AI-supported manufacturability analysis

Challenge

The initial set-up process for profile wrapping systems is still carried out by the wrapper.

Solution

Self-learning AI cycle around the digital twin of the machine, which always gives recommendations for the optimal machine setting.

Added value

Recommendations on the manufacturability of new products and on parameters of future coating processes.

Düspohl is considered the world's most innovative company in the development and manufacture of profile wrapping technologies, as well as surface lamination and peripheral systems for the wood and plastics industry. Because the wrapping of wood, metal or plastic profiles using so-called pressure rollers is done fully automatically and intelligently at Düspohl Maschinenbau GmbH with the help of the "RoboWrap". In the AI marketplace, the setup process of the production plant is now to be optimized and recommendations for the manufacturability of products are to be determined using a digital twin.

Profile wrapping is a process by which a decorative surface is laminated to a substrate. The wrapping takes place on a profile wrapping machine using pressure rollers. The RoboWrap system positions them fully automatically. Currently, an employee still carries out the positioning himself when first setting the pressure rollers to a profile geometry: he “teaches” them. The combination with which the optimal wrapping result is produced is saved at the end and can be called up again at a later point in time. The robots then automatically reproduce the positions of the pinch rollers.

Automated feature extraction and machine training thanks to AI

As part of the AI marketplace, experts from Fraunhofer IEM are working together with Düspohl to complete the automation and replace the previously non-automated teach-in. In addition, the manufacturability of new product specifications should be able to be assessed automatically. For this purpose, an algorithm for the extraction of features is first developed, with the help of which all types of profiles at Düspohl can be examined for their properties. In the next step, these characteristics are assigned to individual robots of the RoboWrap. In the third step, it is determined for each robot which roller geometry can be used to process its assigned feature in the best possible way and where the roller must be positioned exactly.

With the help of a self-learning AI cycle around the digital twin of the machine, Düspohl will always receive recommendations for optimal machine settings in the future. At the same time, Düspohl's customers receive recommendations on the manufacturability of new products and on the parameters of future coating processes. For them, this means not only process optimization but also a further increase in efficiency. The findings from the project are generalized for the AI marketplace in order to derive an application for the platform.