincere gratitude to Ho Chi Minh City University of Technology and Education, with special acknowledgments to the esteemed members of the Mechanical Engineering Faculty. Their unwavering guidance and support have been invaluable throughout our academic journey, playing a crucial role in the successful completion of a significant portion of our training program. The exceptional instruction provided by the faculty has not only equipped us with specialized knowledge but has also significantly shaped our understanding and approach to academic pursuits.
A profound appreciation is extended to M.E Tran Ngoc Thien, whose expertise and guidance have been instrumental in navigating the complexities of this project. His insightful contributions and dedicated mentorship have enriched the depth and quality of our research. Mistakes are an inherent part of any academic endeavor, and we welcome any constructive criticism and recommendations from the academic community to enhance the overall completeness of this report. Your valuable insights will contribute to the refinement and improvement of our work.
Lastly, we express our gratitude to the Faculty of Mechanical Engineering for their generous funding support, which has played a pivotal role in enabling us to undertake and successfully complete this research. Their commitment to fostering academic exploration is deeply appreciated. Sincerely, Tong Huynh Tanh Tran Van Thoai Luong Hoang Hien iii ABSTRACT The assessment of weld quality in universities is an integral part of welding engineering education, ensuring that students grasp the knowledge and skills required for proficiency in welding. Developing a weld assessment system yields practical benefits such as enhancing the quality of welding training, minimizing errors, and increasing objectivity in evaluations.
Recognizing the need for such a system, our team has chosen to undertake the project "Design and Fabrication of a Visual Inspection Workstation for Manual Arc Welding Applied in Welding Training." The objective of this project is to develop a system capable of receiving input from a student's manual arc weld product and providing results on the quality, defects present in the weld, and suggesting causes and remedies based on the severity and type of defect. The system will consist of compact and convenient hardware with full functionality, and AI-powered weld assessment software with high accuracy and a user-friendly interface. As the initial step in implementing the system, our team collected image data of welds produced by students. We analyzed the characteristics of this collected image dataset and consulted relevant technical literature to contribute to the development of a highly accurate weld assessment software system.
Regarding the hardware, our research focused on ensuring the quality of the input image, providing a display for results, and maintaining an aesthetically pleasing design. The results of this project demonstrate that the system achieves high accuracy in classifying weld quality, detecting defects, and suggesting corrective actions. This product provides students with detailed, rapid, and objective feedback, effectively supporting their learning and practice. iv TABLE OF CONTENTS NHIỆM VỤ ĐỒ ÁN TỐT NGHIỆP.
iv TABLE OF CONTENTS. v LIST OF TABLES. viii LIST OF FIGURES. ix LIST OF ABBREVIATIONS.
Scientific and practical significances. Scope of the study. Structure of the report. Theoretical basis of manual metal arc welding (MMAW).
Artificial intelligence (AI) and related terms. Artificial intelligence, machine learning (ML) and deep learning (DL) 16 2. Computer vision and Convolution Neural Network (CNN). Image Classification and Vision Transformer (ViT).
Object Detection and YOLOv8. Retrieval augmented generation (RAG) in large language models (LLMs). Data transmission and processing via cloud server. Data transmission method.
Google Cloud run. DESIGN MECHANICAL SYSTEM. Mechanical Design Process. Weld placement block.
Mechanical system summary. DESIGN ELECTRONICS-CONTROL SYSTEM. System electrical requirements. Electrical System Connection Diagram.
The selection of devices. DESIGN AI SYSTEM. Criteria for evaluating weld quality. Weld quality classification.
Weld defects detection. AI process overview. Developing weld quality classification model. Model evaluation metric.
Developing weld defect detection model. Model Evaluation Metrics. Developing weld feedback generating model. Developing data transmission and processing feature.
Developing graphical user interface. Overall System Result. CONCLUSION AND RECOMMENDATION. 83 vii LIST OF TABLES Table 2.
Welding rod diameter reference. Camera setups comparison. Comparison of common lighting types used for cameras. Frame materials Comparison.
