How quality control in vehicle manufacturing is changing
In car factories, quality is no longer just inspected, but increasingly safeguarded in real time. Cameras, AI, radar, lasers and digital inspection systems are intervening in production ever earlier. This is transforming not only final inspection, but the entire production logic.
Of course, the days of purely manual inspection after final assembly are long gone. The value of visual and data-based quality control has been increasing for years. Recently, however, the momentum has grown noticeably once again. As model variety, new drive systems, the share of software and the complexity of production processes all increase, the number of concrete use cases in the plants is rising rapidly. Camera systems, AI-supported inspections, in-line measuring technology and digital quality models are therefore no longer entering the factory only at isolated points, but are increasingly becoming part of everyday industrial series production.
There is a high level of ambition behind this. Errors are not only to be documented, but ideally detected where they occur. It is striking that the technological reality in the factories looks far more differentiated than some buzzwords suggest. Large-scale AI is not yet in use everywhere. In many cases it is robust, camera-based inspection methods, classical image processing, radar and laser measurements or data-supported decision logics that deliver the greatest leverage. It is precisely this blend of pragmatism and technological concentration that currently makes the topic so interesting.
Quality moves directly into the process
The extent to which quality logic is changing can be seen in the body shop at the Volkswagen plant in Bratislava, where Porsche is preparing the electric Cayenne. There, cameras are integrated directly into the equipment. They measure geometries, check tolerances and assess defined measuring points during the ongoing process. As a result, not only individual random samples are inspected, but every body-in-white. Target and actual values are compared in real time, deviations are documented and evaluated at downstream stations.
What is crucial here is not just the measuring technology itself, but its integration into a closed quality chain. Errors are intended to be detected exactly where they arise and not only at the later finish stage or even after the vehicle has left the line. Bratislava thus provides a prime example of how far quality assurance is moving away from classic end-of-line thinking and becoming a process-integrated discipline.
The fact that this requires not only sensors but also stable environmental conditions is illustrated by an inconspicuous detail. At Porsche, a shielded system, internally called Green Pigskin, protects certain mixed welding processes from interference. Solutions like this are not a side issue, but a prerequisite for camera-based measurements to function reliably at industrial cycle times.
Between AI and robust image processing
The new role of visual quality inspection becomes even more tangible at the Opel plant in Rüsselsheim. There, a young employee has integrated a cobot into an inspection station ahead of the so‑called marriage. It is precisely at this point that the body‑in‑white and the pre‑assembled chassis are brought together. The cobot moves to defined points on the body, holds a camera to the relevant areas and checks whether safety‑critical or assembly‑critical features have been implemented correctly.
The technological decision is particularly noteworthy. Instead of complex AI, the system deliberately relies on classical image processing with contrast and edge detection. This is exactly what makes the use case so interesting for many plants. In industrial reality, the primary concern is not the most spectacular algorithm, but whether a system is robust, flexible and can be deployed in line with cycle times.
In Rüsselsheim, the line stops automatically when safety‑relevant defects are detected. Other deviations are documented for each individual vehicle so that rework can be carried out in a targeted manner. The example illustrates very clearly that advanced inspection in the factory often does not appear as a gigantic transformation programme, but as a precise solution at a critical point. This approach can be particularly effective in brownfield environments with limited investment capacity.
AI helps where model variety demands precision
Audi is setting a different emphasis at its main Ingolstadt plant. There, an AI‑supported system is used in door assembly to measure bodies and doors to ensure the best possible fits. Applications like this show where AI in production is currently particularly convincing: wherever a high variety of models, complex geometries and tight tolerances come together.
This is because the quality of doors, tailgates or panel gaps not only determines function, but also has a strong influence on the perceived value of a vehicle. AI‑supported systems can help here by recognising patterns more quickly, assigning components more precisely and identifying fit issues earlier. In Ingolstadt this is supplemented by digital inspection systems that are intended to ensure that individual assembly steps are carried out correctly.
