AI supported Producibility Analysis
The aim of this pilot project is the application of artificial intelligence to optimize the set-up process in an automated manufacturing process.
This production line in question is the fully automatic profile wrapping machine RoboWrap, in which robots have replaced the manual positioning of the wrapping rolls, which press the decorative material wetted with adhesive onto the profiles. It is the only profile wrapping machine worldwide that provides the process data needed to derive algorithms for an AI-based system.
Its development began in the early 2000s as a cooperation between düspohl Maschinenbau GmbH and the universities of Bielefeld and Paderborn, the Fraunhofer IEM and Mitsubishi Electric as supplier of the articulated arm robots. According to the current state of the art, the wrapping robot performs the positioning itself during the initial adjustment of the rollers to a profile geometry, using an intuitive interface on a tablet or PC: it teaches it. The combination that produces the optimum wrapping result is saved at the end. If a batch of the same profile geometry is produced again at a later date, it is sufficient to call up the stored recipe. The robots position the pressure rollers automatically, even taking into account the wear marks on the rollers.
The newly launched project within the context of the AI Marketplace is intended to complete the automation and replace the manual teach-in process. Furthermore, it shall be possible to automatically assess the expected producibility of new product specifications. In the first step, the manufacturing system, the set-up process and the availability of production data will be analysed. Subsequently, the software modules and interfaces to be developed are determined. Here, various methods of artificial intelligence are combined to form an overall concept. In a third step, modules for feature extraction, producibility analysis and process parameter prediction are developed and validated. Finally, the project will analyze how the findings can be generalized for the AI Marketplace and thus transferred to other use cases in order to derive a generic application for the platform.