How Precision Manufacturing Gets Done

With the help of edge analytics and machine learning, Audi is transforming manufacturing.

At a glance:

  • Intel is taking analytics to the edge and helping Audi automate and enhance critical quality control processes in its factories.

  • By creating a data-driven platform solution, Audi can reduce human error and ensure all cars are built with even more accuracy and precision.



Applying Digital Technology to Manufacturing

For Audi, the secret to success is a commitment to using cutting-edge technology to build high-quality vehicles that deliver precision engineering, exceptional performance, and luxury. Audi auto manufacturing is very advanced with many production jobs, from spot welding to riveting, fully automated. But their ultimate goal is to create smart factories and achieve an Industry 4.0 level of production. To accomplish that goal, the Audi engineers would need to move beyond the traditional approach of creating customized hardware and software solutions to handle individual use cases. Instead, they would need a scalable and flexible platform that would utilize advanced digital capabilities such as data analytics, machine learning, and edge computing.

“If you look at factories today, Audi’s auto manufacturing operation is very advanced and extremely sophisticated,” says Christine Boles, Vice President of the Internet of Things Group and General Manager of the Industrial Solutions Division at Intel. “But custom-made use cases are difficult to maintain and scale, and they can actually hinder innovation because of the time and money required to get the necessary approvals and deploy individual solutions. Audi was ready to look at things in a new way and try a different approach.”

Improving Quality Through Inline Inspection

Audi partnered with Intel on a proof of concept experiment focused on improving the quality control process for the welds on its vehicles. The POC took place at Audi’s factory in Neckarsulm, Germany, one of the company’s two principal assembly plants.

The Neckarsulm plant has 2,500 autonomous robots on its production line. Each robot is equipped with a tool, from glue guns to screwdrivers, and performs a specific task required to assemble an Audi automobile. Nine hundred of those robots carry welding guns to do spot welds that hold pieces of metal together. The production line is organized into a series of cells and vehicles that are being assembled and move down the line from cell to cell. Each cell can contain up to 20 robots and several milling machines. The milling machines are used to clean on the welding guns, as needed, between operations.

Audi assembles up to approximately 1,000 vehicles every day at the Neckarsulm factory, and there are 5,000 welds in each car—which equates to more than 5 million welds in a single day of production. To ensure the quality of its welds, Audi performs manual quality control inspections using the industry’s standard sampling method. “Audi pulls one car off the line each day and takes it to a large room, where 18 engineers with clipboards use ultrasound probes to test the welding spots and record the quality of every spot,” says Rita Wouhaybi, Principal Engineer for the Internet of Things Group in the Industrial Solutions Division at Intel and lead architect for Intel’s Industrial Edge Insights software.

Sampling is costly and labor-intensive, and the process was leaving too many unanswered questions about the quality of the other 999 cars produced each day. Unfortunately, Audi had no feasible and cost-effective way to test the quality of those other welds. “Our big goal for this solution is to make it possible for us to inspect 100% of our welds with a very high degree of accuracy,” says Mathias Mayer, who leads automation technology planning at Audi. “Right now, we don’t have that kind of assurance. We inspect one finished car at the end of the production line. We have no in-line inspection process. Intel has both the technology and the expertise to help us improve our processes and achieve our goals.”

Creating an Edge Solution That’s Scalable

Together with Audi, Intel created algorithms for streaming analytics using Intel’s Industrial Edge Insights software. The algorithms resulted in predictive analytics and modeling that transformed factory data into valuable insights. The solution absorbs data from the welding-gun controllers and analyzes it at the edge.

Intel’s data scientists created a machine-learning algorithm and trained it for accuracy by comparing the predictions it generated to actual inspection data that Audi provided. The model uses data generated by the welding controllers, which showed electric voltage and current curves during the welding operation. The data also includes other parameters such as the configuration of the welds, the types of metal, and the health of the electrodes. A dashboard lets Audi employees visualize the data, and the system alerts technicians whenever it detects a faulty weld or a potential change in the configuration that could minimize or eliminate the faults altogether.

Figure 1. Audi’s solution ingests data from the welding-gun controllers and analyzes it at the edge. A machine-learning algorithm was trained to detect faulty welds. Audi’s solution ran an Intel® Xeon® processor, but it can scale from Intel® Core™ processors all the way to Intel® Xeon® processor E and Intel® Xeon® scalable processors with no changes to the software.

Optimizations on the factory floor can go beyond one process to the rest of the factory. Audi can use this platform solution for other use cases involving robots and controllers such as riveting, gluing and painting. “The value of putting the analytics platform at the edge is that it allows you to draw more data into it and look at correlations, causalities, and other interesting analytics—even some you might not think of at first,” says Brian McCarson, Vice President of the Internet of Things Group and Director of Industrial Systems Engineering and Architecture at Intel. “This platform gives Audi a lot of headroom. It’s not just for this one use case. After making the initial platform investment, Audi can grow into it and scale it across facilities and to other use cases.”

This solution is like a blueprint for future solutions. We have a lot of technologies in the factory, and this is a model we can use to create quality inspection solutions for those other technologies so that we don’t have to rely on manual inspections.”—Henning Löser, Senior Manager Audi Production Lab

Increasing Efficiency and Precision

Moving from manual inspections to an automated, data-driven process has allowed Audi to increase the scope and accuracy of its quality control processes. But there are other benefits that come along with that.

At the Neckarsulm factory, we are already seeing a 30%-50% reduction in labor costs.”—Michael Häffner, Head of Production Planning, Automation and Digitization Audi

Häffner emphasized that increasing automation and efficiency is not about replacing workers, but rather about giving them new knowledge and skills and creating new opportunities for them. It’s also a necessity, because many skilled factory workers are retiring and taking valuable knowledge with them. So automating some of those jobs and channeling younger employees in new directions is good for the business and good for workers.

Another key benefit of the new system and the precise inspections it enables is that Audi can be proactive and focus on avoiding problems rather than merely reacting to them. “Let’s say we do an overall inspection of 5,000 or more welds on one car a day, and maybe 95% of those welds are good and 5% are not,” Mathias Mayer says. “In the future, we can focus on the 5% because we know where they are in the factory, and we can take action much sooner.”

Looking to the Future

Having a transparent system that lets Audi understand and learn from the data their equipment is generating is inspiring them to consider new possibilities and providing additional benefits—some of them unexpected. “Because of the analytics we are now running and the increased visibility of our data, Audi lowered its corporate tax bill,” Häffner says. “In the past, we had to make a lot of assumptions, and our taxes were based on those assumptions. Now, the real data shows that our tax obligation is less, which is a significant cost savings.”

Audi already has plans to use the platform for other use cases at the Neckarsulm factory, and eventually intends to deploy predictive welding inspection and other solutions across all Volkswagen Group production facilities. “We are at the very beginning of collecting and analyzing our data,” Henning Löser says. “As we continue this journey, there will be many more happy surprises and new opportunities for us.”

Building more precise cars, sustainably is wonderful—and Intel, with partners like Audi, is how wonderful gets done.

Download the PDF ›