At the half-time of the KEEN project, the 4th project meeting was planned as a hybrid event to promote the networking of the partners. However, due to the critical Corona situation, the meeting was changed to a virtual meeting on short notice.
Andreas Förster, Executive Director of DECHEMA e.V., regretted the cancellation and put the participants off to next year regarding a face-to-face meeting. In his welcoming, he gave a positive summary of the results so far. "The cooperation is based on trust, which is an important factor for KEEN's success." He also mentioned the technological hurdles that need to be overcome.
A special characteristic of the meeting was the presentations by the industry partners Air Liquide, Bayer, Covestro, Evonik, Merck, ABB, Krohne, Samson, CGC, DDBST, Leikon, INOSIM and X-Visual. The project results were put in their respective industrial contexts. The partners agreed that it is not about the development of AI, but about the transferability of the results. Real data from industrial processes was made available in order to test the methods, which presents a particular challenge in the development and implementation of AI.
Bayer is one of the five companies already using AI solutions in an industrial environment. The company participates in all subprojects and provides various use cases. For example, it supports the development of the ML-based matrix completion method for the prediction of substance data and tests its suitability regarding their internal processes.
As a technology provider, Krohne contributes expertise in the field of instrumentation. In KEEN, the company pursues, among other things, the goal of applying sensors in such a way that they acquire additional information, which are relevant for AI applications, in addition to the primary measured value and provide it in a way that is compatible with the information models used in KEEN.
Felix Riedl represented the company INOSIM, one of the five SMEs in the consortium. The company is working on the development of a Decision Support System for operators, managers and planners of industrial batch plants. The dashboard was demonstrated to the consortium and offered for testing.
After the lunch break, a virtual poster session was held. The KEEN PhD students explained their current results to the project participants on the basis of their posters. Another highlight of the day was the virtual plenary discussion about the three topics "ML Models for Online Applications", "The Cycle between Data and Models" and "Architecture". Lively discussions took place after respective keynote presentations. Leon Urbas, consortium leader of KEEN, commented that it was normal to make mistakes when applying ML. It would therefore be important to bravely move on.
Kai Dadhe, Evonik Operations GmbH