Lecture
Physical AI as part of real-time Automation Processes
- at -
- Hall C6C6.501
- Language: English
- Type: Lecture
Lecture description
From Process Expertise to AI-Driven Automation Systems** Artificial Intelligence is transforming industrial automation, but successful deployment requires more than data scientists and AI frameworks. The real challenge is combining deep process expertise with technologies that enable experimental AI solutions to become robust, maintainable, and real-time capable industrial systems. This presentation demonstrates how domain knowledge from process and mechanical engineering can be systematically transformed into machine learning applications and seamlessly integrated into professional automation architectures. Using a laser welding process as a practical example, the talk shows how process expertise is used to derive training data and develop both numerical and vision-based machine learning models. The presentation further explains how these models can be deployed on a real-time capable aggregation platform, where AI functions operate alongside traditional control and automation logic. By combining IoT, AI, and OT technologies on a unified Linux-based platform and utilizing available CPU, GPU, and NPU resources, entirely new classes of automation solutions become possible. Attendees will gain insight into the complete development workflow—from process understanding and model creation to industrial deployment. The talk also highlights the technologies, architectures, and engineering practices required to make AI accessible to automation engineers, enabling them not only to use AI-based systems but also to develop and maintain them independently. The result is a practical roadmap for turning process knowledge into intelligent, deterministic, and production-ready automation systems.
