During the manufacturing process of a car, the operations carried out in the paint shop are the most delicate steps. Along the process, quality controls are performed, so the process is controlled, and adjustments are made because, for example, each layer thickness variations or repetitive appearance of defects. These reworks are often required and can significantly increase manufacturing costs.

Historically, to detect and correct these defects, inspection zones have been implemented. Here, specialized personnel carefully examine the surface of the body. However, the weak point of this method is the consistency of detection since human vision differs between individuals and, in addition, fatigue.

Due to the advancement of artificial vision technology, these defect detection activities have been automated in such a way that they allow for constant evaluation criteria. What Mercedes has done is to use the data provided by an automatic fault detection process that has been applied to the entire production to develop a predictive model that allows improving the efficiency of processes.

By developing this model in the workshop, operators can now see defects in real-time and predict when these defects will appear in the paint film layer, being able to identify and correct the factors that are leading to their appearance of them and preventing them. At the end of the project, the system could capture this large amount of data and correlate them according to the developed model to predict quality problems before they happen.