RECAST: AI-Powered Quality Assurance for Welding Processes

Overview

RECAST (ÜbeRwachung dEr SChweißquAlität durch künStliche InTelligenz) is a research initiative funded by the Federal Ministry of Education and Research (BMBF). The project focuses on developing an intuitive and adaptable AI framework for process monitoring and quality assurance, with a primary application in welding processes for thermoplastic composite materials.

Objectives

The main goals of the RECAST project are:

1. Automating Manual Processes: Replace manual steps in production chains with AI-driven solutions.
2. Quality Assurance and Process Optimization: Enable non-destructive, automated quality control for welding processes.
3. Open-Source Development: Deliver a freely available framework to encourage broad adoption in small and medium-sized enterprises (SMEs).
4. Ease of Integration: Allow SMEs to utilize AI technologies without prior expertise.

Key Features

• End-to-End Process Monitoring: Utilize mobile and embedded devices for seamless data aggregation throughout the production lifecycle.
• AI Integration: Leverage AutoML and Transfer Learning to develop adaptable AI models for real-time decision-making.
• Low-Code Accessibility: Provide user-friendly tools for configuring processes and analyzing data via a web-based interface.
• Open Source Framework: Share the framework and related datasets to facilitate experimentation and knowledge sharing.

Use Cases

The framework is being validated on welding processes such as:

• Continuous ultrasonic welding.
• Continuous resistance welding.

Project Milestones

1. Sensor Data Collection and Process Visualization: Establish robust data acquisition and visualization systems.
2. AI-Assisted Quality Control: Develop models for detecting defects in welded components.
3. Automated Model Creation: Implement AutoML for building AI models directly on production equipment.
4. Adaptive Process Optimization: Use AI to optimize parameters for varying materials and forms.
5. Transfer Learning: Extend findings to additional processes, such as resistance welding.

Publications

The RECAST team has contributed to advancing knowledge in the field of AI and welding through the following publications:

1. "RECAST: An Open-Source Digital Twin Framework for Industrial Production Environments"
   Authors: Lars Larsen, Thomas Fraunholz, Tim Köhler, Dennis Rall, Veronika Langner, Dominik Görick & Alfons Schuster
   Journal: FAIM2024
   DOI: 10.1007/978-3-031-74482-2_19

2. "RECAST: An Open Source Platform for Item-Specific Capturing of Real Production Processes"
   Authors: Tim Köhler, Thomas Fraunholz, Dennis Rall, Lars Larsen, Dominik Görick, Alfons Schuster
   Journal: ICCBDC 2024
   DOI: 10.1145/3694860

3. "A Comparative Study of Open Source Computer Vision Models for Application on Small Data: The Case of CFRP Tape Laying"
   Authors: Thomas Fraunholz, Dennis Rall, Tim Köhler, Alfons Schuster, Monika Mayer, Lars Larsen
   DOI: 10.48550/arXiv.2409.10104

Funding

• Duration: 3 years

How to Contribute

We welcome contributions from the community. To get involved:

1. Check out the repository.
2. Review our contribution guidelines.
3. Join discussions in the issues section.

License

This project is licensed under the MIT License.

Contact

For more information, contact Tim Köhler at [email protected].