Cover materials comparison. Jetson Nano Specification. Miniature Circuit Breaker specification. AC-DC power supply specification.
Buck Converter specification. Relay module specification. Classify model accuracy comparison. 68 viii LIST OF FIGURES Figure 1.
Metal welding class at vocational college. Manual metal arc welding process. Movement of the welding rod. Burn through defect.
Incomplete fusion (infusion) defect. Machine learning process. Comparing Artificial Intelligence related terms. Neural network structure.
Convolutional neural network example. Image Classification Example. Vision Transformer model overview. Object Detection On Image.
YOLO models comparison. Large language models operation. Retrieval-Augmented Generation operation. Client-Server model.
REST API operation. Cloud Run service. Camera mounting solution. Weld placement block.
Monitor setup options. Using right-angle brackets and bolts to assemble the model frame. Metal covers for machines. Plastic cover for mini fan.
Composite mudguards for cars. Aluminum composite panel’s structure. The cover panels for the light-blocking chamber of the model. The cover for the electrical cabinet.
Completed mechanical system design. Completed mechanical system fabrication. Block diagram of electrical device connection. Electronics-control system fabrication.
Jetson Nano Developer kit. GigE Grasshopper 2 camera. Miniature circuit breaker. AC-DC Power Supply.
Buck DC-DC Converter. Switches & indicator lights. B class weld with some spatters. C class weld with multiple defects.
D class weld which is burned through. D class weld which has not form a weld. Burn through sample. Weld Quality Class Balance.
Testing classify model on real weld image. Classification model accuracy. Testing detection model on real weld image. F1-Confidence curve of defect detection model.
Precision-Recall curve of defect detection model. The data transmission and processing workflow. LED indicator when power on. Correct weld position.
Wrong weld position. Displaying results on GUI. 82 x LIST OF ABBREVIATIONS AC Alternating Current ACP Aluminum Composite Panel AI Artificial Intelligence ANN Artificial Neural Network AP Average Precision API Application Programming Interface CNN Convolutional Neural Network DC Direct Current DL Deep Learning GMAW Gas Metal Arc Welding GUI Graphical User Interface LCD Liquid Crystal Display LLMs Large Language Models MCB Miniature Circuit Breaker MIG Metal Inert Gas ML Machine Learning MMAW Manual Metal Arc Welding NIMs Nvidia Inference Microservices NLP Natural Language Processing RAG Retrieval-Augmented Generation RNN Recurrent Neural Network SDK Software Development Kit SVM Support Vector Machine ViT Vision Transformer YOLO You Only Look Once xi CHAPTER 1. Thesis background In the field of welding, particularly manual metal arc welding (MMAW), the quality of welds is a critical determinant of the structural integrity and durability of fabricated products.
Ensuring high-quality welds requires a combination of skill, experience, and meticulous attention to detail. However, traditional methods of welding training often rely heavily on subjective assessments by instructors, which can lead to inconsistent evaluations and a steep learning curve for students. This subjectivity can result in inadequate feedback, hindering students' ability to identify and correct mistakes effectively. Moreover, with the advancement of manufacturing technologies and the increasing complexity of welding tasks, there is a growing demand for more precise and reliable methods of weld quality assessment.
Automation and digitalization in welding processes have the potential to enhance training methodologies, making them more efficient and objective. Despite the potential benefits, there is a lack of comprehensive systems specifically designed for welding training that utilize these technologies. Existing solutions are often tailored for industrial quality control rather than educational purposes. Therefore, there is a need to develop a specialized system that can capture and process weld bead images from metal manual arc welding, classify the weld quality, and provide actionable feedback to students in a training environment.
Metal welding class at vocational college (source: [1]) 1 Additionally, the integration of image processing and machine learning technologies offers a promising solution to this challenge. By capturing images of weld beads in metal manual arc welding and analyzing them using advanced algorithms, it is possible to classify weld quality with a high degree of accuracy. This approach not only provides consistent and objective assessments but also facilitates immediate feedback, enabling students to learn and improve their welding skills more effectively.