ARC EIF 2023 Review - AI & Machine Learning

ARC’s European Industry Forum, part of the successful series of worldwide conferences in Europe, America and Asia, has been held in Sitges (Barcelona), Spain on May 15 - 17, 2023. The event offered exclusive presentations and workshops on strategies and case studies in line with this year’s topic “Managing Industry Evolution in Times of Global Disruption” to its 160 international participants from 20 countries.

AI in Deterministic Real-Time Control

Participants of the Artificial Intelligence Workshop were: Dr. Daniel Klein, Siemens; Claus Botzenhardt, Multivac; Michael Hoff, ai.dopt; Viacheslav Gromov, AITAD and Paco Masip, Air Products. The target of the workshop was to look at the role of AI in deterministic real-time control in the future. Several factors were identified beforehand. These factors included technological factors such as the capability of AI models, the rise of embedded AI, the availability of high level data and legal ones such as labor laws. People factors were also taken into account, like the availability of workers, especially of AI specialists and last but not least economic and market factors such as the cost of AI implementation or the high competition between companies.

Before these factors were analyzed, the definition of real-time was discussed, with some interesting discrepancies. While Claus from Multivac was thinking in microseconds for their machinery, Paco from Air Products considered even hours as real time, as they do not require updates from their processes that often. The next point of discussion was the necessity of 100 percent accuracy in automation. In fact, there are many applications where closed loop AI control could be used without being that accurate, excluding those applications with safety and security requirements.

Availability of workers

AI will become more attractive in the plant and be used increasingly for deterministic real time control, to compensate for the growing lack of workers. But this lack will not just manifest on the plant floor, but also affect AI specialists, hence low code and no code programming for almost autonomous AI will be a key aspect.

Low code and no code programming.

This factor was identified as another drive for the adoption of AI. Yet the ease of making these models and creating output may create a false sense of security about the accuracy of the output. The conclusion was that one of the key areas will be large language models which can provide initial solutions, but whatever AI produces will require quality checks by people.

Availability of data

Data availability has improved steadily over the last years and will continue to do so. The field level will connect directly to the data lake and, but we will still need experts to map the data and create contextualized data. Progress on this is expected over the next years and has been identified as one of the key factors.

Scalability of models

One of the key challenges of AI has been the poor scalability of the models. The fact that one model can be used for one machine doesn't mean it will work for an any other machine or application. Over the next two to five years, more AI use cases will be brought to maturity by extended teams, increasing the relevant databases as well as the scalability and adoption of edged base AI models.

Embedded AI

One of the hottest topic currently embedded AI, i.e. running on a chip and not in a software. We expect embedded AI to expand into many aspects and down to the sensor level over the next two to four years. It will be found in autonomous vehicles but also be in production applications. We expect embedded AI to be a playing a key role in energy and sustainability while moving towards swarms of systems and more abstract models. The technology will probably be present in every aspect of production, but so deeply embedded into the systems that users are not even aware of it.

Summary

All these factors point into the direction that AI in deterministic real time control will become reality over the next 5+ years, moving us closer to autonomous production Humans will be supervising and coordinating production rather than doing repetitive tasks and surveilling standard processes.

ARC’s European Industry Forum & Platform 

We like to thank our sponsors, who supported the ARC EIF, and all our  speakers and presenters, who made the event interesting, exciting and memorable. To complement the Forum, ARC Europe offers its online platform accessible 365 days a year. Its goal is to maintain the flow of information, and to foster network opportunities. On the platform you will find information concerning relevant economic developments and information on automation markets, as well as our quarterly industry webinars. 

Materials and presentations from the Forum will be made available on our platform https://arc-industry-forum-europe.arcweb.com/en/ after registration, depending on the package chosen. Forum attendees can download presentation slides, view video recordings and receive further materials on the platform.

For more information, please contact Ann-Kathrin Blech (mailto:ablech@arcweb.com).