Turning Factories Into Smart Factories with AI
Artificial intelligence (AI) has been on everybody's lips since the triumphant launch of the ChatGPT chatbot. AI is also making great strides in industrial production technology. Machine learning can increase the efficiency of manufacturing. But just how does it work? Find out how at EMO Hannover 2023, from 18 to 23 September. Under the banner of "Innovate Manufacturing", the world's leading trade fair for production technology will be inspiring its trade audience by presenting plenty of fresh ideas, with artificial intelligence featuring prominently.
Can production machines really self-optimize? Can they learn from their mistakes? And is it possible for them to acquire know-how from other machines? Artificial intelligence (AI) makes all of this possible. When self-learning production machines function intelligently, this leads to greater productivity, lower costs, improved quality and reduced downtimes.
"We have spent a great deal of time on optimizing our production technology processes and have built up a competitive edge here. We now want to do the same in the digital transformation of industrial production," explains Markus Spiekermann, Head of the Data Economy Department at the Fraunhofer Institute for Software and Systems Engineering ISST. "Artificial intelligence is playing a decisive role in meeting the new requirements, " says Spiekermann. "Because only through the use of AI methods can high levels of automation be achieved."
Predictive Maintenance for Lathes
The AI trend is taking hold in industry. Machine tool manufacturer Weisser Söhne GmbH & Co. KG, for example, relies on AI models that enable predictive maintenance of its lathes.
"Predictive maintenance uses AI to forecast when a machine will nrequire servicing to prevent it from breaking down," explains Dr.-Ing. Robin Hirt, CEO and founder of Karlsruhe-based startup Prenode GmbH. Then software company helps machine builders equip their plants with customized AI-based features.
Modern production machines can self-optimize with the help of artificial intelligence, says Hirt. "They generally use so-called machine learning methods for this. These enable the machines to recognize patterns and correlations in the production data and automatically derive improvements from them." In many cases it is also possible for them to learn from their mistakes and adopt know-how from other machines.
Decentralized Data Used to Generate a Common AI model
The federated learning technique is often used, as the data obtained from a single lathe is often insufficient as the basis for an accurate AI model. Federated learning facilitates the "training" of a common AI model, with data stored in decentralized form but with no direct sharingn of data. The individual data therefore remains on the respective machines and does not have to be stored centrally in one place (such as in the machine manufacturer's cloud).
The AI models use ongoing lathe data to estimate the present status of the plant, and then forward this to the operating personnel. Deep learning neural networks are used for this.
Images: R. Eberhard, messekompakt.com, EBERHARD print & medien agentur gmbh
Source: Deutsche Messe AG