AWS re:Invent 2025 - Revolutionizing Audi's Welding Inspection System through AI (IND367)

Revolutionizing Audi's Welding Inspection through AI

Audi Company Overview

• Audi is the leading brand within the Volkswagen Group, which also includes Lamborghini, Ducati, and Bentley
• In 2024, Audi delivered over 1.6 million cars to customers, with Lamborghini, Bentley, and Ducati delivering over 10,000, 10,000, and 50,000 units respectively
• Audi has over 88,000 employees worldwide and reported a profit of 3.9 billion euros, a return on sales of 6%, and a net cash flow of 3.1 billion euros
• Audi's main production sites are located globally, including in Germany, Slovakia, Brazil, Argentina, and Mexico

Challenges in Manufacturing

• Top 500 manufacturers lose $1.4 trillion annually due to inefficiencies caused by production bottlenecks
• The cost of one hour of unplanned downtime has doubled from 2019 to 2024, indicating increasing complexity and production costs
• Manufacturers are under pressure to deliver products faster (from years to months or weeks), provide unique value, and create smart, intelligent products

How AI Can Help

• Synthesize information across departments to leverage centralized data
• Automate routine tasks and workflows to increase efficiency
• Enable data-driven decision making
• Preserve and centralize knowledge from experts

Volkswagen Group's Digital Production Platform (DPP)

• Volkswagen Group faced challenges with increasing complexity, decreasing efficiency, and increasing production costs, as well as a fragmented landscape across their 120+ factories
• The vision for the DPP was to create a centralized platform with centralized data management, connecting all factories to the cloud and building smart use cases
• This led to decreased complexity, increased efficiency, decreased production costs, and harmonized the IT landscape across the factories
• Audi and AWS have been collaborating on the DPP for 5 years, developing joint principles, exchanging on cultures, and building smart technologies

Use Case 1: Resistance Spot Welding Analytics

• Audi's body shop has approximately 1,000 cars per day, 1,200 robots, 800 welding guns, and 5.5 million welding spots per vehicle
• Previously, quality assurance checked 10,000 points per day on random samples, potentially missing defects
• The new AI-based solution checks all 5 million welding spots, providing a dynamic, real-time quality dashboard
• The architecture includes:
  • Data collection from welding controllers to an MQTT gateway
  • Streaming and batching of data to the cloud for storage in a data hub
  • Machine learning models trained on the welding data and ultrasonic check results
  • Inference models deployed in the cloud to provide real-time quality insights
• This solution increased the quality assurance coverage from 10,000 points per day to 5 million points, enabling proactive detection and correction of issues.

Use Case 2: Weld Splatter Detection

• Weld splatters can cause issues like damage to wiring, risk of injury to employees, and corrosion on the car body
• Previously, employees manually inspected and ground down weld splatters in a 20-second window, a labor-intensive and error-prone process
• The new AI-based solution uses computer vision to detect weld splatters in real-time, guiding employees or robots to the specific areas that need cleaning
• The architecture includes:
  • Cameras capturing high-resolution images of the car body
  • An edge gateway running the AI inference model to identify weld splatters
  • Integration with the PLC to provide real-time guidance to employees or robots
• This solution reduced the time and effort required to inspect and clean weld splatters, improving efficiency and quality.

Key Learnings

• Data quality is fundamental for successful AI use cases, requiring significant effort to understand and standardize the data
• Start small and iterate quickly, rather than trying to solve everything at once
• Adopt a composable, scalable architecture that allows for continuous improvement and refactoring
• Bridge the gap between IT and operational technology (OT) teams, as they often speak different languages
• Establish a central data management platform to enable reuse and sharing of data across multiple use cases

Business Impact

• Increased quality assurance coverage from 10,000 points per day to 5 million points, enabling proactive detection and correction of issues
• Reduced the time and effort required to inspect and clean weld splatters, improving efficiency and quality
• Enabled the development of smart, AI-powered solutions that enhance productivity, quality, and safety in Audi's manufacturing operations

Conclusion

Audi, in collaboration with AWS, has successfully implemented two AI-powered use cases to revolutionize its welding inspection and weld splatter detection processes. These solutions have demonstrated the power of leveraging data and AI to drive efficiency, quality, and safety improvements in the manufacturing environment. The key learnings and technical details provided offer valuable insights for other manufacturers looking to embark on their own digital transformation journeys.