Battery assembly also shows how significantly quality inspection is gaining in importance today. Leak tightness, dielectric strength and insulation are systematically tested and every battery passes through defined test steps. Quality is therefore not understood only as a visual category, but as the sum of dimensional accuracy, function, safety and traceability.
A seamless digital quality system
The quality architecture at Xpeng appears particularly tightly integrated. The plant in Guangzhou was conceived as a smart factory from the outset, and this is clearly reflected in the inspection system. In body‑in‑white, a high‑precision radar system monitors key joining processes with a tolerance of 0.05 millimetres. During installation of the panoramic roof, a robot scans the component using a laser sight and positions it with an accuracy of 0.1 millimetres. Only then does a second robot apply the adhesive and carry out the assembly.
This is complemented by an exceptionally thorough final inspection. According to the company, Xpeng checks 587 points on the exterior, interior and dimensions before the vehicle is subjected to an intensive rain test. This is followed by a further static inspection with 572 checkpoints before the car is tested on different road types. This is supplemented by close data integration, for example during battery installation, where the parameters are linked to the vehicle identity and stored centrally.
This is precisely where it becomes clear where the topic is heading. Quality is no longer just an inspection step, but a data system. Sensor technology, assembly parameters, test results and traceability are merging into a unified quality architecture. Xpeng thus presents the counter-model to the pragmatically retrofitted brownfield plant. It is a factory in which quality control is conceived from the outset as digital infrastructure.
Data as an increasingly important basis for decisions
Not every advance in quality control is immediately visible. At the VW Palmela plant it becomes clear how strongly data-driven systems can change inspection strategies. There, 100% of vehicles are checked at the end of each line for functional dimensions relevant to final assembly. The data are available online, so operators can identify deviations immediately and carry out corrections.
In addition, numerous camera systems act as poka-yoke solutions to help prevent errors. Particularly interesting, however, is the so-called road test predictor. This system decides, on the basis of collected production data, which vehicles are sent on a longer or shorter test route. Quality is therefore not only measured here, but, in a sense, also predicted and prioritised. This points to a development that is only just beginning in many plants. In future, inspection intensity will be controlled to a much greater extent on a risk-based and data-driven basis.
Even in the design-to-build process, early precursors of this logic can already be identified. In Palmela, virtual models and 3D-printed components are used to detect potential assembly issues at an early stage. This is not yet a fully developed digital quality twin, but it clearly indicates the direction of travel. Quality assurance is increasingly beginning even before the physical start of series production.
The quality question is expanding
With the transition to software-defined and autonomous vehicles, the concept of quality is expanding further. At the Hannover plant, Volkswagen Commercial Vehicles is preparing the industrialisation of the ID. Buzz AD. There, the focus is no longer only on classic assembly quality, but also on the series-ready integration of cameras, radar and lidar systems, as well as high-performance computers. Once installation is complete, the sensors are calibrated and the vehicle is commissioned.
This is shifting the quality question. In future, it will no longer be sufficient to check whether a component has been correctly assembled. It will also be crucial to determine whether a sensor is precisely adjusted, whether the overall system interacts correctly, and whether the software-based function will later operate reliably. Quality assurance is therefore becoming more systemic. It no longer covers only material, fit and surface finish, but increasingly also calibration, data consistency and the functional capability of complex technical systems.
The real transformation is organisational
As different as the examples from Rüsselsheim, Bratislava, Ingolstadt, Palmela, Hanover or Guangzhou may be, they all point in the same direction. In the car plant, quality is shifting from downstream inspection into the running process. It is becoming more visual, more data based and, in many areas, also more intelligent. At the same time, practice shows that not every effective solution has to be based on highly complex AI. Often it is robust vision systems, cleanly linked data, clear inspection logic and rapid feedback into the process that achieve the greatest effect.
The real transformation is therefore not only technological but organisational. Plants must learn to make quality data usable in real time, to move inspection decisions closer to the line and to build new capabilities at the interface between production, automation and data analysis. This is precisely where the greatest leverage currently lies. The factory of the future does not simply inspect more. It inspects earlier, more selectively and with significantly more knowledge about each individual vehicle.
