How augmented reality is changing quality inspection in automotive production

With each new vehicle generation, the complexity of automotive production continues to increase. Platform strategies, a growing number of derivatives and shorter development cycles mean that production processes are constantly changing. At the same time, the pressure to detect errors as early as possible is increasing. After all, every deviation that is only discovered at the end of production leads to rework, costs time and, in the worst case, money.

As a result, quality assurance is increasingly caught between two worlds. On the one hand, there are highly precise measurement systems that reliably detect even the smallest tolerances. On the other hand, a large part of daily quality control is still carried out visually – supported by drawings, paper documents or digital inspection plans. But there is a gap between these two approaches. And this is precisely where Visometry sees the field of application for its augmented reality solution Twyn.

“It is not about replacing traditional measurement technology,” explains Florian Schmitt, Product Manager at Visometry. Rather, it is about providing a fast and intuitive way to compare the target state of a component with the real object. In many cases, this makes it possible to identify at an early stage whether components are missing, have been installed in the wrong position or whether production defects are present. For highly precise measurements in the micrometer range, coordinate measuring machines or laser measurement systems are still required. But many inspection tasks sit exactly between these two worlds. This is where augmented reality can deliver its greatest added value.

From Fraunhofer research to the shopfloor

The roots of Twyn go back more than a decade. Visometry as a company was founded as a spin-off from the Fraunhofer Institute, where today’s developers had already been working on industrial applications of augmented reality at an early stage. Together with industry partners from the automotive sector, the first prototypes were created, demonstrating how digital CAD data could be projected directly onto real components.

Over time, these projects gave rise to the idea of turning individual customer solutions into a standardized product. Instead of developing each project from scratch, the aim was to create a solution that companies could deploy as easily as possible: load CAD data, transfer it to a tablet and start quality inspection immediately – without complex infrastructure or lengthy implementation projects.

Another aspect played a decisive role from the very beginning. The software was not only intended to work for digital natives, but also for employees who had previously had little contact with augmented reality. “The system is intended to be an assistance tool that simplifies workflows rather than making them more complicated,” says Schmitt. For this reason, the most intuitive operation possible was a central development goal from the outset.

Why robust tracking determines practical success

For outsiders, augmented reality often appears spectacular. From the developers’ perspective, however, a comparatively inconspicuous function determines its actual usefulness in everyday production: tracking.

Only when the system precisely recognizes the position of a real component can digital information be overlaid on it with millimeter accuracy. This is based on an edge-based process. Characteristic contours are derived from the CAD data and then recognized in the camera image. As a result, the system always knows where the tablet is located in space and can project the digital target state precisely onto the real object.

“Tracking is the core of our product,” says Schmitt. Only if this overlay works reliably can users determine, for example, whether a bolt is missing or a component is positioned incorrectly. If tracking constantly breaks off or the user has to restart the process again and again, acceptance on the shopfloor quickly declines. This is precisely why Visometry has invested a large part of its development work in recent years in the stability of this function.

With the current version, initialization in particular has been significantly simplified. While users previously had to align the virtual model relatively precisely, an intelligent initialization process now supports them when getting started. This considerably reduces the time required and makes it easier for less experienced users in particular to begin working with the system.

The visualization itself also plays an important role. The more precisely the CAD model and the real component are aligned, the easier it is to identify deviations. According to the company, high-quality CAD rendering is therefore, alongside tracking, one of the areas in which it deliberately seeks to differentiate itself from competitors.

Particularly interesting for pre-series production and complex assembly processes

The benefit of such a solution depends heavily on the respective field of application. In highly automated series production lines, many inspection tasks are already handled by specialized measurement systems. The situation is different where new vehicle projects are starting up or processes have not yet been fully standardized.

Schmitt therefore sees pre-series production in particular as an important application area. In many cases, fully defined inspection procedures do not yet exist there. At the same time, the first vehicles already have to be checked for correct assembly and production quality. In such cases, Twyn makes it possible to transfer existing CAD data into digital inspection processes with comparatively little effort.

In addition, companies have very different data structures. Some work with Excel lists, others with PLM or PDM systems, while others already have structured quality data. The software was therefore developed in such a way that it can handle different data sources and supplement existing processes as easily as possible, rather than replacing them completely.

Typical areas of application today include body-in-white inspections, assembly checks, incoming goods inspections and the digital inspection of joining elements. In many cases, conventional templates can also be replaced by virtual overlays. This not only saves costs for complex physical fixtures, but also increases flexibility when products are changed.

Twyn explicitly sees itself as a complement to existing quality systems. “We are not competing with measurement technology,” Schmitt emphasizes. Rather, the software supplements existing inspection processes in areas where inspections were previously often not carried out at all or only with considerable effort. As a result, errors are detected earlier, rework is reduced and the overall efficiency of the quality process is improved.

AI Is intended to support, not decide

Like many software companies, Visometry is also intensively examining artificial intelligence. However, Schmitt observes that the term is often used in a very broad and general way. Not every new function is automatically AI, and not every problem can be sensibly solved with AI.

Nevertheless, the company sees great potential. In the future, image-based machine learning methods could help automatically detect small joining elements, bolts or weld points, for example. It is precisely with such components that traditional edge-based methods reach their limits, because distinctive geometries are missing or the appearance changes due to wear and contamination.

Machine learning could provide additional robustness here. At the same time, Visometry is deliberately pursuing a cautious approach. AI is intended to support workers, not replace them. The final decision on the quality of a component remains with the human operator. “An assistance system must not take over responsibility,” says Schmitt. The software provides guidance and accelerates the inspection process. However, responsibility for the result remains with the user.

This attitude is also reflected in further product development. In addition to AI, the company is working on stronger integration into existing PLM, MES and quality management systems. The goal is to create the most seamless digital quality chain possible, in which inspection data can be exchanged between different systems without media disruptions.

For Visometry, augmented reality has therefore long since become more than just a technical demonstration object. From the company’s perspective, the technology is increasingly developing into a new user interface for digital product data. Just as tablets have replaced paper documents in many areas, AR-based assistance systems could play a similar role in industrial quality inspection in the future. Not as a replacement for existing measurement technology, but as a link between digital development data and practical work on the shopfloor. This is precisely where the company sees the greatest potential for the coming years.