© Hengst SE
Where is the fault? Artificial intelligence ensures production quality
Cooperation partner: NFT automates GmbH and Hengst SE
Manufacturers in industry are facing ever greater challenges: Production should run faster and faster, but the production steps should still be traceable. To meet these requirements, automatic and reliable controls of processes and work steps are necessary. Thanks to artificial intelligence, a company from Münsterland has succeeded in enabling automatic controls that guarantee higher production quality.
NFT automates GmbH, Ibbenbüren
NFT automates GmbH from Ibbenbüren has set itself the goal of improving the industrial production of its customers through automation and digitalisation at all levels. NFT automates GmbH advises, plans and designs Production 4.0.
Stallion SE, Münster
Challenge: Automatic visual inspection
In manufacturing, quality is one of the most important features. Every production step must fit, every component must be checked. However, producers often discover errors only at the end, when the product has already been manufactured. Accordingly, the effort to rectify the error is high. Because the earlier defects are detected, the fewer production steps have to be undone.
Hengst SE produces oil filter housings in which several differently shaped seals are inserted. Automatically acquired images, which are also automatically evaluated (automatic visual inspection), are used to ensure that the seals have been correctly inserted into the recesses. If the gaskets protrude, it can happen in the production process that they slip in further assembly steps, get jammed between the housing elements and the assembly leaks.
Because there are different seals and surfaces, classic approaches to visual inspection meet their limits here. That's why the use of artificial intelligence (AI) is an obvious choice. But when using AI solutions, it is difficult to understand the AI's decisions. AI is therefore often a black box with opaque processes. Together, the cooperation partners Hengst SE and NFT automates GmbH therefore developed a solution for this problem.
Solution approach: AI-supported object recognition
The innovative AI-based application checks whether the different seals are correctly positioned in their recesses. The special feature here is the error display. The machine operator is shown via a web application exactly where the error is located. With this knowledge, the operator can rework the faulty seal so that the assembly does not leak later in the process. Such a quality control would be far too time-consuming manually. The AI-supported object recognition works via a web application that can be displayed in any browser.
Another advantage of the application: it works both day and night. The AI can be trained with different lighting conditions so that the visual inspection works in direct as well as indirect lighting.
Success: Use of state-of-the-art technology for quality assurance
The application is optimal for the manufacturing industry: by using AI in automatic visual inspection, it enables consistent production quality. The innovative aspect is not only the use of artificial intelligence for quality control. What is special is that the AI's decisions are transparent and comprehensible, something that has not existed in this form in this context until now. In addition to the visual inspection, the camera images in which a defect was detected are archived. With the help of these images, the production process can be optimised even further.
As a regional joint initiative, Enabling Networks Münsterland supports companies and universities in the Münsterland region in developing innovations, implementing them and finding the right partners for the project. In order to show how innovative and at the same time cooperative Münsterland already is, the project also went in search of innovative cooperation projects from the region. The results are presented on this page. The Enabling Networks Münsterland project is funded by the European Union and the NRW Ministry of Economics as part of the ERDF call "Regio.NRW".
